1 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
4 * config/i386/i386.c (construct_container): Use TImode for
5 BLKmode values in 2 integer registers.
7 2018-09-29 Jeff Law <law@redhat.com>
9 * builtins.c (unterminated_array): Pass in c_strlen_data * to
10 c_strlen rather than just a tree *.
11 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
12 Update recursive calls appropriately. If caller did not provide a
13 suitable data pointer, create a local one. When a non-terminated
14 string is discovered, bubble up information about the string via the
16 * builtins.h (c_strlen): Update prototype.
17 (c_strlen_data): New structure.
18 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
19 For a type 2 call, if c_strlen indicates a non-terminated string
20 use the length of the non-terminated string.
21 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
23 2018-09-29 Jakub Jelinek <jakub@redhat.com>
26 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
27 __m512d type for __A argument rather than __m512.
29 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
31 * match.pd (simple_comparison): Don't optimize if either operand is
32 a function pointer when target needs function pointer canonicalization.
34 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
36 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
37 power5 .. power9 to remove indirection.
38 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
39 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
40 ASM_CPU_476_SPEC): Delete.
41 (ASM_CPU_SPEC): Adjust.
42 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
43 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
45 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
47 * config.in: Delete HAVE_AS_DCI.
48 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
49 * config/rs6000/rs6000.h: Ditto.
50 * configure.ac: Delete HAVE_AS_DCI.
51 * configure: Regenerate.
53 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
55 * config.in (HAVE_AS_LWSYNC): Delete.
56 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
57 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
59 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
60 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
62 * configure.ac: Delete HAVE_AS_LWSYNC.
63 * configure: Regenerate.
65 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
66 Pierre-Marie de Rodat <derodat@adacore.com>
68 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
69 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
70 (cgraph_node::create_thunk): Add indirect_offset parameter.
71 (thunk_adjust): Likewise.
72 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
73 and initialize the corresponding field with it.
74 (cgraph_node::dump): Dump indirect_offset field.
75 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
76 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
77 (thunk_adjust): Add indirect_offset parameter and deal with it.
78 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
79 pass it to thunk_adjust. Do not call the target hook if it's non-zero
80 or if the thunk is external or local. Fix formatting. Do not chain
81 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
82 if any, in the GIMPLE representation.
83 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
84 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
85 (input_node): Read indirect_offset field.
86 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
88 * tree-nested.c (struct nesting_info): Add thunk_p field.
89 (create_nesting_tree): Set it.
90 (convert_all_function_calls): Copy static chain from targets to thunks.
91 (finalize_nesting_tree_1): Return early for thunks.
92 (unnest_nesting_tree_1): Do not finalize thunks.
93 (gimplify_all_functions): Do not gimplify thunks.
95 2018-09-28 David Malcolm <dmalcolm@redhat.com>
97 * opt-suggestions.c (option_proposer::build_option_suggestions):
98 Release "option_values".
100 2018-09-28 David Malcolm <dmalcolm@redhat.com>
102 * coverage.c (get_coverage_counts): Convert problem-reporting dump
103 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
104 * dumpfile.c (kind_as_string): New function.
105 (dump_loc): Rather than a hardcoded prefix of "note: ", use
106 kind_as_string to vary the prefix based on dump_kind.
107 (selftest::test_capture_of_dump_calls): Update for above.
109 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
111 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
112 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
114 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
116 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
117 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
118 INVALID_REGNUM instead of FPSR_REG.
119 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
120 * config/i386/i386.md: Update comment of FP compares.
121 (fldenv): Do not clobber FPSR_REG.
123 2018-09-28 Richard Biener <rguenther@suse.de>
125 * tree.h (BLOCK_ORIGIN): New.
126 * omp-expand.c (grid_expand_target_grid_body): Assign
127 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
128 * tree-inline.c (remap_block): Likewise.
129 * auto-profile.c (get_function_decl_from_block): Simplify
130 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
131 * langhooks.c (lhd_print_error_function): Likewise.
132 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
134 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
135 * tree.c (block_nonartificial_location): Likewise.
136 (block_ultimate_origin): Likewise.
137 * tree-pretty-print.c (percent_K_format): Likewise. Remove
138 no longer needed LTO case.
140 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
141 Jan Hubicka <jh@suse.cz>
142 Martin Jambor <mjambor@suse.cz>
144 * simplify-rtx.c (simplify_merge_mask): New function.
145 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
146 same masks are used in op1 or op2.
147 (test_vec_merge): New function.
148 (test_vector_ops): Call test_vec_merge.
150 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
152 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
153 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
154 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
155 (cypress_costs): Set it.
156 (supersparc_costs): Likewise.
157 (hypersparc_costs): Likewise.
158 (leon_cost): Likewise.
159 (leon3_costs): Likewise.
160 (sparclet_costs): Likewise.
161 (ultrasparc_costs): Likewise.
162 (ultrasparc_costs): Likewise.
163 (niagara_costs): Likewise.
164 (niagara2_costs): Likewise.
165 (niagara3_costs): Likewise.
166 (niagara4_costs): Likewise.
167 (niagara7_costs): Likewise.
168 (m8_costs): Likewise.
169 (TARGET_CAN_FOLLOW_JUMP): Define.
170 (pass_work_around_errata::gate): Minor tweak.
171 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
172 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
173 Automaitcally clear MASK_FSMULD mask for V7 processors.
174 (sparc_can_follow_jump): New static function.
175 (output_ubranch): Deal with CROSSING_JUMP_P.
176 (sparc_use_sched_lookahead): Rewrite using switch statement.
177 (sparc_issue_rate): Reorder.
178 (sparc_branch_cost): New function.
180 2018-09-27 Martin Sebor <msebor@redhat.com>
182 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
183 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
184 (int_fits_type_p): Same.
186 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
188 * config/i386/i386.md (FPCR_REG): Remove.
189 (UNSPEC_FLDCW): Remove.
190 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
191 (x86_fldcw_1): Remove insn pattern.
192 (fnstenv): Do not clobber FPCR_REG.
194 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
195 (CALL_USED_REGISTERS): Ditto.
196 (REG_ALLOC_ORDER): Ditto.
197 (REG_CLASS_CONTENTS): Ditto.
198 (HI_REGISTER_NAMES): Ditto.
199 (ADDITIONAL_REGISTER_NAMES): Use defines instead
200 of numerical constants.
201 * config/i386/i386.c (regclass_map): Remove fpsr register.
202 (dbx_register_map): Ditto.
203 (dbx64_register_map): Ditto.
204 (svr4_dbx_register_map): Ditto.
205 (print_reg): Do not handle FPCR_REG.
207 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
210 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
211 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
212 Delete, always treat as true.
213 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
214 Ditto. Simplify remaining code.
215 * config/powerpcspe/powerpcspe.h: Ditto.
216 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
217 Simplify remaining code.
218 (rs6000_expand_builtin): Ditto.
219 * config/rs6000/rs6000.h: Ditto.
220 * configure.ac: Ditto.
221 * configure: Regenerate.
223 2018-09-27 Martin Liska <mliska@suse.cz>
225 * coverage.c (get_coverage_counts): Revert the formatting
226 of missing profile opt info.
228 2018-09-27 Richard Biener <rguenther@suse.de>
232 * dwarf2out.c (set_block_origin_self): Do not mark outermost
233 block as we do not output that.
234 (gen_inlined_subroutine_die): Elide the originally outermost
235 block, matching what we do for concrete instances.
236 (decls_for_scope): Add parameter specifying whether to recurse
239 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
240 Tom de Vries <tom@codesourcery.com>
244 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
245 STORE_FLAG_VALUE == 1.
247 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
249 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
250 constant definitions.
251 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
252 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
253 ("speculation_barrier"): New expander definition.
255 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
257 PR gcov-profile/86957
258 * common.opt: New warning option -Wmissing-profile.
259 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
260 * doc/invoke.texi: Document -Wmissing-profile.
262 2018-09-26 Jim Wilson <jimw@sifive.com>
264 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
265 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
267 2018-09-26 Martin Sebor <msebor@redhat.com>
269 * tree.c (zerop): Change return type to bool.
270 (integer_zerop, integer_onep, integer_each_onep): Same.
271 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
272 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
273 (real_onep, real_minus_onep, chain_index): Same.
274 (print_type_hash_statistics, type_list_equal): Same.
275 * tree.h (zerop): Same.
276 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
277 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
278 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
279 (real_onep, real_minus_onep, chain_index): Same.
280 (print_type_hash_statistics, type_list_equal): Same.
282 2018-09-26 Jim Wilson <jimw@sifive.com>
284 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
286 2018-09-26 Jakub Jelinek <jakub@redhat.com>
289 * config/i386/i386.c: Include debug.h and dwarf2out.h.
290 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
293 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
295 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
297 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
299 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
300 and FP_SECOND_SSE_REGS.
301 (REG_CLASS_NAMES): Ditto.
302 (REG_CLASS_CONTENTS): Ditto.
303 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
304 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
305 (ix86_preferred_output_reload_class): Ditto.
306 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
307 clobber constraint to "=&f".
308 (fix_truncdi_i387): Ditto.
310 (fistdi2_<rounding>): Ditto.
311 (fpremxf4_i387): Change "=u" constraint to "=f".
312 (fprem1xf4_i387): Ditto.
314 (fptanxf4_i387): Ditto.
315 (fxtractxf3_i387): Ditto.
316 (fscalexf4_i387): Ditto.
317 (atan2xf3): Change "u" constraint to "f".
318 (fyl2xxf3_i387): Ditto.
319 (fyl2xp1xf3_i387): Ditto.
321 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
324 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
325 for removed I387_MASK_PM entity.
328 2018-09-26 Jeff Law <law@redhat.com>
330 2018-09-26 Alexey Neyman <stilor@att.net>
332 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
333 headers are no longer pulled in by <isl/val.h>.
335 2018-09-26 Richard Biener <rguenther@suse.de>
338 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
339 or concrete instance DIE to the tree. Create abstract origin
340 attributes also for concrete instances.
342 2018-09-26 Alexey Neyman <stilor@att.net>
344 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
345 headers are no longer pulled in by <isl/val.h>.
347 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
349 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
350 Use new helper functions.
351 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
352 Use new helper functions.
353 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
354 aarch_mm_needs_release): New declarations.
355 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
356 aarch_mm_needs_release): New.
358 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
360 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
361 (arm32_output_mi_thunk): Deal with long calls.
363 2018-09-26 Richard Biener <rguenther@suse.de>
367 * tree-inline.c (expand_call_inline): When the location
368 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
369 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
370 the inserted BLOCK to make inlined_function_outer_scope_p
372 * dwarf2out.c (add_call_src_coords_attributes): Do not add
373 coords for reserved locations.
375 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
377 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
378 (*call_indirect_nonlocal_sysv<mode>): Ditto.
379 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
380 (*sibcall_nonlocal_sysv<mode>): Ditto.
381 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
382 (<bd>_<mode>): Ditto.
383 (<bd>tf_<mode>): Ditto.
385 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
387 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
388 control string as a list of templates instead of as C code.
389 (*altivec_movti): Ditto.
390 * config/rs6000/darwin.md (movdf_low_di): Ditto.
392 2018-09-25 Jim Wilson <jimw@sifive.com>
394 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
395 when target symbol is weak.
397 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
400 * builtins.c (unterminated_array): Simplify.
401 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
402 where pointer arithmetic is safe.
404 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
407 * config/rs6000/altivec.md (altivec_vspltb): Use
408 const_0_to_15_operand instead of u5bit_cint_operand.
409 (*altivec_vspltb_internal): Ditto.
410 (altivec_vspltb_direct): Ditto.
411 (altivec_vsplth): Use const_0_to_7_operand instead of
413 (*altivec_vsplth_internal): Ditto.
414 (altivec_vsplth_direct): Ditto.
415 (altivec_vspltw): Use const_0_to_3_operand instead of
417 (*altivec_vspltw_internal): Ditto.
418 (altivec_vspltw_direct): Ditto.
419 (altivec_vspltsf): Ditto.
420 (*altivec_vspltsf_internal): Ditto.
421 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
422 various splats with the proper size immediate. Reorder the various
423 cases by ascending size of immediate, and put all such together.
425 2018-09-25 Richard Biener <rguenther@suse.de>
428 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
430 (maybe_create_die_with_external_ref): Do not create
431 DW_TAG_imported_unit here.
432 (add_abstract_origin_attribute): Handle external BLOCK refs.
433 (dwarf2out_abstract_function): Simplify LTO case.
434 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
435 rather than using maybe_create_die_with_external_ref.
437 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
440 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
441 (frndintxf2_mask_pm_i387): Ditto.
442 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
443 Enable for !flag_trapping_math.
444 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
445 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
446 Change operand 1 predicate to nonimmediate_operand.
447 (attr "i387_cw"): Remove mask_pm.
448 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
449 (enum ix86_entity): Remove I387_MASK_PM.
450 * config/i386/i386.c (ix86_i387_mode_needed): Do not
452 (ix86_mode_needed): Ditto.
453 (ix86_mode_after): Ditto.
454 (ix86_mode_entry): Ditto.
455 (ix86_mode_exit): Ditto.
456 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
458 2018-09-25 Jakub Jelinek <jakub@redhat.com>
460 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
461 to_update_switch_stmts to vNULL instead of calling create on them
464 2018-09-25 Richard Biener <rguenther@suse.de>
466 PR tree-optimization/87402
467 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
468 (visit_phi): Re-instantiate handling of supposed to be VARYING
469 but non-VARYING backedge value.
471 2018-09-25 Richard Biener <rguenther@suse.de>
474 * dwarf2out.c (struct sym_off_pair): New.
475 (external_die_map): New global.
476 (lookup_decl_die): When in LTO create DIEs lazily from the
478 (lookup_block_die): New function, create DIEs lazily in LTO.
479 (equate_block_to_die): New function.
480 (dwarf2out_die_ref_for_decl): During WPA get the association
481 from the external DIE map.
482 (dwarf2out_register_external_die): Record mapping into the
484 (maybe_create_die_with_external_ref): New function split out from
485 DIE generation part of old dwarf2out_register_external_die.
486 (add_abstract_origin_attribute): Do not return the DIE. When
487 in LTO reference externals directly.
488 (dwarf2out_abstract_function): When in LTO ignore calls for
489 decls with external DIEs (already present abstract instances).
490 (gen_call_site_die): Adjust.
491 (add_high_low_attributes): Likewise.
492 (gen_lexical_block_die): Likewise.
493 (gen_inlined_subroutine_die): Likewie.
494 (gen_block_die): Likewise.
495 (dwarf2out_inline_entry): Likewise.
496 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
499 2018-09-25 Martin Liska <mliska@suse.cz>
501 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
502 integers and not by a float value.
504 2018-09-25 Martin Liska <mliska@suse.cz>
507 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
509 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
510 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
511 * godump.c (go_format_type): Remove extra parenthesis.
513 2018-09-25 Martin Liska <mliska@suse.cz>
515 * alias.c (set_dest_equal_p): Remove unused function.
516 * config/i386/i386.c (def_builtin_pure2): Likewise.
517 * diagnostic-show-locus.c (class layout): Remove
519 (layout::layout): Likewise here.
520 * dump-context.h (class temp_dump_context): Likewise.
521 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
522 (add_AT_loclistsptr): Likewise.
523 (add_AT_offset): Likewise.
524 (get_AT_hi_pc): Likewise.
525 (is_comdat_die): Likewise.
526 (type_is_enum): Likewise.
528 (add_AT_vms_delta): Likewise.
529 (is_class_die): Likewise.
530 * edit-context.c (class line_event): Remove unused field.
531 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
533 * ipa-cp.c (ipa_get_vr_lat): Likewise.
534 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
535 (ok_for_base_p_nonstrict): Likewise.
536 * tree-chrec.c (is_not_constant_evolution): Likewise.
537 (chrec_fold_poly_cst): Likewise.
538 * tree-if-conv.c (has_pred_critical_p): Likewise.
539 * tree-ssa-coalesce.c (print_exprs): Likewise.
540 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
541 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
542 * tree-vrp.c (value_ranges_intersect_p): Likewise.
543 (value_range_nonnegative_p): Likewise.
545 2018-09-25 Martin Liska <mliska@suse.cz>
547 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
548 Do not handle "GNU Pascal".
549 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
551 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
552 from documentation. Likewise.
553 * dbxout.c (dbxout_range_type): Likewise.
554 * doc/cpp.texi: Likewise.
555 * doc/extend.texi: Likewise.
556 * doc/frontends.texi: Likewise.
557 * doc/invoke.texi: Remove Pascal entry.
558 * tree.def (CLEANUP_POINT_EXPR): Likewise.
559 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
561 2018-09-25 Martin Liska <mliska@suse.cz>
564 * doc/invoke.texi: Document all parameters and remove default
567 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
570 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
571 contains HOST_WIDE_INTs when computing its size.
573 2018-09-24 Jim Wilson <jimw@sifive.com>
576 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
578 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
580 2018-09-24 Andrew Pinski <apinski@marvell.com>
582 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
583 access prev before checking it for NULLness in the
584 AARCH64_FUSE_CMP_BRANCH case.
586 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
589 * config/i386/i386.c (rest_of_insert_endbranch): Set
590 endbr_queued_at_entrance to true and don't insert ENDBR if
591 x86_function_profiler will be called.
592 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
594 * config/i386/i386.h (machine_function): Add
595 endbr_queued_at_entrance.
597 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
599 * genattrtab.c (mk_attr_alt): Use alternative_mask.
600 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
602 (check_attr_test): Use alternative_mask.
603 (get_attr_value): Likewise.
604 (compute_alternative_mask): Use alternative_mask and XWINT.
605 (make_alternative_compare): Use alternative_mask.
606 (attr_alt_subset_p): Use XWINT.
607 (attr_alt_subset_of_compl_p): Likewise.
608 (attr_alt_intersection): Use alternative_mask and XWINT.
609 (attr_alt_union): Likewise.
610 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
611 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
612 (simplify_test_exp): Use alternative_mask and XWINT.
613 (write_test_expr): Use alternative_mask and XWINT, adjust bit
614 number calculation to support 64 bits. Generate code that
616 (main): Use alternative_mask.
617 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
619 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
622 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
623 RETURN+USE when returning via %r14.
625 2018-09-24 Martin Liska <mliska@suse.cz>
627 * gcov.c (output_lines): Print colorization legend
628 for both flag_use_colors and flag_use_hotness_colors.
631 2018-09-24 Martin Liska <mliska@suse.cz>
633 * coverage.c (get_coverage_counts): Use warning_at
634 with current_function_decl location. Use %qD in warning
637 2018-09-24 Martin Liska <mliska@suse.cz>
639 * memory-block.h (memory_block_pool::release): Annotate with
640 valgrind that the memory is not accessible.
642 2018-09-24 Martin Liska <mliska@suse.cz>
645 * asan.c: Make asan_handled_variables extern.
647 * cfgexpand.c (expand_stack_vars): Make sure
648 a representative is unpoison if another
649 variable in the partition is handled by
650 use-after-scope sanitization.
652 2018-09-24 Richard Biener <rguenther@suse.de>
654 PR tree-optimization/63155
655 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
656 the worklist when the edge of the respective argument isn't
659 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
661 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
662 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
663 (MASK_CLASS_P): Update for rename.
664 (MAYBE_MASK_CLASS_P): Ditto.
665 (REG_CLASS_NAMES): Update.
666 (REG_CLASS_CONTENT): Update.
667 * config/i386/i386.c (regclass_map): Update for MASK_REG
668 and ALL_MASK_REGS rename.
669 * config/i386/constraints.md (Yk): Update for rename.
672 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
674 * config/i386/i386.h (enum reg_class): Remove
675 EVEX_SSE_REGS and MOD4_SSE_REGS.
676 (REG_CLASS_NAMES): Update.
677 (REG_CLASS_CONTENT): Update.
678 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
679 registers as ALL_SSE_REGS.
680 (ix86_additional_allocno_class_p): Remove.
681 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
682 (ix86_register_priority): Lower priority of EVEX SSE registers.
683 Use IN_RANGE macro where appropriate.
684 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
685 AVX-5124VNNIW checks.
686 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
687 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
688 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
689 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
690 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
691 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
692 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
693 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
694 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
695 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
696 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
697 * config/i386/constraints.md (Yh): Remove.
699 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
701 * config/i386/i386.c (regclass_map): Declare integer REX registers
704 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
706 * doc/service.texi (Service): Switch the fsf.org link to https.
708 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
711 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
712 Define to speculation_safe_value_not_needed.
714 2018-09-21 Florian Weimer <fweimer@redhat.com>
717 * doc/extend.texi (Common Function Attributes): Mention that
718 noreturn suppresses tail call optimization.
720 2018-09-21 Jeff Law <law@redhat.com>
722 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
723 vr_values::cleanup_edges_and_switches.
724 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
726 (identify_jump_threads): Remove EDGE_IGNORE handling.
727 (execute_vrp): Move handling of to_remove_edges and
728 to_update_switch_stmts into vr_values class member functions.
729 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
730 (to_update_switch_stmts): Likewise.
731 * vr-values.c: Include cfghooks.h.
732 (vr_values::vr_values): Initialize to_remove_edges and
733 to_update_switch_stmts.
734 (vr_values::~vr_values): Verify to_remove_edges and
735 to_update_switch_stmts are empty.
736 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
737 (vr_values::cleanup_edges_and_switches): New member function.
738 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
739 function. Add new data members.
741 2018-09-21 David Malcolm <dmalcolm@redhat.com>
743 PR tree-optimization/87309
744 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
745 calls with pflags and alt_flags.
746 (selftest::test_capture_of_dump_calls): Add test of interaction of
747 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
749 2018-09-21 Olivier Hainque <hainque@adacore.com>
751 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
753 2018-09-21 Olivier Hainque <hainque@adacore.com>
755 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
757 2018-09-21 Olivier Hainque <hainque@adacore.com>
759 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
760 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
761 (VXWORKS_LIBS_RTP): Minor reordering.
763 2018-09-21 Olivier Hainque <hainque@adacore.com>
765 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
766 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
768 2018-09-21 Olivier Hainque <hainque@adacore.com>
770 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
771 (PTRDIFF_TYPE): Likewise.
773 2018-09-21 Olivier Hainque <hainque@adacore.com>
775 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
776 triplet, similar to support for VxWorks7.
777 * config/vxworks-dummy.h: Provide a default definition
778 of TARGET_VXWORKS64 to 0.
780 2018-09-21 Olivier Hainque <hainque@adacore.com>
782 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
783 * config/vxworks-dummy.h: here.
785 2018-09-21 Olivier Hainque <hainque@adacore.com>
787 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
789 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
790 Bo Zhou <zbo.zhou@hisilicon.com>
792 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
793 * config/aarch64/aarch64-tune.md: Regenerated.
794 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
795 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
796 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
798 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
799 Julian Brown <julian@codesourcery.com>
801 * builtins.c (get_builtin_sync_mem): Handle address spaces.
803 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
805 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
806 if the call takes a static chain.
808 2018-09-21 Martin Liska <mliska@suse.cz>
810 * auto-profile.c (autofdo_source_profile::read): Do not
812 (read_profile): Do not add working sets.
813 (read_autofdo_file): Remove sum_all.
814 (afdo_callsite_hot_enough_for_early_inline): Remove const
816 * coverage.c (struct counts_entry): Remove gcov_summary.
817 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
818 do not support GCOV_TAG_PROGRAM_SUMMARY.
819 (get_coverage_counts): Remove summary and expected
821 * coverage.h (get_coverage_counts): Likewise.
822 * doc/gcov-dump.texi: Remove -w option.
823 * gcov-dump.c (dump_working_sets): Remove.
824 (main): Do not support '-w' option.
825 (print_usage): Likewise.
826 (tag_summary): Likewise.
827 * gcov-io.c (gcov_write_summary): Do not dump
829 (gcov_read_summary): Likewise.
830 (gcov_histo_index): Remove.
831 (gcov_histogram_merge): Likewise.
832 (compute_working_sets): Likewise.
833 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
835 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
836 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
837 (GCOV_HISTOGRAM_SIZE): Remove.
838 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
839 (struct gcov_summary): Simplify rapidly just
840 to runs and sum_max fields.
841 (gcov_histo_index): Remove.
842 (NUM_GCOV_WORKING_SETS): Likewise.
843 (compute_working_sets): Likewise.
844 * gcov-tool.c (print_overlap_usage_message): Remove
846 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
847 (output_lines): Remove program related line.
848 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
849 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
851 (input_profile_summary): Do not read it.
852 (merge_profile_summaries): And do not merge it.
853 (input_symtab): Do not call removed function.
854 * modulo-sched.c (sms_schedule): Do not print sum_max.
855 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
856 removed when histogram method was invented.
857 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
859 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
861 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
862 and dump selected value.
863 * profile.c (add_working_set): Remove.
864 (get_working_sets): Likewise.
865 (find_working_set): Likewise.
866 (get_exec_counts): Do not work with working sets.
867 (read_profile_edge_counts): Do not inform as sum_max is removed.
868 (compute_branch_probabilities): Likewise.
869 (compute_value_histograms): Remove argument for call of
871 * profile.h: Do not make gcov_summary const.
873 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
875 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
877 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
879 PR tree-optimization/86990
880 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
881 Check that the entire merged store group is made of constants only for
884 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
886 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
887 (VTABLE_VERIFICATION_SPEC): Likewise.
888 (SANITIZER_EARLY_SPEC): Likewise.
889 (SANITIZER_SPEC): Likewise.
890 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
891 * doc/invoke.texi (Link Options): Document -r.
893 2018-09-20 Richard Biener <rguenther@suse.de>
896 * gimplify.c (gimplify_expr): Retain alignment of
897 addressable lvalue in dereference.
899 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
902 * configure.ac: Check for .loc is_stmt support.
903 * configure, config.in: Rebuilt.
904 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
907 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
909 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
910 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
913 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
915 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
917 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
920 2018-09-20 Marek Polacek <polacek@redhat.com>
922 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
924 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
926 PR tree-optimization/87288
927 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
928 into account when determining PEELING_FOR_NITERS.
930 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
932 PR tree-optimization/86877
933 * tree-vect-loop.c (vect_analyze_loop_2): Call
934 vect_verify_datarefs_alignment.
936 2018-09-19 Marek Polacek <polacek@redhat.com>
938 * doc/invoke.texi: Document -Wclass-conversion.
940 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
942 * config/pa/pa.c (pa_adjust_priority): Delete.
943 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
945 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
946 (atomic_storehi): Likewise.
947 (atomic_storesi): Likewise.
948 (atomic_loaddi): Restore compare and swap exchange loop code.
950 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
952 PR rtl-optimization/86902
953 * combine.c (try_combine): When changing the CC mode used, don't change
954 an unrelated mode in other_insn to that new CC mode.
956 2018-09-19 David Malcolm <dmalcolm@redhat.com>
958 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
959 with %T in place of calls to dump_generic_expr.
960 (prune_runtime_alias_test_list): Likewise.
961 (create_runtime_alias_checks): Likewise.
962 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
963 (vect_analyze_data_ref_dependence): Likewise.
964 (vect_slp_analyze_data_ref_dependence): Likewise.
965 (vect_record_base_alignment): Likewise. Use %G in place of call
967 (vect_compute_data_ref_alignment): Likewise.
968 (verify_data_ref_alignment): Likewise.
969 (vect_find_same_alignment_drs): Likewise.
970 (vect_analyze_group_access_1): Likewise.
971 (vect_analyze_data_ref_accesses): Likewise.
972 (dependence_distance_ge_vf): Likewise.
973 (dump_lower_bound): Likewise.
974 (vect_prune_runtime_alias_test_list): Likewise.
975 (vect_find_stmt_data_reference): Likewise.
976 (vect_analyze_data_refs): Likewise.
977 (vect_create_addr_base_for_vector_ref): Likewise.
978 (vect_create_data_ref_ptr): Likewise.
979 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
980 (vect_can_advance_ivs_p): Likewise.
981 (vect_update_ivs_after_vectorizer): Likewise.
982 (vect_gen_prolog_loop_niters): Likewise.
983 (vect_prepare_for_masked_peels): Likewise.
984 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
985 (vect_determine_vectorization_factor): Likewise.
986 (vect_is_simple_iv_evolution): Likewise.
987 (vect_analyze_scalar_cycles_1): Likewise.
988 (vect_analyze_loop_operations): Likewise.
989 (report_vect_op): Likewise.
990 (vect_is_slp_reduction): Likewise.
991 (check_reduction_path): Likewise.
992 (vect_is_simple_reduction): Likewise.
993 (vect_create_epilog_for_reduction): Likewise.
994 (vect_finalize_reduction:): Likewise.
995 (vectorizable_induction): Likewise.
996 (vect_transform_loop_stmt): Likewise.
997 (vect_transform_loop): Likewise.
998 (optimize_mask_stores): Likewise.
999 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
1000 (vect_split_statement): Likewise.
1001 (vect_recog_over_widening_pattern): Likewise.
1002 (vect_recog_average_pattern): Likewise.
1003 (vect_determine_min_output_precision_1): Likewise.
1004 (vect_determine_precisions_from_range): Likewise.
1005 (vect_determine_precisions_from_users): Likewise.
1006 (vect_mark_pattern_stmts): Likewise.
1007 (vect_pattern_recog_1): Likewise.
1008 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1009 (vect_record_max_nunits): Likewise.
1010 (vect_build_slp_tree_1): Likewise.
1011 (vect_build_slp_tree_2): Likewise.
1012 (vect_print_slp_tree): Likewise.
1013 (vect_analyze_slp_instance): Likewise.
1014 (vect_detect_hybrid_slp_stmts): Likewise.
1015 (vect_detect_hybrid_slp_1): Likewise.
1016 (vect_slp_analyze_operations): Likewise.
1017 (vect_slp_analyze_bb_1): Likewise.
1018 (vect_transform_slp_perm_load): Likewise.
1019 (vect_schedule_slp_instance): Likewise.
1020 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1021 (vect_mark_stmts_to_be_vectorized): Likewise.
1022 (vect_init_vector_1): Likewise.
1023 (vect_get_vec_def_for_operand): Likewise.
1024 (vect_finish_stmt_generation_1): Likewise.
1025 (vect_check_load_store_mask): Likewise.
1026 (vectorizable_call): Likewise.
1027 (vectorizable_conversion): Likewise.
1028 (vectorizable_operation): Likewise.
1029 (vectorizable_load): Likewise.
1030 (vect_analyze_stmt): Likewise.
1031 (vect_is_simple_use): Likewise.
1032 (vect_get_vector_types_for_stmt): Likewise.
1033 (vect_get_mask_type_for_stmt): Likewise.
1034 * tree-vectorizer.c (increase_alignment): Likewise.
1036 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
1038 * doc/rtl.texi: Adjust vec_select description.
1039 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
1040 non-constant selectors.
1042 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
1044 * config/aarch64/aarch64-protos.h
1045 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
1046 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
1047 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
1048 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
1049 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
1050 (AARCH64_FL_PROFILE): Move index so flags are ordered.
1051 (AARCH64_ISA_RCPC8_4): New flag.
1052 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
1053 to aarch64_offset_9bit_signed_unscaled_p.
1054 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
1056 * config/aarch64/constraints.md (Ust): New constraint.
1057 * config/aarch64/predicates.md.
1058 (aarch64_9bit_offset_memory_operand): New predicate.
1059 (aarch64_rcpc_memory_operand): New predicate.
1061 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
1063 PR rtl-optimization/87361
1064 * rtlanal.c (nonzero_bits1): Revert accidental change.
1066 2018-09-19 Richard Biener <rguenther@suse.de>
1068 PR tree-optimization/87349
1069 PR tree-optimization/87342
1070 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
1072 2018-09-18 Marek Polacek <polacek@redhat.com>
1074 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
1075 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
1077 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
1079 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
1081 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
1083 PR rtl-optimization/86882
1084 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
1086 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
1088 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
1089 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
1091 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
1094 * doc/invoke.texi (Link Options): Fix formatting and grammar.
1096 2018-09-18 Richard Biener <rguenther@suse.de>
1099 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
1100 (compute_samebase_partition_bases): Likewise.
1101 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
1102 (gimple_can_coalesce_p): Simplify.
1104 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
1106 Handle a library implementation of ffs calling __builtin_ffs.
1107 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
1108 (mmix_init_libfuncs): New function: make __builtin_ffs expand
1111 2018-09-17 David Malcolm <dmalcolm@redhat.com>
1113 * diagnostic-show-locus.c (class layout_range): Add field
1115 (layout_range::layout_range): Add "original_idx" param and use it
1116 to initialize new field.
1117 (make_range): Use 0 for original_idx.
1118 (layout::layout): Pass in index to calls to
1119 maybe_add_location_range.
1120 (layout::maybe_add_location_range): Add param "original_idx" and
1121 pass it on to layout_range.
1122 (layout::print_any_labels): Pass on range->m_original_idx to
1124 (gcc_rich_location::add_location_if_nearby): Use 0 for
1126 * gcc-rich-location.h (text_range_label::get_text): Update for new
1128 (range_label_for_type_mismatch::get_text): Likewise.
1130 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
1132 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
1134 2018-09-17 David Malcolm <dmalcolm@redhat.com>
1136 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
1137 format_string_diagnostic_t.
1138 (fmtwarn_n): Likewise.
1139 * substring-locations.c
1140 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
1141 (format_warning_n_va): Convert to...
1142 (format_string_diagnostic_t::emit_warning_n_va): ...this.
1143 (format_warning_va): Convert to...
1144 (format_string_diagnostic_t::emit_warning_va): ...this.
1145 (format_warning_at_substring): Convert to...
1146 (format_string_diagnostic_t::emit_warning): ...this.
1147 (format_warning_at_substring_n): Convert to...
1148 (format_string_diagnostic_t::emit_warning_n): ...this.
1149 * substring-locations.h (class format_string_diagnostic_t): New
1151 (format_warning_va): Convert to
1152 format_string_diagnostic_t::emit_warning_va.
1153 (format_warning_n_va): Convert to
1154 format_string_diagnostic_t::emit_warning_n_va.
1155 (format_warning_at_substring): Convert to
1156 format_string_diagnostic_t::emit_warning.
1157 (format_warning_at_substring_n): Convert to
1158 format_string_diagnostic_t::emit_warning_n.
1160 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
1161 Bernd Schmidt <bernds_cb1@t-online.de>
1163 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
1166 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
1168 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
1169 operand 0 predicate to nonimmediate operand.
1170 (rint<mode>2_frndint): Remove insn pattern.
1171 (rint<mode>2): Change operand 1 predicate to general_operand.
1172 Extend operand 1 to XFmode and generate rintxf2 insn.
1173 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
1174 Do not use X87MODEF mode macro.
1175 (frndintxf2_<rounding>_i387): Rename from
1176 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
1177 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
1178 to XFmode and generate significandxf3 insn.
1180 2018-09-17 Richard Biener <rguenther@suse.de>
1182 PR tree-optimization/87328
1183 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
1184 visiting unexecutable backedges when not iterating.
1185 (do_rpo_vn): Mark all edges not executable even when not
1188 2018-09-17 Martin Jambor <mjambor@suse.cz>
1191 * doc/invoke.texi (Warning Options): Likewise.
1193 2018-09-17 Richard Biener <rguenther@suse.de>
1195 PR tree-optimization/87301
1196 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
1197 clean EH info from leftover copy assignments.
1199 2018-09-17 Martin Liska <mliska@suse.cz>
1201 PR gcov-profile/85871
1202 * gcov.c (output_intermediate_file): Fix out of bounds
1205 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
1207 * config/arc/arc.c: Object attributes for core4 not reflected
1209 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
1212 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
1214 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
1216 2018-09-17 Martin Liska <mliska@suse.cz>
1218 * doc/gcov.texi: Document new option --use-hotness-colors.
1219 * gcov.c (struct source_info): Declare new field.
1220 (source_info::source_info): Set default for maximum_count.
1221 (print_usage): Add new -q option.
1222 (process_args): Process it.
1223 (accumulate_line_info): Save src->maximum_count.
1224 (output_line_beginning): Make color line number if
1225 flag_use_hotness_colors is set.
1226 (output_line_details): Pass default argument value.
1227 (output_lines): Pass src->maximum_count.
1229 2018-09-17 Martin Liska <mliska@suse.cz>
1231 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
1232 Use processor_names table.
1233 * config/i386/i386.c (ix86_default_align): Use
1234 processor_cost_table for alignment values.
1235 (ix86_option_override_internal): Use processor_names.
1236 (ix86_function_specific_print): Likewise.
1237 * config/i386/i386.h (struct processor_costs):
1238 Add alignment values.
1239 (struct ptt): Remove and replace with const char *.
1240 * config/i386/x86-tune-costs.h (struct processor_costs):
1241 Declare default alignments for all costs.
1243 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
1245 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
1246 symbolics or VR_VARYING ranges for ABS_EXPR.
1247 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
1248 when range will wrap.
1250 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
1253 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
1254 before and after a JUMP_TABLE_DATA.
1256 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
1259 * dojump.c (do_compare_and_jump): Canonicalize function pointers
1260 when one operand is a function pointer. Use POINTER_TYPE_P and
1261 FUNC_OR_METHOD_TYPE_P.
1262 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
1263 * fold-const.c (build_range_check): Likewise.
1264 * match.pd (simple_comparison): Likewise.
1266 2018-09-14 David Malcolm <dmalcolm@redhat.com>
1269 * spellcheck.c (get_edit_distance_cutoff): New function.
1270 (selftest::test_edit_distance_unit_test_oneway): Rename to...
1271 (selftest::test_get_edit_distance_one_way): ...this.
1272 (selftest::test_get_edit_distance_unit): Rename to...
1273 (selftest::test_get_edit_distance_both_ways): ...this.
1274 (selftest::test_edit_distances): Move tests to this new function,
1275 and test some more pairs of strings. Update for above renaming.
1276 (selftest::get_old_cutoff): New function.
1277 (selftest::test_get_edit_distance_cutoff): New function.
1278 (selftest::assert_suggested_for): New function.
1279 (ASSERT_SUGGESTED_FOR): New macro.
1280 (selftest::assert_not_suggested_for): New function.
1281 (ASSERT_NOT_SUGGESTED_FOR): New macro.
1282 (selftest::test_suggestions): New function.
1283 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
1284 tests to selftest::test_edit_distances and call it. Add calls to
1285 selftest::test_get_edit_distance_cutoff and
1286 selftest::test_suggestions.
1287 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
1288 (best_match::consider): Replace hard-coded cutoff calculation with
1290 (best_match::get_cutoff): New declaration.
1291 (best_match::get_best_meaningful_candidate): Likewise.
1293 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
1295 * builtins.c (fold_builtin_strlen): Remove TODO comment.
1297 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
1300 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1302 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1303 terminated string literal.
1305 2018-09-14 Martin Sebor <msebor@redhat.com>
1307 * builtins.c (unterminated_array): Handle ARRAY_REF.
1308 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
1309 * builtins.h (unterminated_array): Declare extern.
1310 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
1312 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
1315 2018-09-14 Martin Sebor <msebor@redhat.com>
1316 Jeff Law <law@redhat.com>
1318 * builtins.c (unterminated_array): New.
1319 (expand_builtin_strcpy): Adjust.
1320 (expand_builtin_strcpy_args): Detect unterminated arrays.
1321 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
1322 unterminated arrays.
1323 * gimple-fold.h (get_maxval_strlen): Add argument.
1324 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
1326 * gimple-fold.c (get_range_strlen): Add argument.
1327 (get_maxval_strlen): Adjust.
1328 * gimple-fold.h (get_range_strlen): Add argument.
1330 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
1332 * config/i386/movdirintrin.h: Fix copyright year.
1334 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
1336 * reg-stack.c: Include regs.h.
1337 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
1338 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
1339 FIRST_STACK_REG, not DFmode.
1340 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
1341 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
1342 (change stack): Default register mode to the reg_raw_mode of
1343 FIRST_STACK_REG, not DFmode.
1344 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
1345 (*swapxf): Rename from swapxf.
1347 2018-09-14 Carl Love <cel@us.ibm.com>
1349 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
1350 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
1352 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
1355 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
1358 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
1361 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
1363 2018-09-14 Jason Merrill <jason@redhat.com>
1365 Fix --enable-gather-detailed-mem-stats.
1366 * hash-table.c (hash_table_usage): Change from variable to function.
1367 * hash-table.h: Adjust.
1368 * Makefile.in: Add missing dependencies on hash-table.h.
1370 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1372 PR tree-optimization/87259
1374 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
1375 execute_cse_reciprocals_1 has tried transforming.
1377 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
1379 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
1380 VR_VARYING for PLUS/MINUS_EXPR.
1382 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
1384 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
1387 2018-09-14 Richard Biener <rguenther@suse.de>
1390 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
1391 bits for the merged partition.
1393 2018-09-13 Martin Sebor <msebor@redhat.com>
1394 Bernd Edlinger <bernd.edlinger@hotmail.de>
1396 * builtins.h (c_srlen): Add argument.
1397 * builtins.c (warn_string_no_nul): New function.
1398 (c_strlen): Add argument and use it. Update recursive calls.
1399 Pass DECL argument to string_constant to get info on non
1400 terminated strings. Update *NONSTR as needed.
1401 (fold_builtin_strlen): Add argument to calls to c_strlen.
1402 Warn for unterminated arrays.
1403 (warn_string_no_null): Add prototype.
1404 * expr.c (string_constant): Update arguments. Update recursive
1405 calls appropriately. Detect missing NUL terminator and outermost
1406 declaration its missing in.
1407 Improve checks for arrays with nonzero lower bound or elements
1408 that are not a single byte. Simplify offset computation.
1409 Simplify checks for non-NUL terminated strings.
1410 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
1411 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
1413 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
1415 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
1417 * fold-const.c (c_getstr): Fix function comment. Remove unused third
1418 argument. Fix range checks.
1419 * fold-const.h (c_getstr): Adjust protoype.
1420 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
1421 string is constant but contains no NUL byte.
1423 * expr.c (string_constant): Adjust function comment.
1424 Remove bogus check for zero termination.
1426 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
1428 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
1429 (get_constant_size): Don't make STRING_CSTs larger than they are.
1430 (check_string_literal): New check function for STRING_CSTs.
1431 (output_constant): Use it.
1433 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
1436 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
1438 2018-09-13 Richard Biener <rguenther@suse.de>
1440 PR tree-optimization/87263
1441 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
1442 (struct unwind_state): Add max_rpo field.
1443 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
1444 Compute max_rpo, the max RPO number a block can be backwards reached
1445 from. Re-write non-iterating mode to a RPO ordered worklist approach,
1446 separating it from the iterating mode.
1448 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1450 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
1451 (rfs_decision): New scheduling decision.
1453 2018-09-13 Richard Biener <rguenther@suse.de>
1456 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
1457 (vn_nary_op_insert_pieces_predicated): Do not write useless
1458 valid_dominated_by_p entry outside of the allocated storage.
1460 2018-09-13 Omar Sandoval <osandov@osandov.com>
1461 Tom de Vries <tdevries@suse.de>
1464 * dwarf2out.c (is_c): New function.
1465 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
1467 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
1470 * config/aarch64/aarch64.md (*aarch64_bfxil):
1472 * config/aarch64/constraints.md (Ulc): Define.
1473 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
1475 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
1478 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1480 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
1481 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
1482 aarch64_layout_frame call.
1483 (aarch64_expand_epilogue): Likewise.
1484 (aarch64_initial_elimination_offset): Likewise.
1485 (aarch64_get_separate_components): Likewise.
1486 (aarch64_use_return_insn_p): Likewise.
1487 (aarch64_layout_frame): Remove unneeded check.
1489 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1491 * configure.ac: Only append
1492 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
1493 gcc_config_arguments if it was never reconfigured or last reconfigure
1494 was with different arguments.
1495 * configure: Regenerated.
1497 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1498 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1501 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
1502 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
1504 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1506 PR tree-optimization/87287
1507 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
1508 X % C == 0 to X % (unsigned) C == 0 optimization to ...
1509 * match.pd (X % C == 0): ... here. New optimization.
1511 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1514 * expr.h (maybe_optimize_mod_cmp): Declare.
1515 * expr.c (mod_inv): New function.
1516 (maybe_optimize_mod_cmp): New function.
1517 (do_store_flag): Use it.
1518 * cfgexpand.c (expand_gimple_cond): Likewise.
1520 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
1521 Julian Brown <julian@codesourcery.com>
1524 * gimplify.c (gimplify_scan_omp_clauses): Set
1525 target_firstprivatize_array_bases in OpenACC parallel and kernels
1526 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
1527 OpenACC data regions.
1529 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
1531 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
1532 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
1534 2018-09-12 Richard Biener <rguenther@suse.de>
1536 PR tree-optimization/87280
1537 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
1538 edge but unreachable target.
1539 (do_rpo_vn): For conservatively handling a PHI only mark
1540 the backedge executable but not the block reachable.
1542 2018-09-12 Richard Biener <rguenther@suse.de>
1544 PR tree-optimization/87266
1545 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
1548 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
1550 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
1552 ("trunc<BFP:mode><DFP_ALL:mode>2")
1553 ("trunc<DFP_ALL:mode><BFP:mode>2")
1554 ("extend<BFP:mode><DFP_ALL:mode>2")
1555 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
1556 according to the target operand type.
1558 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1559 Andreas Krebbel <krebbel@linux.ibm.com>
1561 PR tree-optimization/86844
1562 * gimple-ssa-store-merging.c
1563 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
1564 there are any overlapping stores in between them, make sure they are
1565 also coalesced or we give up completely.
1567 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1570 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
1571 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
1573 2018-09-12 Tom de Vries <tdevries@suse.de>
1575 * common.opt (gdescribe-dies): Add option.
1576 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
1577 attribute for artifical and nameless decls.
1578 (dwarf2out_register_external_die): Add description attribute to
1579 external reference die.
1580 (add_desc_attribute): New functions.
1581 (gen_subprogram_die): Add description attribute to
1582 DW_TAG_call_site_parameter.
1583 * tree-pretty-print.c (print_generic_expr_to_str): New function.
1584 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
1585 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
1587 (@item -gdescribe-dies): Add.
1589 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1591 * tree-vrp.c (vrp_shift_undefined_p): Remove.
1592 (extract_range_from_binary_expr_1: Call
1593 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
1594 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
1597 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1599 * gimple-ssa-warn-alloca.c
1600 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
1601 field for ALLOCA_BOUND_*_LARGE.
1603 2018-09-11 Nathan Sidwell <nathan@acm.org>
1605 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
1607 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1609 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
1610 for clobbers. Remove obsolete comment.
1612 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1614 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
1615 mpxchk, mpxld and mpxst types.
1616 (define_attr length_immediate): Remove all processing of mpx types.
1617 (define_attr prefix_0f): Ditto.
1618 (define_attr memory): Ditto.
1620 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1622 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
1623 (log<mode>2): Change operand 1 predicate to general_operand.
1624 Extend operand 1 to XFmode and generate logxf3 insn.
1625 (log10<mode>2): Change operand 1 predicate to general_operand.
1626 Extend operand 1 to XFmode and generate log10xf3 insn.
1627 (log2<mode>2): Change operand 1 predicate to general_operand.
1628 Extend operand 1 to XFmode and generate log2xf3 insn.
1629 (fyl2xp1_extend<mode>xf3_i387): Remove.
1630 (log1p<mode>2): Change operand 1 predicate to general_operand.
1631 Extend operand 1 to XFmode and generate log1pxf3 insn.
1632 (fxtract_extend<mode>xf3_i387): Remove.
1633 (logb<mode>2): Change operand 1 predicate to general_operand.
1634 Extend operand 1 to XFmode and generate logbxf3 insn.
1635 (ilogb<mode>2): Change operand 1 predicate to general_operand.
1636 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
1637 (significand<mode>2): Change operand 1 predicate to general_operand.
1638 Extend operand 1 to XFmode and generate significandxf3 insn.
1640 2018-09-11 Nathan Sidwell <nathan@acm.org>
1642 * gcc.c (perror_with_name, pfatal_with_name): Delete.
1643 (load_specs): Use fatal_error.
1644 (DELETE_IF_ORDINARY, process_command): Use error.
1645 (execute, run_attempt): Use fatal_error.
1647 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
1649 * diagnostic-core.h (sorry_at): New prototype.
1650 * diagnostic.c (sorry_at): New function.
1652 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
1654 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
1655 by zero as VR_UNDEFINED.
1657 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
1659 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
1660 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
1661 (<sincos>mode2): New expander.
1662 (sincos_extend<mode>xf3_i387): Remove insn pattern.
1663 (sincos -> sin, cos splitters): Remove splitter patterns.
1664 (sincos<mode>3): Change operand 2 predicate to general_operand.
1665 Extend operand 2 to XFmode and generate sincosxf3 insn.
1666 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
1667 Change operand 3 predicate to const1_operand.
1668 (fptan_extend<mode>xf4_i387): Remove insn pattern.
1669 (tanxf2): Update operands in the call to fptanxf4_i387.
1670 (tan<mode>2): Change operand 1 predicate to general_operand.
1671 Extend operand 1 to XFmode and generate tanxf3 insn.
1672 (atan2xf3): Rename from *fpatanxf3_i387.
1673 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
1674 (atan2xf3): Remove expander.
1675 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
1676 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
1677 (atan<mode>2): Change operand 1 predicate to general_operand.
1678 Extend operand 1 to XFmode and generate atanxf3 insn.
1680 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
1682 * config/i386/i386.md (x87/SSE constant load splitter): Use
1683 memory_operand instead of nonimmediate_operand for input operand
1686 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
1688 * config/i386/i386.md (float partial SSE register stall splitter): Move
1689 splitter near its instruction pattern.
1690 (float_extend partial SSE register stall splitter): Ditto.
1691 (float_truncate partial SSE register stall splitter): Ditto.
1693 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
1696 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1697 to speculation_safe_value_not_needed.
1700 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
1701 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
1702 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
1703 leaf_function_p, instead use has_hard_reg_initial_val.
1705 2018-09-09 Nathan Sidwell <nathan@acm.org>
1707 * gcc.h (pfatal_with_name): Don't declare here.
1708 * gcc.c (pfatal_with_name): Make static.
1710 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
1712 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
1715 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
1717 PR rtl-optimization/85458
1718 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
1719 priority hook to reduce the priority of EXPR.
1721 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
1723 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
1724 DImode for x87 on 32bit targets. Conditionally disable x87 modes
1725 with X87_ENABLE_FLOAT. Remove preparation code.
1726 (*float<SWI48:mode><MODEF:mode>2): Rename from
1727 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
1728 math using "enabled" attribute.
1729 (*floatdi<MODEF:mode>2_i387): Rename from
1730 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
1731 enable for 32bit targets only.
1732 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
1734 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
1735 as operand 1 predicate. Rewrite as define_insn_and_split.
1736 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
1738 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
1740 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
1741 to fallthru to FLOAT case.
1743 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1746 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
1747 around folding of vec_sl to handle out of range shift values.
1749 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
1751 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
1752 Update callers to gen_fix_trunc<mode>_i387_fisttp
1753 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
1754 nonimmediate_operand.
1755 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
1756 and corresponding splitters.
1757 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
1758 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
1759 (fix_truncdi_i387_with_temp): Remove insn pattern
1760 and corresponding splitters.
1761 (fix_trunc<mode>_i387): Change operand 0 predicate to
1762 nonimmediate_operand.
1763 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1764 and corresponding splitters.
1765 (*fistdi2_1): Remove.
1767 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1768 (lrintxfdi2): Remove expander. Reimplement as define_insn.
1769 (*fist<mode>2_1): Remove.
1770 (fist<mode>2): Ditto.
1771 (fist<mode>2_with_temp): Remove insn pattern and corresponding
1773 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
1774 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1775 (fistdi2_<rounding>): Change operand 0 predicate to
1776 nonimmediate_operand.
1777 (fistdi2_<rounding>_with_temp): Remove insn pattern
1778 and corresponding splitters.
1779 (fist<mode>2_<rounding>): Change operand 0 predicate to
1780 nonimmediate_operand.
1781 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1782 and corresponding splitters.
1784 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1786 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
1788 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1791 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1793 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1794 early gimple folding of vec_splat().
1795 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
1796 * gimple-fold.h: Add an extern define for tree_vec_extract().
1798 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
1800 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
1801 wrappers around TREE_TYPE comparisons.
1803 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1806 * config/s390/predicates.md: Add nonsym_memory_operand.
1807 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
1808 contains a SYMBOL_REF, load it into an intermediate pseudo.
1809 (s390_emit_compare_and_swap): Legitimize operand.
1810 * config/s390/s390.md: Use the new nonsym_memory_operand
1811 with UNSPECV_CAS patterns.
1813 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1816 * config/s390/s390-passes.def: New file.
1817 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
1819 (make_pass_s390_early_mach): Add declaration.
1820 * config/s390/s390.c (make_pass_s390_early_mach):
1821 (s390_option_override): Remove dynamic registration.
1822 * config/s390/t-s390: Add s390-passes.def.
1824 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
1826 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1827 Remove UNSPEC_LTREL_BASE check.
1828 (annotate_constant_pool_refs): Likewise.
1829 (find_constant_pool_ref): Likewise.
1830 (find_ltrel_base): Removed.
1831 (replace_ltrel_base): Removed.
1832 (s390_mainpool_finish): Remove replace_ltrel_base call.
1833 (s390_chunkify_start): Remove pending LTREL_BASE logic.
1834 (s390_chunkify_finish): Remove replace_ltrel_base call.
1835 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
1837 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
1840 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1841 to speculation_safe_value_not_needed.
1843 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
1844 Bernd Schmidt <bernds_cb1@t-online.de>
1846 * config/nvptx/nvptx-opts.h: New file.
1847 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
1848 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
1850 (TARGET_SM35): New macro.
1851 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
1853 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
1855 (misa=): New option.
1856 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
1858 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
1860 * config/i386/i386.md (truncdfsf2): Remove expander.
1861 (truncdfsf2_with_temp): Ditto.
1862 (truncxf<mode>2): Ditto.
1863 (*truncdfsf_fast_mixed): Remove insn pattern.
1864 (*truncdfsf_fast_i387): Ditto.
1865 (*truncdfsf_mixed): Ditto.
1866 (*truncdfsf_i387): Ditto.
1867 (*truncdfsf2_i387_1): Ditto.
1868 (*truncxfsf2_mixed): Ditto.
1869 (*truncxfdf2_mixed): Ditto.
1870 (*truncxf<mode>2_i387_noop): Ditto. Update callers
1871 to call gen_truncxf<mode>2 instead.
1872 (*truncxf<mode>2_i387): Remove.
1873 (reg->reg splitters): Remove splitter pattern.
1874 (reg->mem splitters): Ditto.
1876 (truncdfsf2): New insn pattern.
1877 (truncxf<mode>2): Ditto.
1879 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1881 * tree-ssa-math-opts.c (is_mult_by): New function.
1882 (is_square_of): Use the above.
1883 (optimize_recip_sqrt): New function.
1884 (pass_cse_reciprocals::execute): Use the above.
1886 2018-09-05 Richard Biener <rguenther@suse.de>
1889 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
1890 to zero-init the emplaced vec.
1892 2018-09-05 Martin Liska <mliska@suse.cz>
1894 PR tree-optimization/87205
1895 * tree-switch-conversion.c (pass_lower_switch::execute):
1896 Group cases for switch statements.
1898 2018-09-05 Richard Biener <rguenther@suse.de>
1900 PR tree-optimization/87217
1901 * tree-ssa-sccvn.c (vuse_valueize): New.
1902 (vn_reference_lookup_pieces): Use it.
1903 (vn_reference_lookup): Likewise.
1905 2018-09-05 Nathan Sidwell <nathan@acm.org>
1908 * stor-layout.c (place_field): Scan forwards to check last
1909 bitfield when ms_bitfield_placement is in effect.
1911 2018-09-05 Richard Biener <rguenther@suse.de>
1914 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
1917 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
1918 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
1920 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
1921 * config.gcc (extra_objs): Build it.
1922 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
1924 * config/aarch64/aarch64-passes.def
1925 (pass_tag_collision_avoidance): New pass.
1926 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
1927 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
1928 (aarch64_classify_address): Remove static qualifier.
1929 (aarch64_address_info, aarch64_address_type): Move to...
1930 * config/aarch64/aarch64-protos.h: ... here.
1931 (make_pass_tag_collision_avoidance): New function.
1932 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
1935 2018-09-05 Martin Liska <mliska@suse.cz>
1937 * doc/gcov.texi: Update documentation of humar
1939 * gcov.c (format_count): Print one decimal place, it provides
1940 more fine number of situations like '1G' vs. '1.4G'.
1942 2018-09-05 Martin Liska <mliska@suse.cz>
1945 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
1946 * optc-gen.awk: Allow 'Var' for Deprecated options in order
1947 to generate a MASK value.
1949 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
1952 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
1953 if frame pointer isn't used.
1954 (compute_frame_pointer_to_fb_displacement): Likewise.
1956 2018-09-04 Jakub Jelinek <jakub@redhat.com>
1959 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
1960 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
1961 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
1962 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
1963 and OPTION_MASK_ISA_XSAVEC_UNSET.
1965 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
1967 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
1968 XOR operations in NAND case.
1970 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
1972 * wide-int-range.cc (wide_int_range_convert): New.
1973 * wide-int-range.h (wide_int_range_convert): New.
1974 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
1975 code into wide_int_range_convert.
1976 (extract_range_into_wide_ints): Do not munge anti range constants
1977 into the entire domain. Just return the range back.
1979 2018-09-04 Martin Liska <mliska@suse.cz>
1981 * genmatch.c (output_line_directive): Add new argument
1983 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
1985 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
1987 * doc/invoke.texi (Option Summary): Add whitespace.
1989 * doc/invoke.texi (Option Summary): Add -Waligned-new.
1991 2018-09-04 Richard Biener <rguenther@suse.de>
1993 PR tree-optimization/87211
1994 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
1995 backedge value we're supposed to treat as VARYING also number
1996 the PHI to VARYING in case it got a different value-number already.
1998 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
2000 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
2001 (extract_range_from_binary_expr_1): Do not call
2002 vrp_can_optimize_bit_op.
2003 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
2005 (wide_int_range_get_mask_and_bounds): New.
2006 (wide_int_range_optimize_bit_op): New.
2007 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
2008 (wide_int_range_bit_and): Same.
2009 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
2010 (wide_int_range_optimize_bit_op): New.
2011 (wide_int_range_get_mask_and_bounds): New.
2013 2018-09-04 Richard Biener <rguenther@suse.de>
2015 PR tree-optimization/87176
2016 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
2017 variable. When value-numbering a virtual PHI node make sure
2018 to not value-number to the backedge value.
2020 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2022 * doc/extend.texi (Long Long, Hex Floats): Document support for
2023 long long and hex floats in more recent versions of ISO C++.
2025 2018-09-03 Richard Biener <rguenther@suse.de>
2027 PR tree-optimization/87177
2028 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
2031 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2033 * bb-reorder.c (edge_order): Convert to C-qsort-style
2034 tri-state comparator.
2035 (reorder_basic_blocks_simple): Change std::stable_sort to
2038 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2040 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
2041 tri-state comparator.
2042 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
2044 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2046 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
2047 (mergesort): ... here as maximum count for using netsort.
2048 (gcc_qsort): Set nlim to 3 if stable sort is requested.
2049 (gcc_stablesort): New.
2050 * system.h (gcc_stablesort): Declare.
2052 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2054 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
2055 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
2056 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
2058 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
2060 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
2061 lxsdx and stxsdx alternatives.
2062 (*mov<mode>_hardfloat64): Ditto.
2063 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
2065 2018-09-03 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/87200
2068 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
2071 2018-09-03 Martin Liska <mliska@suse.cz>
2073 PR tree-optimization/87201
2074 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2075 Fix parenthesis in an expression.
2077 2018-09-03 Richard Biener <rguenther@suse.de>
2079 PR tree-optimization/87197
2080 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
2081 visited. CSE the VN_INFO hashtable lookup.
2083 PR tree-optimization/87169
2084 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
2085 iterating make sure there's no extra backedges from irreducible
2086 regions feeding the header. Mark the destination block
2089 2018-09-03 Martin Liska <mliska@suse.cz>
2092 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
2093 * common/common-targhooks.c (default_get_valid_option_values):
2095 * common/common-targhooks.h (default_get_valid_option_values):
2097 * common/config/i386/i386-common.c: Move processor_target_table
2099 (ix86_get_valid_option_values): New function.
2100 (TARGET_GET_VALID_OPTION_VALUES): New macro.
2101 * config/i386/i386.c (struct ptt): Move to i386-common.c.
2102 (PTA_*): Move all defined masks into i386-common.c.
2103 (ix86_function_specific_restore): Use new processor_cost_table.
2104 * config/i386/i386.h (struct ptt): Moved from i386.c.
2105 (struct pta): Likewise.
2106 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
2107 * doc/tm.texi.in: Likewise.
2108 * opt-suggestions.c (option_proposer::suggest_option):
2109 Pass prefix to build_option_suggestions.
2110 (option_proposer::get_completions): Likewise.
2111 (option_proposer::build_option_suggestions): Use the new target
2113 * opts.c (struct option_help_tuple): New struct.
2114 (print_filtered_help): Use the new target hook.
2116 2018-09-03 Martin Liska <mliska@suse.cz>
2119 * predict.c (set_even_probabilities): Add likely_edges
2120 argument and handle cases where we have precisely one
2122 (combine_predictions_for_bb): Catch also likely_edges.
2123 (tree_predict_by_opcode): Handle gswitch statements.
2124 * tree-cfg.h (find_case_label_for_value): New declaration.
2125 (find_taken_edge_switch_expr): Likewise.
2126 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2127 Find pivot in decision tree based on probabily, not by number of
2130 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
2132 * doc/standards.texi (Standards): Update Objective-C reference.
2134 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2136 * doc/install.texi (Prerequisites): Update link for MPC.
2138 2018-09-01 Michael Matz <matz@suse.de>
2140 PR tree-optimization/87074
2141 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
2142 PHIs for outer-loop uses.
2144 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2146 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
2147 * doc/invoke.texi (C Dialect Options): Ditto.
2149 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2151 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
2153 2018-08-31 Richard Biener <rguenther@suse.de>
2155 PR tree-optimization/87168
2156 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
2157 (rpo_elim::eliminate_avail): When OP was not visited it must
2160 2018-08-31 David Malcolm <dmalcolm@redhat.com>
2162 * tree-vrp.c (copy_value_range): Convert param "from" from
2163 "value_range *" to "const value_range *".
2164 (range_is_null): Likewise for param "vr".
2165 (range_int_cst_p): Likewise.
2166 (range_int_cst_singleton_p): Likewise.
2167 (symbolic_range_p): Likewise.
2168 (value_ranges_intersect_p): Likewise for both params.
2169 (value_range_nonnegative_p): Likewise for param "vr".
2170 (value_range_constant_singleton): Likewise.
2171 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
2172 (extract_range_into_wide_ints): Likewise for param "vr".
2173 (extract_range_from_multiplicative_op): Likewise for params "vr0"
2175 (vrp_can_optimize_bit_op): Likewise.
2176 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
2178 (extract_range_from_unary_expr): Likewise.
2179 (debug_value_range): Likewise for param "vr".
2180 (value_range::dump): Add "const" qualifier.
2181 (vrp_prop::check_array_ref): Convert local "vr" from
2182 "value_range *" to "const value_range *".
2183 (vrp_prop::check_mem_ref): Likewise.
2184 (vrp_prop::visit_stmt): Likewise for local "old_vr".
2185 (vrp_intersect_ranges_1): Likewise for param "vr_1".
2186 (vrp_intersect_ranges): Likewise.
2187 (simplify_stmt_for_jump_threading): Likewise for local "vr".
2188 (vrp_prop::vrp_finalize): Likewise.
2189 * tree-vrp.h (value_range::dump): Add "const" qualifier.
2190 (vrp_intersect_ranges): Add "const" qualifier to params as above.
2191 (extract_range_from_unary_expr): Likewise.
2192 (value_range_constant_singleton): Likewise.
2193 (symbolic_range_p): Likewise.
2194 (copy_value_range): Likewise.
2195 (extract_range_from_binary_expr_1): Likewise.
2196 (range_int_cst_p): Likewise.
2197 (vrp_set_zero_nonzero_bits): Likewise.
2198 (range_int_cst_singleton_p): Likewise.
2200 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
2202 * config/aarch64/arm_neon.h (vabsd_s64): New.
2203 (vnegd_s64): Likewise.
2205 2018-08-31 Martin Jambor <mjambor@suse.cz>
2207 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
2209 2018-08-31 Martin Liska <mliska@suse.cz>
2211 * ipa-icf.c (sem_item::add_type): Use
2212 sem_item::m_type_hash_cache.
2213 * ipa-icf.h: Move the cache from sem_item_optimizer
2216 2018-08-31 Nathan Sidwell <nathan@acm.org>
2218 * doc/extend.texi (Backwards Compatibility): Remove implicit
2219 extern C leeway of () being (...).
2221 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2223 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
2225 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
2229 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
2231 2018-08-31 Jakub Jelinek <jakub@redhat.com>
2234 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
2235 gen_int_mode. Formatting fixes.
2237 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
2239 * target.def (custom_function_descriptors): Improve documentation.
2240 * doc/tm.texi.in (Trampolines): Expand discussion of function
2241 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
2242 beginning of the section.
2243 * doc/tm.texi: Regenerated.
2245 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
2247 * cfg.h (class auto_edge_flag): Spell out the template-id of the
2248 base class in the initializer list. This is a workaround for
2249 building with older GCC.
2250 (class auto_bb_flag): Likewise.
2252 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
2254 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
2255 (altivec_vcmpequ<VI_char>_p): Remove star.
2256 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
2258 (expand_strncmp_vec_sequence): New function.
2259 (emit_final_str_compare_vec): New function.
2260 (expand_strn_compare): Add support for vector strncmp.
2261 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
2262 length specification to bytes.
2263 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
2264 (vcmpnezb_p): New pattern.
2265 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
2266 for option -mstring-compare-inline-limit.
2268 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
2270 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
2271 (PTA_SKYLAKE): Add PTA_AES.
2272 (PTA_GOLDMONT): Likewise.
2274 2018-08-29 Jan Hubicka <jh@suse.cz>
2277 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
2278 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
2280 2018-08-29 Jan Hubicka <jh@suse.cz>
2282 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
2284 (hash_tree): Do not visit TYPE_STUB_DECL.
2285 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
2286 stream TYPE_STUB_DECL.
2287 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
2288 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
2289 after free_lang_data.
2290 (type_in_anonymous_namespace_p): Likewise.
2292 2018-08-29 Jan Hubicka <jh@suse.cz>
2294 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2295 comment that it has to be even number.
2296 (class sreal): Change m_sig type to int32_t.
2297 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2298 int64_t for temporary calculations.
2299 (sreal_verify_basics): Drop one bit from minimum and maximum.
2301 2018-08-30 Richard Biener <rguenther@suse.de>
2303 PR tree-optimization/87147
2304 * tree-ssa-sccvn.c (SSA_VISITED): New function.
2305 (visit_phi): When the degenerate result is from the backedge and
2306 we didn't visit its definition yet drop to VARYING.
2307 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
2309 2018-08-29 Jan Hubicka <jh@suse.cz>
2311 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
2313 (hash_tree): Likewise.
2315 2018-08-29 Jan Hubicka <jh@suse.cz>
2317 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
2318 and TYPE_NEXT_REF_TO.
2320 2018-08-29 Jan Hubicka <jh@suse.cz>
2322 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2323 comment that it has to be even number.
2324 (class sreal): Change m_sig type to int32_t.
2325 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2326 int64_t for temporary calculations.
2327 (sreal_verify_basics): Drop one bit from minimum and maximum.
2329 2018-08-30 Tamar Christina <tamar.christina@arm.com>
2331 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
2333 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
2336 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
2337 if to_add is negative.
2339 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
2342 * builtins.c (c_strlen): Improve range checks.
2344 2018-08-29 Martin Sebor <msebor@redhat.com>
2345 Jeff Law <law@redhat.com>
2347 PR tree-optimization/86714
2348 PR tree-optimization/86711
2349 * builtins.c (c_strlen): Add arguments to call to string_constant.
2350 * expr.c (string_constant): Add argument. Detect missing nul
2351 terminator and outermost declaration it's missing in.
2352 * expr.h (string_constant): Add argument.
2353 * fold-const.c (read_from_constant_string): Add arguments to call to
2355 (c_getstr): Likewise.
2356 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2358 * tree-ssa-strlen.c (get_stridx): Likewise.
2360 2018-08-29 Jan Hubicka <jh@suse.cz>
2362 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
2363 Do not stream DECL_VINDEX.
2364 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
2365 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
2366 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
2368 2018-08-29 Richard Biener <rguenther@suse.de>
2370 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
2371 virtual operands that are not default defs to honor region
2373 (rpo_vn_valueize): Remove ineffective code here.
2375 2018-08-29 Richard Biener <rguenther@suse.de>
2377 PR tree-optimization/87132
2378 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
2379 when skipping defs reachable over backedges.
2381 2018-08-29 Richard Biener <rguenther@suse.de>
2383 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
2384 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
2385 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
2386 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
2387 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
2388 (vn_reference_lookup_3): Remove use of const_parms.
2389 (free_rpo_vn): Do not free const_parms.
2390 (do_rpo_vn): Do not call init_const_parms.
2391 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
2392 SSA_NAME_POINTS_TO_READONLY_MEMORY.
2393 (call_may_clobber_ref_p_1): Likewise.
2395 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
2398 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
2400 (-ftree-scev-cprop): Document.
2402 2018-08-29 Jan Hubicka <jh@suse.cz>
2404 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
2406 (sreal constructor): Update.
2407 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
2408 sreal:operator/): Update.
2410 2018-08-29 Martin Liska <mliska@suse.cz>
2412 * tree-switch-conversion.c (switch_conversion::expand):
2413 Strenghten assumption about gswitch statements.
2415 2018-08-29 Richard Biener <rguenther@suse.de>
2417 PR tree-optimization/87117
2418 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
2419 re-value-number released SSA VDEFs.
2421 2018-08-29 Richard Biener <rguenther@suse.de>
2423 PR tree-optimization/87126
2424 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
2426 2018-08-28 Jim Wilson <jimw@sifive.com>
2428 * config/riscv/pic.md: Rewrite.
2429 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
2431 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
2432 (SOFTF, default_load, softload, softstore): New.
2434 2018-08-28 Jeff Law <law@redhat.com>
2436 * fold-const.c (fold_binary_loc): Remove recently added assert.
2438 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
2440 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
2441 to OP parmeter of generated function.
2443 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
2445 PR tree-optimization/87009
2446 * match.pd: Add boolean optimizations.
2448 2018-08-28 Martin Sebor <msebor@redhat.com>
2451 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
2452 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
2453 (pass_walloca::gate): Use it.
2454 (alloca_call_type): Same.
2455 (pass_walloca::execute): Same.
2456 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
2458 2018-08-28 David Malcolm <dmalcolm@redhat.com>
2460 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
2461 GCC_VERSION for usage of "__gcc_dump_printf__" format from
2464 2018-08-28 Richard Biener <rguenther@suse.de>
2466 PR tree-optimization/87124
2467 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
2468 constants before looking up avail.
2470 2018-08-28 Jakub Jelinek <jakub@redhat.com>
2473 * calls.c (maybe_warn_nonstring_arg): Punt early if
2474 warn_stringop_overflow is zero. Don't call get_range_strlen
2475 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2476 Swap comparison operands to have constants on rhs. Only use
2477 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
2478 increment lenrng[0].
2480 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
2482 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
2483 use of tree_to_shwi. Remove duplicated test for the size being
2484 a whole number of bytes.
2486 2018-08-28 Richard Biener <rguenther@suse.de>
2488 PR tree-optimization/87117
2489 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
2490 Handle removed stmt without LHS (GIMPLE_NOP).
2492 2018-08-28 Richard Biener <rguenther@suse.de>
2494 PR tree-optimization/87117
2495 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
2496 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
2498 2018-08-28 Richard Biener <rguenther@suse.de>
2500 PR tree-optimization/87117
2501 * tree-ssa-pre.c (compute_avail): Do not make expressions
2502 with predicated values available.
2503 (get_expr_value_id): Assert we do not run into predicated value
2506 2018-08-28 Richard Biener <rguenther@suse.de>
2508 PR tree-optimization/87117
2509 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
2510 get virtual operands.
2511 (get_expr_operands): Handle STRING_CST like other decls.
2513 2018-08-28 Martin Liska <mliska@suse.cz>
2515 * tree.h: Update documentation of fndecl_built_in_p
2519 2018-08-27 Jeff Law <law@redhat.com>
2520 PR tree-optimization/87110
2521 * tree-ssa-dse.c (compute_trims): Handle non-constant
2524 2018-08-27 Martin Sebor <msebor@redhat.com>
2526 PR tree-optimization/86914
2527 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2529 2018-08-27 Martin Sebor <msebor@redhat.com>
2531 PR tree-optimization/87112
2532 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
2533 the type of the bound argument.
2535 2018-08-27 Jeff Law <law@redhat.com>
2537 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
2538 type does not have a TYPE_SIZE_UNIT.
2540 2018-08-27 Steve Ellcey <sellcey@cavium.com>
2542 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
2543 with include of backend.h.
2545 2018-08-27 Richard Biener <rguenther@suse.de>
2547 PR tree-optimization/86927
2548 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2549 use const cond reduction code.
2551 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
2553 PR tree-optimization/85758
2554 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
2556 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2559 * diagnostic-show-locus.c (class layout_range): Update for
2560 conversion of show_caret_p to a tri-state.
2561 (layout_range::layout_range): Likewise.
2562 (make_range): Likewise.
2563 (layout::maybe_add_location_range): Likewise.
2564 (layout::should_print_annotation_line_p): Don't show annotation
2565 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
2566 (layout::get_state_at_point): Update for conversion of
2567 show_caret_p to a tri-state. Bail out early for
2568 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
2569 underlining or source colorization.
2570 (gcc_rich_location::add_location_if_nearby): Update for conversion
2571 of show_caret_p to a tri-state.
2572 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
2573 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
2575 (selftest::test_one_liner_labels): Likewise.
2576 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
2577 conversion of show_caret_p to a tri-state.
2578 * pretty-print.c (text_info::set_location): Likewise.
2579 * pretty-print.h (text_info::set_location): Likewise.
2580 * substring-locations.c (format_warning_n_va): Likewise.
2581 * tree-diagnostic.c (default_tree_printer): Likewise.
2582 * tree-pretty-print.c (newline_and_indent): Likewise.
2584 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2587 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
2588 line above for line-insertion fix-it hints.
2589 (selftest::test_fixit_insert_containing_newline): Update the
2590 expected results, and add a test with line-numbering enabled.
2592 2018-08-27 Martin Liska <mliska@suse.cz>
2595 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
2596 params with DECL_HAS_VALUE_EXPR_P.
2598 2018-08-27 Martin Liska <mliska@suse.cz>
2600 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
2601 selected expansion strategy.
2603 2018-08-27 Martin Liska <mliska@suse.cz>
2605 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
2606 * builtins.c (is_builtin_fn): Likewise.
2607 * attribs.c (diag_attr_exclusions): Use new function
2608 fndecl_built_in_p and remove check for FUNCTION_DECL if
2610 (builtin_mathfn_code): Likewise.
2611 (fold_builtin_expect): Likewise.
2612 (fold_call_expr): Likewise.
2613 (fold_builtin_call_array): Likewise.
2614 (fold_call_stmt): Likewise.
2615 (set_builtin_user_assembler_name): Likewise.
2616 (is_simple_builtin): Likewise.
2617 * calls.c (gimple_alloca_call_p): Likewise.
2618 (maybe_warn_nonstring_arg): Likewise.
2619 * cfgexpand.c (expand_call_stmt): Likewise.
2620 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
2621 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
2622 (cgraph_node::verify_node): Likewise.
2623 * cgraphclones.c (build_function_decl_skip_args): Likewise.
2624 (cgraph_node::create_clone): Likewise.
2625 * config/arm/arm.c (arm_insert_attributes): Likewise.
2626 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
2627 * dse.c (scan_insn): Likewise.
2628 * expr.c (expand_expr_real_1): Likewise.
2629 * fold-const.c (operand_equal_p): Likewise.
2630 (fold_binary_loc): Likewise.
2631 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
2632 * gimple-low.c (lower_stmt): Likewise.
2633 * gimple-pretty-print.c (dump_gimple_call): Likewise.
2634 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
2636 * gimple.c (gimple_build_call_from_tree): Likewise.
2637 (gimple_call_builtin_p): Likewise.
2638 (gimple_call_combined_fn): Likewise.
2639 * gimplify.c (gimplify_call_expr): Likewise.
2640 (gimple_boolify): Likewise.
2641 (gimplify_modify_expr): Likewise.
2642 (gimplify_addr_expr): Likewise.
2643 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
2644 * ipa-cp.c (determine_versionability): Likewise.
2645 * ipa-fnsummary.c (compute_fn_summary): Likewise.
2646 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
2647 * ipa-split.c (visit_bb): Likewise.
2648 (split_function): Likewise.
2649 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2650 * lto-cgraph.c (input_node): Likewise.
2651 * lto-streamer-out.c (write_symbol): Likewise.
2652 * omp-low.c (setjmp_or_longjmp_p): Likewise.
2653 (lower_omp_1): Likewise.
2654 * predict.c (strip_predict_hints): Likewise.
2655 * print-tree.c (print_node): Likewise.
2656 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
2657 * trans-mem.c (is_tm_irrevocable): Likewise.
2658 (is_tm_load): Likewise.
2659 (is_tm_simple_load): Likewise.
2660 (is_tm_store): Likewise.
2661 (is_tm_simple_store): Likewise.
2662 (is_tm_abort): Likewise.
2663 (tm_region_init_1): Likewise.
2664 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
2665 * tree-cfg.c (verify_gimple_call): Likewise.
2666 (move_stmt_r): Likewise.
2667 (stmt_can_terminate_bb_p): Likewise.
2668 * tree-eh.c (lower_eh_constructs_2): Likewise.
2669 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2670 * tree-inline.c (remap_gimple_stmt): Likewise.
2671 (copy_bb): Likewise.
2672 (estimate_num_insns): Likewise.
2673 (fold_marked_statements): Likewise.
2674 * tree-sra.c (scan_function): Likewise.
2675 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
2676 (optimize_stack_restore): Likewise.
2677 (pass_fold_builtins::execute): Likewise.
2678 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
2679 (mark_all_reaching_defs_necessary_1): Likewise.
2680 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
2681 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2682 (pass_forwprop::execute): Likewise.
2683 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2684 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2685 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
2686 * tree-ssa-strlen.c (get_string_length): Likewise.
2687 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
2688 (find_func_aliases_for_call): Likewise.
2689 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
2690 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
2691 * tree-tailcall.c (find_tail_calls): Likewise.
2692 * tree.c (need_assembler_name_p): Likewise.
2693 (free_lang_data_in_decl): Likewise.
2694 (get_call_combined_fn): Likewise.
2695 * ubsan.c (is_ubsan_builtin_p): Likewise.
2696 * varasm.c (incorporeal_function_p): Likewise.
2697 * tree.h (DECL_BUILT_IN): Remove and replace with
2699 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
2700 (fndecl_built_in_p): New.
2702 2018-08-27 Martin Liska <mliska@suse.cz>
2704 PR tree-optimization/86847
2705 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
2706 Dump also subtree probability.
2707 (switch_decision_tree::do_jump_if_equal): New function.
2708 (switch_decision_tree::emit_case_nodes): Handle special
2709 situations in balanced tree that can be emitted much simpler.
2710 Fix calculation of probabilities that happen in tree expansion.
2711 * tree-switch-conversion.h (struct cluster): Add
2713 (struct simple_cluster): Likewise.
2714 (struct case_tree_node): Add new function has_child.
2715 (do_jump_if_equal): New.
2717 2018-08-27 Martin Liska <mliska@suse.cz>
2719 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
2720 Add new argument to bit_test_cluster constructor.
2721 (bit_test_cluster::emit): Set bits really number of values
2723 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
2724 probability argument.
2725 * tree-switch-conversion.h (struct bit_test_cluster):
2726 Add m_handles_entire_switch.
2728 2018-08-27 Martin Liska <mliska@suse.cz>
2730 PR tree-optimization/86702
2731 * tree-switch-conversion.c (jump_table_cluster::emit):
2732 Make probabilities even for values in jump table
2733 according to number of cases handled.
2734 (switch_decision_tree::compute_cases_per_edge): Pass
2735 argument to reset_out_edges_aux function.
2736 (switch_decision_tree::analyze_switch_statement): Likewise.
2737 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
2740 2018-08-27 Martin Liska <mliska@suse.cz>
2742 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
2743 cfun argument explicitly.
2744 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
2745 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
2746 function gimple_switch_default_bb.
2747 (convert_switch_statements):
2749 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
2750 * stmt.c (label_to_block_fn): Use label_to_block and pass
2751 cfun argument explicitly and use gimple_switch_label_bb.
2752 (expand_case): Likewise.
2753 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
2754 cfun argument explicitly. Likewise.
2755 (make_edges_bb): Likewise.
2756 (make_cond_expr_edges): Likewise.
2757 (get_cases_for_edge): Likewise.
2758 (make_gimple_switch_edges): Likewise.
2759 (label_to_block_fn): Likewise.
2760 (label_to_block): Likewise.
2761 (make_goto_expr_edges): Likewise.
2762 (make_gimple_asm_edges): Likewise.
2763 (main_block_label): Likewise.
2764 (group_case_labels_stmt): Likewise.
2765 (find_taken_edge_computed_goto): Likewise.
2766 (find_taken_edge_switch_expr): Likewise.
2767 (gimple_verify_flow_info): Likewise.
2768 (gimple_redirect_edge_and_branch): Likewise.
2769 (gimple_switch_label_bb): New function.
2770 (gimple_switch_default_bb): Likewise.
2771 (gimple_switch_edge): Likewise.
2772 (gimple_switch_default_edge): Likewise.
2773 * tree-cfg.h (label_to_block_fn): Remove and replace ...
2774 (label_to_block): ... with this.
2775 (gimple_switch_label_bb): New.
2776 (gimple_switch_default_bb): Likewise.
2777 (gimple_switch_edge): Likewise.
2778 (gimple_switch_default_edge): Likewise.
2779 * tree-cfgcleanup.c (convert_single_case_switch): Use
2780 new gimple functions and pass new argument to label_to_block.
2781 (cleanup_control_flow_bb):
2782 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2783 cfun argument explicitly.
2784 (make_eh_edges): Likewise.
2785 (redirect_eh_dispatch_edge): Likewise.
2786 (lower_resx): Likewise.
2787 (lower_eh_dispatch): Likewise.
2788 (maybe_remove_unreachable_handlers): Likewise.
2789 (unsplit_eh): Likewise.
2790 (cleanup_empty_eh): Likewise.
2791 (verify_eh_edges): Likewise.
2792 (verify_eh_dispatch_edge): Likewise.
2793 * tree-ssa-dom.c (record_edge_info): Likewise.
2794 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2795 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
2796 (thread_through_normal_block): Likewise.
2797 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
2798 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
2799 * tree-switch-conversion.c (switch_conversion::collect): Use new
2801 (switch_conversion::check_final_bb): Likewise.
2802 (switch_conversion::gather_default_values): Pass new argument
2804 (switch_conversion::build_constructors): Likewise.
2805 (switch_decision_tree::compute_cases_per_edge): Use new
2806 gimple_switch_edge function.
2807 (switch_decision_tree::analyze_switch_statement): Pass new argument
2809 (switch_decision_tree::try_switch_expansion): Use
2810 gimple_switch_default_edge.
2811 * tree-vrp.c (find_switch_asserts): Pass new argument
2813 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
2814 (vr_values::simplify_switch_using_ranges): Likewise.
2816 2018-08-27 Richard Biener <rguenther@suse.de>
2818 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
2819 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
2821 * tree-ssa-sccvn.h (struct vn_pval): New structure.
2822 (struct vn_nary_op_s): Add unwind_to member. Add
2823 predicated_values flag and put result into a union together
2824 with a linked list of vn_pval.
2825 (struct vn_ssa_aux): Add name member to make maintaining
2826 a map of SSA name to vn_ssa_aux possible. Remove no longer
2827 needed info, dfsnum, low, visited, on_sccstack, use_processed
2828 and range_info_anti_range_p members.
2829 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
2830 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
2832 (vn_valueize): New global.
2833 (vn_context_bb): Likewise.
2834 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
2835 VN_INFO_PTR_INFO): Remove.
2836 * tree-ssa-sccvn.c: ... (rewrite)
2837 (pass_fre::execute): For -O2+ initialize loops and run
2838 RPO VN in optimistic mode (iterating). For -O1 and -Og
2839 run RPO VN in non-optimistic mode.
2840 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
2841 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
2842 * doc/invoke.texi (sccvn-max-scc-size): Remove.
2843 (rpo-vn-max-loop-depth): Document.
2844 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
2845 when valuezing the VUSE signals we walked out of the region.
2846 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
2847 (phi_translate): Set VN context block to use for availability
2849 (compute_avail): Likewise.
2850 (pre_valueize): New function.
2851 (pass_pre::execute): Adjust to the RPO VN API.
2853 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
2854 (propagate_constants_for_unrolling): Remove.
2855 (tree_unroll_loops_completely): Perform value-numbering
2856 on the unrolled bodies loop parent.
2858 2018-08-27 Richard Biener <rguenther@suse.de>
2860 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
2861 for partial antic compute.
2863 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2865 PR rtl-optimization/87065
2866 * combine.c (simplify_if_then_else): Formatting fix.
2867 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2869 (known_cond): Don't return const_true_rtx for vector modes. Use
2870 CONST0_RTX instead of const0_rtx. Formatting fixes.
2872 2018-08-27 Martin Liska <mliska@suse.cz>
2874 PR gcov-profile/87069
2875 * gcov.c (process_file): Record files already processed
2876 and warn about a file being processed multiple times.
2878 2018-08-27 Martin Liska <mliska@suse.cz>
2881 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2882 Set default values for x_aarch64_*_string strings.
2883 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
2884 prefix. For -mabi do not print '=ABI' in help and use
2885 <option_value> format for -msve-vector-bits and -moverride
2888 2018-08-26 Jeff Law <law@redhat.com>
2890 * config/mips/frame-header-opt.c: Include "backend.h" rather than
2893 2018-08-26 Marek Polacek <polacek@redhat.com>
2895 PR c++/87029, Implement -Wredundant-move.
2896 * doc/invoke.texi: Document -Wredundant-move.
2898 2018-08-25 Martin Sebor <msebor@redhat.com>
2900 PR tree-optimization/87059
2901 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
2902 to the same type as the other.
2903 * fold-const.c (fold_binary_loc): Assert expectation.
2905 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
2907 * config/darwin.c (machopic_legitimize_pic_address): Clean up
2908 extraneous parentheses, dead code section and formatting.
2910 2018-08-24 David Malcolm <dmalcolm@redhat.com>
2913 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
2914 wide enough for jumps in the line-numbering to be visible.
2915 (layout::print_gap_in_line_numbering): New member function.
2916 (layout::calculate_line_spans): When using line numbering, merge
2917 line spans that are only 1 line apart.
2918 (diagnostic_show_locus): When printing line numbers, show gaps in
2919 line numbering directly, rather than printing headers.
2920 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
2921 line-numbering with multiple line spans.
2922 (selftest::test_fixit_insert_containing_newline_2): Add test of
2923 line-numbering, in which the spans are close enough to be merged.
2925 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
2927 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
2928 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
2929 * tree-vrp.c (range_is_nonnull): Remove.
2930 (range_includes_zero_p): Accept value_range instead of min/max.
2931 (extract_range_from_binary_expr_1): Do not early bail on
2933 Use range_includes_zero_p instead of range_is_nonnull.
2934 (extract_range_from_unary_expr): Use range_includes_zero_p instead
2935 of range_is_nonnull.
2936 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
2937 special case VR_ANTI_RANGE.
2938 (vrp_finalize): Same.
2939 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
2941 (range_is_nonnull): Remove.
2942 * vr-values.c (vrp_stmt_computes_nonzero): Use
2943 range_includes_zero_p instead of range_is_nonnull.
2944 (extract_range_basic): Pass value_range to range_includes_zero_p
2945 instead of range_is_nonnull.
2947 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
2949 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
2950 * emit-rtl.h (rtl_data): Remove return_bnd.
2951 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
2952 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
2953 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
2954 (POINTER_BOUNDS_MODE): Remove definition.
2955 (make_pointer_bounds_mode): Remove.
2956 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
2957 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
2958 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
2959 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
2960 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
2961 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
2962 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
2964 * config/i386/i386-modes.def (BND32, BND64): Remove.
2965 * config/i386/i386.c (dbx_register_map): Remove bound registers.
2966 (dbx64_register_map): Ditto.
2967 (svr4_dbx_register_map): Ditto.
2968 (indirect_thunk_bnd_needed): Remove.
2969 (indirect_thunks_bnd_used): Ditto.
2970 (indirect_return_bnd_needed): Ditto.
2971 (indirect_return_via_cx_bnd): Ditto.
2972 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
2973 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
2974 (output_indirect_thunk): Ditto. Remove need_prefix argument.
2975 (output_indirect_thunk_function): Remove handling of
2976 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
2977 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
2978 (ix86_save_reg): Remove handling of crtl->return_bnd.
2979 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
2980 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
2981 and UNSPEC_BNDLX_ADDR.
2982 (ix86_output_indirect_branch_via_reg): Remove handling of
2983 indirect_thunk_prefix_bnd.
2984 (ix86_output_indirect_branch_via_push): Ditto.
2985 (ix86_output_function_return): Ditto.
2986 (ix86_output_indirect_function_return): Ditto.
2987 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
2988 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
2989 (CALL_USED_REGISTERS): Ditto.
2990 (REG_ALLOC_ORDER): Update for removal of bound registers.
2991 (HI_REGISTER_NAMES): Ditto.
2992 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
2993 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
2994 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
2995 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
2996 (FIRST_PSEUDO_REG): Update.
2997 (BND): Remove mode iterator.
2998 * config/i386/predicates.md (bnd_mem_operator): Remove.
3000 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
3002 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
3005 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
3007 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
3008 the case in which the permute needs only a single element and
3009 repeats for every vector of the result. Extend that case to
3010 handle variable-length vectors.
3011 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3013 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
3016 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
3017 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
3019 2018-08-24 Richard Biener <rguenther@suse.de>
3021 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
3022 bb_flags_allocated members.
3023 (auto_flag): New RAII class for allocating flags.
3024 (auto_edge_flag): New RAII class for allocating edge flags.
3025 (auto_bb_flag): New RAII class for allocating bb flags.
3026 * cfgloop.c (verify_loop_structure): Allocate temporary edge
3028 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
3029 in favor of temporarily allocated BB flag.
3030 * hsa-brig.c: Re-order includes.
3031 * hsa-dump.c: Likewise.
3032 * hsa-regalloc.c: Likewise.
3033 * print-rtl.c: Likewise.
3034 * profile-count.c: Likewise.
3036 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
3039 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
3042 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3045 * wide-int-range.cc (wide_int_range_div): Do not ignore result
3046 from wide_int_range_multiplicative_op.
3048 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3050 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
3053 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
3055 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
3058 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
3060 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
3061 full GENERIC expression used for replacement.
3063 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
3065 * tree-vrp.c (abs_extent_range): Remove.
3066 (extract_range_into_wide_ints): Pass wide ints by reference.
3067 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
3068 Pass wide ints by reference in all calls to
3069 extract_range_into_wide_ints.
3070 * wide-int-range.cc (wide_int_range_div): New.
3071 * wide-int-range.h (wide_int_range_div): New.
3072 (wide_int_range_includes_zero_p): New.
3073 (wide_int_range_zero_p): New.
3075 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
3077 * config/aarch64/aarch64.md (arches): New enum.
3078 (arch): New enum attr.
3079 (arch_enabled): New attr.
3080 (enabled): Now uses arch_enabled only.
3081 (simd, sve, fp16): Removed attribute.
3082 (fp): Attr now defined in terms of 'arch'.
3083 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
3084 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
3085 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
3086 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
3087 attributes into 'arch'.
3088 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
3089 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
3090 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
3091 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
3092 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
3093 'simd' attribute into 'arch'.
3094 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
3095 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
3096 Convert use of 'fp' attribute to 'arch'.
3097 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
3098 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
3100 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
3101 (different modes) Merge 'fp' and 'simd' into 'arch'.
3102 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
3105 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
3107 PR rtl-optimization/87026
3108 * expmed.c (canonicalize_comparison): If we can no longer create
3109 pseudoregisters, don't.
3111 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
3114 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
3116 * config/arm/arm.c (speculation_barrier_libfunc): New static
3118 (arm_init_libfuncs): Initialize it.
3119 (arm_emit_speculation_barrier): New function.
3120 * config/arm/arm.md (speculation_barrier): Call
3121 arm_emit_speculation_barrier for architectures that do not have
3123 (speculation_barrier_insn): Only match on Armv7 or later.
3125 2018-08-23 Richard Biener <rguenther@suse.de>
3128 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
3131 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
3133 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
3134 of single-vector TBLs.
3135 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
3138 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
3141 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
3142 aarch64_evpc_tbl guard.
3144 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
3146 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
3149 2018-08-22 Martin Sebor <msebor@redhat.com>
3152 * tree-pretty-print.c (pretty_print_string): Add argument.
3153 (dump_generic_node): Call to pretty_print_string with string size.
3155 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
3157 PR rtl-optimization/86771
3158 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
3159 of two SETs into those two SETs, one to be placed at i2, if that SETs
3160 destination is modified between i2 and i3.
3162 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3164 PR tree-optimization/86725
3165 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
3167 (vect_analyze_scalar_cycles_1): Check it.
3169 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3171 PR tree-optimization/86725
3172 * tree-vect-loop.c (vect_is_simple_reduction): When treating
3173 an outer loop phi as a double reduction, make sure that the
3174 single user of the phi result is an inner loop phi.
3176 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3178 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
3179 grouped stores with gaps to a strided group.
3181 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3183 * tree-vect-stmts.c (get_group_load_store_type)
3184 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
3185 first statement in a group.
3187 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3189 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
3190 the sequence used in gcc/gcc.c.
3192 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3195 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
3198 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3200 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
3201 Darwin10-specific unwinder-shim.
3202 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
3203 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
3204 New to cater for Darwin10 Rosetta.
3206 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3208 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
3209 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
3211 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3216 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
3217 (dwarf2out_switch_text_section): Generate a local label for the
3218 second function sub-section and apply it as the second FDE start
3220 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
3221 second sub-section start.
3223 2018-08-22 Richard Biener <rguenther@suse.de>
3225 PR tree-optimization/86988
3226 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
3228 2018-08-22 Richard Biener <rguenther@suse.de>
3230 PR tree-optimization/86945
3231 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
3233 2018-08-22 Alexandre Oliva <oliva@adacore.com>
3235 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
3236 a comment about how uses of r2 for .sdata2 come about.
3238 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
3240 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
3242 2018-08-21 Marek Polacek <polacek@redhat.com>
3244 PR c++/86981, Implement -Wpessimizing-move.
3245 * doc/invoke.texi: Document -Wpessimizing-move.
3247 2018-08-21 Jan Hubicka <jh@suse.cz>
3249 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
3250 * tree.h (is_redundant_typedef): Remove.
3251 * dwarf2out.c (is_redundant_typedef): Turn into static function.
3253 2018-08-21 Jan Hubicka <jh@suse.cz>
3255 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
3258 2018-08-21 Tamar Christina <tamar.christina@arm.com>
3260 * expmed.c (extract_low_bits): Reject invalid subregs early.
3262 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
3265 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
3268 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3270 * config/vxworks.h: Guard vxworks_asm_out_constructor and
3271 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
3272 * config/vxworks.c: Likewise.
3274 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3276 * config/vxworks.c: Set targetm.have_ctors_dtors
3277 if HAVE_INITFINI_ARRAY_SUPPORT.
3278 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
3279 if HAVE_INITFINI_ARRAY_SUPPORT.
3281 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3283 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
3284 default search path for VxWorks < 7.
3286 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
3288 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
3289 (get_format_string): Refer to c_getstr.
3291 2018-08-21 Tom de Vries <tdevries@suse.de>
3293 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3294 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3295 (debuginfo_early_stop): Declare.
3296 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3297 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3298 (debuginfo_early_stop): New function.
3299 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
3300 and debuginfo_early_stop.
3301 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
3302 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
3303 (general_init): Call debuginfo_early_init.
3304 (finalize): Call debuginfo_fini.
3305 (do_compile): Call debuginfo_init.
3306 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
3308 (@item -fdump-debug, @item -fdump-earlydebug): Add.
3310 2018-08-21 Tom de Vries <tdevries@suse.de>
3312 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
3313 flag_dump_noaddr and flag_dump_unnumbered.
3315 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
3317 * wide-int-range.cc (wide_int_range_abs): New.
3318 (wide_int_range_order_set): Rename from wide_int_range_min_max.
3319 * wide-int-range.h (wide_int_range_abs): New.
3320 (wide_int_range_min_max): New.
3321 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
3322 case to call wide_int_range_abs.
3323 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
3324 (extract_range_from_abs_expr): Delete.
3326 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
3329 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
3330 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
3333 2018-08-20 Nathan Sidwell <nathan@acm.org>
3334 Jeff Law <law@redhat.com>
3336 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
3337 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
3339 2018-08-20 David Malcolm <dmalcolm@redhat.com>
3342 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
3343 (decl_attributes): Likewise.
3344 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
3346 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3347 * diagnostic-core.h (class auto_diagnostic_group): New class.
3348 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
3349 (diagnostic_report_diagnostic): Handle the first diagnostics within
3351 (emit_diagnostic): Add auto_diagnostic_group instance.
3353 (inform_n): Likewise.
3354 (warning): Likewise.
3355 (warning_at): Likewise.
3356 (warning_n): Likewise.
3357 (pedwarn): Likewise.
3358 (permerror): Likewise.
3360 (error_n): Likewise.
3361 (error_at): Likewise.
3363 (fatal_error): Likewise.
3364 (internal_error): Likewise.
3365 (internal_error_no_backtrace): Likewise.
3366 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
3367 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
3368 * diagnostic.h (struct diagnostic_context): Add fields
3369 "diagnostic_group_nesting_depth",
3370 "diagnostic_group_emission_count", "begin_group_cb",
3372 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
3373 Add auto_diagnostic_group instance(s).
3374 (find_explicit_erroneous_behavior): Likewise.
3375 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
3376 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
3377 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
3378 (gimplify_va_arg_expr): Likewise.
3379 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
3380 (HSA_SORRY_AT): Likewise.
3381 * ipa-devirt.c (compare_virtual_tables): Likewise.
3382 (warn_odr): Likewise.
3383 * multiple_target.c (expand_target_clones): Likewise.
3384 * opts-common.c (cmdline_handle_error): Likewise.
3385 * reginfo.c (globalize_reg): Likewise.
3386 * substring-locations.c (format_warning_n_va): Likewise.
3387 * tree-inline.c (expand_call_inline): Likewise.
3388 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
3389 * tree-ssa-loop-niter.c
3390 (do_warn_aggressive_loop_optimizations): Likewise.
3391 * tree-ssa-uninit.c (warn_uninit): Likewise.
3392 * tree.c (warn_deprecated_use): Likewise.
3394 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
3397 * config/i386/i386.md (eh_return): Always update EH return
3398 address in word_mode.
3400 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
3402 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
3403 TARGET_SPLIT_COMPLEX_ARG is defined.
3405 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3407 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
3409 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3412 * expr.c (expand_assignment): Assert that bitpos is positive.
3413 (store_field): Likewise
3414 (expand_expr_real_1): Make sure that bitpos is positive.
3415 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
3418 2018-08-20 Nathan Sidwell <nathan@acm.org>
3420 * Makefile.in (CPP_ID_DATA_H): Delete.
3421 (CPP_INTERNAL_H): Don't add it.
3422 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
3423 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
3425 2018-08-20 Richard Biener <rguenther@suse.de>
3427 PR tree-optimization/78655
3428 * tree-vrp.c (extract_range_from_binary_expr_1): Make
3429 pointer + offset nonnull if either operand is nonnull work.
3431 2018-08-20 Tom de Vries <tdevries@suse.de>
3433 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
3434 unless the referenced die describes the added property using
3435 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
3436 Otherwise, add a DW_AT_location to the referenced die.
3438 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
3441 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
3442 register_operand when calling ix86_set_reg_reg_cost.
3443 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
3444 Set *total to 0 for operands that satisfy x86_64_immediate_operand
3445 predicate and to 1 otherwise.
3447 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
3449 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
3450 emit a diagnostic that it is not supported and reset the option.
3451 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
3452 supported and consume the option. (ASM_FINAL_SPEC): New.
3454 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3456 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
3459 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
3461 C-SKY port: Documentation
3463 * doc/extend.texi (C-SKY Function Attributes): New section.
3464 * doc/invoke.texi (Option Summary): Add C-SKY options.
3465 (C-SKY Options): New section.
3466 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
3468 2018-08-17 Jojo <jijie_rong@c-sky.com>
3469 Huibin Wang <huibin_wang@c-sky.com>
3470 Sandra Loosemore <sandra@codesourcery.com>
3471 Chung-Lin Tang <cltang@codesourcery.com>
3473 C-SKY port: Backend implementation
3475 * config/csky/*: New.
3476 * common/config/csky/*: New.
3478 2018-08-17 Jojo <jijie_rong@c-sky.com>
3479 Huibin Wang <huibin_wang@c-sky.com>
3480 Sandra Loosemore <sandra@codesourcery.com>
3481 Chung-Lin Tang <cltang@codesourcery.com>
3482 Andrew Jenner <andrew@codesourcery.com>
3484 C-SKY port: Configury
3486 * config.gcc (csky-*-*): New.
3487 * configure.ac: Add csky to targets for dwarf2 debug_line support.
3488 * configure: Regenerated.
3490 2018-08-17 David Malcolm <dmalcolm@redhat.com>
3492 * dump-context.h: Include "dumpfile.h".
3493 (dump_context::dump_printf_va): Convert final param from va_list
3494 to va_list *. Convert from ATTRIBUTE_PRINTF to
3495 ATTRIBUTE_GCC_DUMP_PRINTF.
3496 (dump_context::dump_printf_loc_va): Likewise.
3497 * dumpfile.c: Include "stringpool.h".
3498 (make_item_for_dump_printf_va): Delete.
3499 (make_item_for_dump_printf): Delete.
3500 (class dump_pretty_printer): New class.
3501 (dump_pretty_printer::dump_pretty_printer): New ctor.
3502 (dump_pretty_printer::emit_items): New member function.
3503 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
3505 (dump_pretty_printer::emit_item): New member function.
3506 (dump_pretty_printer::stash_item): New member function.
3507 (dump_pretty_printer::format_decoder_cb): New member function.
3508 (dump_pretty_printer::decode_format): New member function.
3509 (dump_context::dump_printf_va): Reimplement in terms of
3510 dump_pretty_printer.
3511 (dump_context::dump_printf_loc_va): Convert final param from va_list
3513 (dump_context::begin_scope): Reimplement call to
3514 make_item_for_dump_printf.
3515 (dump_printf): Update for change to dump_printf_va.
3516 (dump_printf_loc): Likewise.
3517 (selftest::test_capture_of_dump_calls): Convert "stmt" from
3518 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
3519 with %T, %E, and %G.
3520 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
3521 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
3522 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
3523 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
3524 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
3525 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
3526 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
3527 within a dump_printf_loc call to "%wu".
3528 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
3529 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
3530 missing space after "=".
3531 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
3532 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
3533 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
3534 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
3535 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
3536 duplicate "vectorized" from message.
3538 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
3540 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
3541 polyNxK_t element's TYPE_STRING_FLAG.
3543 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3545 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
3546 (they were unnamed before). Fix comments.
3548 2018-08-17 Nathan Sidwell <nathan@acm.org>
3550 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
3552 2018-08-17 Richard Biener <rguenther@suse.de>
3554 PR tree-optimization/86841
3555 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
3557 2018-08-17 Martin Liska <mliska@suse.cz>
3559 * common.opt: Remove Warn, Init and Report for options with
3560 Ignore/Deprecated flag. Warning is done automatically for
3562 * config/i386/i386.opt: Likewise.
3563 * config/ia64/ia64.opt: Likewise.
3564 * config/rs6000/rs6000.opt: Likewise.
3565 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
3566 Remove usage of flag_check_pointer_bounds.
3567 * lto-wrapper.c (merge_and_complain): Do not handle
3568 OPT_fcheck_pointer_bounds.
3569 (append_compiler_options): Likewise.
3570 * opt-functions.awk: Do not handle Deprecated.
3571 * optc-gen.awk: Check that Var, Report and Init are not
3572 used for an option with Ignore/Deprecated flag.
3573 * opts-common.c (decode_cmdline_option): Do not report
3575 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
3577 * opts.h (struct cl_option): Remove cl_deprecated flag.
3578 (CL_ERR_DEPRECATED): Remove error enum value.
3580 2018-08-17 Richard Biener <rguenther@suse.de>
3583 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
3584 across a va-arg-pack using call adjust its return value accordingly.
3586 2018-08-16 Martin Sebor <msebor@redhat.com>
3588 PR tree-optimization/86853
3589 * gimple-ssa-sprintf.c (struct format_result): Rename member.
3590 (struct fmtresult): Add member and initialize it in ctors.
3591 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
3592 (format_string): Handle %S the same as %ls. Set MAYFAIL.
3593 (format_directive): Set POSUNDER4K when MAYFAIL is set.
3594 (parse_directive): Handle %C same as %c.
3595 (sprintf_dom_walker::compute_format_length): Adjust.
3596 (is_call_safe): Adjust.
3598 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3600 * builtins.c (c_strlen): Add new parameter eltsize. Use it
3601 for determining how to count the elements.
3602 * builtins.h (c_strlen): Adjust prototype.
3603 * expr.c (string_constant): Add new parameter mem_size.
3604 Set *mem_size appropriately.
3605 * expr.h (string_constant): Adjust protoype.
3606 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
3607 * gimple-fold.h (get_range_strlen): Adjust prototype.
3608 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
3609 (format_string): Call get_string_length with eltsize.
3611 2018-08-16 David Malcolm <dmalcolm@redhat.com>
3613 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
3614 to emit the span, rather than setting it as the prefix.
3616 2018-08-16 David Malcolm <dmalcolm@redhat.com>
3618 * diagnostic-show-locus.c (layout::start_annotation_line): Add
3619 "margin_char" parameter, defaulting to space. Use it in place
3620 of pp_space for the initial part of the margin.
3621 (layout::print_leading_fixits): Use '+' when filling the margin
3622 of line-insertion fix-it hints.
3624 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3626 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
3629 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3631 * config/rs6000/altivec.md: Don't set length attribute to the default
3633 * config/rs6000/darwin.md: Ditto.
3634 * config/rs6000/dfp.md: Ditto.
3635 * config/rs6000/htm.md: Ditto.
3636 * config/rs6000/rs6000.md: Ditto.
3637 * config/rs6000/sync.md: Ditto.
3638 * config/rs6000/vsx.md: Ditto.
3640 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3642 * config/rs6000/altivec.md: Don't set length attribute to the default
3643 value, for branch instructions.
3644 * config/rs6000/darwin.md: Ditto.
3645 * config/rs6000/rs6000.md: Ditto.
3647 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3649 * config/rs6000/rs6000.md (length): Always define as const_int 4.
3650 (unnamed conditional branch define_insn): Set length to 4 or 8
3651 depending on offset.
3652 (<bd>_<mode>): Similar, for alternative 0.
3653 (<bd>tf_<mode>): Ditto.
3655 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3657 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
3659 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
3661 * doc/rtl.texi: Replace old RTX class names with new names.
3664 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
3666 * expmed.h (canonicalize_comparison): New declaration.
3667 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
3668 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
3669 * optabs.c (prepare_cmp_insn): Likewise.
3670 * rtl.h (unsigned_condition_p): New function which checks if a
3671 comparison operator is unsigned.
3673 2018-08-16 Nathan Sidwell <nathan@acm.org>
3675 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
3676 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
3678 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3681 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
3682 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
3683 (mov<mov>): ..this and enable unconditionally.
3685 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3687 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
3689 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
3691 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
3692 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
3693 "Common" with "Target".
3695 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
3697 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
3698 * doc/invoke.texi (mmitigate-rop): Remove.
3699 * config/i386/i386.c: Do not include "regrename.h".
3700 (ix86_rop_should_change_byte_p, reg_encoded_number)
3701 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
3703 (ix86_reorg): Remove call to ix86_mitigate_rop.
3704 * config/i386/i386.md (attr "modrm_class"): Remove.
3705 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
3706 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
3707 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
3709 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
3711 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
3712 allow folding of mergeh() and mergel() for the float and double types.
3713 (fold_mergehl_helper): Rework to handle building a permute tree
3716 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
3718 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
3721 2018-08-15 David Malcolm <dmalcolm@redhat.com>
3723 * common.opt (fdiagnostics-show-labels): New option.
3724 * diagnostic-show-locus.c (class layout_range): Add field
3726 (class layout): Add field "m_show_labels_p".
3727 (layout_range::layout_range): Add param "label" and use it to
3729 (make_range): Pass in NULL for new "label" param of layout_range's
3731 (layout::layout): Initialize m_show_labels_p.
3732 (layout::maybe_add_location_range): Pass in loc_range->m_label
3733 when constructing layout_range instances.
3734 (struct line_label): New struct.
3735 (layout::print_any_labels): New member function.
3736 (layout::print_line): Call it if label-printing is enabled.
3737 (selftest::test_one_liner_labels): New test.
3738 (selftest::test_diagnostic_show_locus_one_liner): Call it.
3739 * diagnostic.c (diagnostic_initialize): Initialize
3740 context->show_labels_p.
3741 * diagnostic.h (struct diagnostic_context): Add field
3743 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3744 -fno-diagnostics-show-labels.
3745 * dwarf2out.c (gen_producer_string): Add
3746 OPT_fdiagnostics_show_labels to the ignored options.
3747 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
3749 (gcc_rich_location::maybe_add_expr): Likewise.
3750 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
3751 label" param, defaulting to NULL.
3752 (gcc_rich_location::add_expr): Add "label" param.
3753 (gcc_rich_location::maybe_add_expr): Likewise.
3754 (class text_range_label): New class.
3755 (class range_label_for_type_mismatch): New class.
3756 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
3757 of format_warning_va.
3758 (fmtwarn_n): Likewise for new params of format_warning_n_va.
3759 * lto-wrapper.c (merge_and_complain): Add
3760 OPT_fdiagnostics_show_labels to the "pick one setting" options.
3761 (append_compiler_options): Likewise to the dropped options.
3762 (append_diag_options): Likewise to the passed-on options.
3763 * opts.c (common_handle_option): Handle the new option.
3764 * selftest-diagnostic.c
3765 (test_diagnostic_context::test_diagnostic_context): Enable
3767 * substring-locations.c: Include "gcc-rich-location.h".
3768 (format_warning_n_va): Add "fmt_label" and "param_label" params
3769 and use them as appropriate.
3770 (format_warning_va): Add "fmt_label" and "param_label" params,
3771 passing them on to format_warning_n_va.
3772 (format_warning_at_substring): Likewise.
3773 (format_warning_at_substring_n): Likewise.
3774 * substring-locations.h (format_warning_va): Add "fmt_label" and
3775 "param_label" params.
3776 (format_warning_n_va): Likewise.
3777 (format_warning_at_substring): Likewise.
3778 (format_warning_at_substring_n): Likewise.
3779 * toplev.c (general_init): Initialize global_dc->show_labels_p.
3781 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
3784 * builtins.c (expand_builtin_memcmp): Do not expand the call
3785 when overflow is detected.
3787 2018-08-15 Martin Sebor <msebor@redhat.com>
3789 PR tree-optimization/71625
3790 * config/aarch64/aarch64-builtins.c
3791 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3793 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
3795 * config/s390/s390.c (s390_reorg): Remove loop.
3797 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3800 (darwin_function_switched_text_sections): Delete.
3801 * gcc/config/darwin.h
3802 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
3804 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
3807 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
3808 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
3809 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
3811 2018-08-15 Martin Liska <mliska@suse.cz>
3813 PR tree-optimization/86925
3814 * predict.c (expr_expected_value_1): When taking
3815 later predictor, assign also probability.
3816 Use fold_build2_initializer_loc in order to fold
3817 the expression in -frounding-math.
3819 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
3821 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
3823 (expand_vec_perm_1): Try the new method.
3825 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
3828 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
3829 Check whether lra_live_max_point is 0 before dividing.
3831 2018-08-14 Martin Sebor <msebor@redhat.com>
3833 PR tree-optimization/86650
3834 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
3835 (vrp_prop::check_mem_ref): Same.
3837 2018-08-13 Liu Hao <lh_mouse@126.com>
3839 * pretty-print.c (eat_esc_sequence): Swap the foreground and
3840 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
3841 and clear it thereafter, as it only works for DBCS.
3843 2018-08-13 Liu Hao <lh_mouse@126.com>
3845 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
3846 handle returned by _get_osf_handle().
3848 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3850 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
3851 for folding vec_perm.
3853 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3855 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
3856 Add support for gimple-folding of vec_pack() and vec_unpack()
3859 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
3861 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
3862 vec_xst variants to the list.
3863 (rs6000_gimple_fold_builtin): Add support for folding unaligned
3864 vector loads and stores.
3866 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
3868 * config.gcc (rs6000-ibm-aix4.x): Delete.
3869 (rs6000-ibm-aix5.1): Delete.
3870 (rs6000-ibm-aix5.2): Delete.
3871 (rs6000-ibm-aix5.3): Delete.
3872 * config/rs6000/aix43.h: Delete.
3873 * config/rs6000/aix51.h: Delete.
3874 * config/rs6000/aix52.h: Delete.
3875 * config/rs6000/t-aix43: Delete.
3877 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
3879 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3881 (s390_decompose_address): Factor out constant pool ref
3884 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3886 * config/nds32/nds32-predicates.c
3887 (nds32_can_use_bclr_p): Change return type as bool.
3888 (nds32_can_use_bset_p): Ditto.
3889 (nds32_can_use_btgl_p): Ditto.
3890 (nds32_can_use_bitci_p): Ditto.
3891 * config/nds32/nds32-protos.h
3892 (nds32_can_use_bclr_p): Change declaration.
3893 (nds32_can_use_bset_p): Ditto.
3894 (nds32_can_use_btgl_p): Ditto.
3895 (nds32_can_use_bitci_p): Ditto.
3897 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3899 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
3900 Support -msched-prolog-epilog option.
3901 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
3903 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3905 * common/config/nds32/nds32-common.c
3906 (nds32_option_optimization_table): Enalbe -malways-align.
3908 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
3912 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
3913 OPT_misr_secure_ case.
3914 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
3915 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
3916 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
3918 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
3919 (nds32_isr_info): New field security_level.
3920 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
3921 * config/nds32/nds32.md (return_internal): Consider critical attribute.
3922 * config/nds32/nds32.opt (misr-secure): New option.
3923 * config/nds32/nds32_init.inc: New file.
3924 * config/nds32/nds32_isr.h: New file.
3926 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
3928 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
3929 Update comment for atomic instructions.
3930 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
3931 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
3933 (atomic_loaddi): Revise fence expansion to only emit fence prior to
3934 load for __ATOMIC_SEQ_CST model.
3935 (atomic_loaddi_1): Remove float register target.
3936 (atomic_storedi): Handle CONST_INT values.
3937 (atomic_storedi_1): Remove float register source. Add special case
3939 (memory_barrier): New expander and insn.
3941 2018-08-11 Jakub Jelinek <jakub@redhat.com>
3943 PR tree-optimization/86835
3944 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
3945 new_stmt after def_gsi, make sure to insert new_square_stmt after
3946 that stmt, not 2 stmts before it.
3948 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
3951 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
3954 2018-08-10 Martin Liska <mliska@suse.cz>
3957 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
3959 * builtins.c (expand_builtin_expect_with_probability):
3961 (expand_builtin_expect_with_probability): New function.
3962 (build_builtin_expect_predicate): Add new argumnet probability
3963 for BUILT_IN_EXPECT_WITH_PROBABILITY.
3964 (fold_builtin_expect):
3967 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
3968 * builtins.h (fold_builtin_expect): Set new argument.
3969 * doc/extend.texi: Document __builtin_expect_with_probability.
3970 * doc/invoke.texi: Likewise.
3971 * gimple-fold.c (gimple_fold_call): Pass new argument.
3972 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
3973 also BUILT_IN_EXPECT_WITH_PROBABILITY.
3974 * predict.c (get_predictor_value): New function.
3975 (expr_expected_value): Add new argument probability. Assume
3976 that predictor and probability are always non-null.
3977 (expr_expected_value_1): Likewise. For __builtin_expect and
3978 __builtin_expect_with_probability set probability. Handle
3979 combination in binary expressions.
3980 (tree_predict_by_opcode): Simplify code by simply calling
3981 get_predictor_value.
3982 (pass_strip_predict_hints::execute): Add handling of
3983 BUILT_IN_EXPECT_WITH_PROBABILITY.
3984 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
3986 * tree.h (DECL_BUILT_IN_P): New function.
3988 2018-08-10 Martin Liska <mliska@suse.cz>
3990 PR tree-optimization/85799
3991 * passes.def: Add argument for pass_strip_predict_hints.
3992 * predict.c (class pass_strip_predict_hints): Add new argument
3994 (strip_predictor_early): New function.
3995 (pass_strip_predict_hints::execute): Call the function to
3997 (strip_predict_hints): New function.
3998 * predict.def: Fix comment.
4000 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
4002 * Makefile.in: Clarify which tm.texi to copy over to assert the
4003 right to grant a GFDL license for all.
4005 2018-08-09 Jeff Law <law@redhat.com>
4007 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
4010 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
4012 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
4015 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4017 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
4018 reductions for variable-length vectors.
4020 2018-08-09 David Malcolm <dmalcolm@redhat.com>
4023 * common.opt (fdiagnostics-show-line-numbers): New option.
4024 * diagnostic-show-locus.c (class layout): Add fields
4025 "m_show_line_numbers_p" and "m_linenum_width";
4026 (num_digits): New function.
4027 (test_num_digits): New function.
4028 (layout::layout): Initialize new fields. Update m_x_offset
4029 logic to handle any left margin.
4030 (layout::print_source_line): Print line number when requested.
4031 (layout::start_annotation_line): New member function.
4032 (layout::print_annotation_line): Call it.
4033 (layout::print_leading_fixits): Likewise.
4034 (layout::print_trailing_fixits): Likewise. Update calls to
4035 move_to_column for new parameter.
4036 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
4037 it to potentially call start_annotation_line.
4038 (layout::show_ruler): Call start_annotation_line.
4039 (selftest::test_line_numbers_multiline_range): New selftest.
4040 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
4041 and selftest::test_line_numbers_multiline_range.
4042 * diagnostic.c (diagnostic_initialize): Initialize
4043 show_line_numbers_p.
4044 * diagnostic.h (struct diagnostic_context): Add field
4045 "show_line_numbers_p".
4046 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4047 -fno-diagnostics-show-line-numbers.
4048 * dwarf2out.c (gen_producer_string): Add
4049 OPT_fdiagnostics_show_line_numbers to the ignored options.
4050 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
4051 one setting" options.
4052 (append_compiler_options): Likewise to the dropped options.
4053 (append_diag_options): Likewise to the passed-on options.
4054 * opts.c (common_handle_option): Handle the new option.
4055 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
4057 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
4059 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
4060 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
4061 third argument of type "const signed char" to descriptions of
4062 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
4063 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
4064 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
4065 __builtin_bcdsub_ov functions.
4067 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4069 PR tree-optimization/86858
4070 * tree-vect-loop.c (vect_is_simple_reduction): Restore
4071 flow_bb_inside_loop_p calls.
4073 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4075 PR tree-optimization/86871
4076 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
4077 instead of gimple_assign_lhs.
4079 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
4082 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
4083 register constraint to operand 0.
4084 (add<mode>3_carryinC): Likewise.
4085 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
4087 2018-08-09 Martin Liska <mliska@suse.cz>
4090 * common.opt: Remove extra line.
4092 2018-08-09 Martin Liska <mliska@suse.cz>
4094 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
4095 at the end of a line, make first letter capital and end up
4096 a sentence with a dot.
4097 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
4098 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
4099 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
4100 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
4101 (PARAM_MAX_ISL_OPERATIONS): Likewise.
4102 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
4103 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
4104 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
4105 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
4106 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
4107 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
4108 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
4109 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
4110 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
4111 (PARAM_TREE_REASSOC_WIDTH): Likewise.
4112 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
4113 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
4114 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
4116 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
4119 * config/s390/s390.c (s390_option_override_internal): Reduce the
4120 stack-clash-protection-probe-interval param if it would be too big
4123 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
4126 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
4127 * config/m68k/m68k.c (handle_move_double): Don't call
4128 m68k_final_prescan_insn.
4129 (m68k_adjust_decorated_operand): Renamed from
4130 m68k_final_prescan_insn, remove first and third operand and
4132 (print_operand): Call it.
4133 (print_operand_address): Call it.
4135 2018-08-08 Nathan Sidwell <nathan@acm.org>
4137 * diagnostic.c (diagnostic_report_current_module): Use
4138 linemap_included_from & linemap_included_from_linemap.
4140 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
4142 * config/aarch64/aarch64-cores.def: Add phecda core.
4143 * config/aarch64/aarch64-tune.md: Regenerate.
4144 * doc/invoke.texi: Add phecda core.
4146 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
4149 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
4151 * config/s390/s390.md ("movti"): Add more alternatives for
4152 constant to GPR copies.
4154 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
4156 * config/s390/s390.c: Fix whitespace damage throughout the file.
4157 * config/s390/s390.h: Likewise.
4158 * config/s390/tpf.h: Likewise.
4160 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
4162 * config/s390/s390.c (s390_loadrelative_operand_p):
4163 Remove TARGET_CPU_ZARCH usages.
4164 (s390_rtx_costs): Likewise.
4165 (s390_legitimate_constant_p): Likewise.
4166 (s390_cannot_force_const_mem): Likewise.
4167 (legitimate_reload_constant_p): Likewise.
4168 (s390_preferred_reload_class): Likewise.
4169 (legitimize_pic_address): Likewise.
4170 (legitimize_tls_address): Likewise.
4171 (s390_split_branches): Removed.
4172 (s390_add_execute): Removed.
4173 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
4174 (s390_mainpool_start): Likewise.
4175 (s390_mainpool_finish): Likewise.
4176 (s390_mainpool_cancel): Removed.
4177 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
4178 (s390_chunkify_cancel): Likewise.
4179 (s390_return_addr_rtx): Likewise.
4180 (s390_register_info): Remove split_branches_pending_p uages.
4181 (s390_optimize_register_info): Likewise.
4182 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
4183 split_branches_pending_p usages.
4184 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
4185 (s390_load_got): Likewise.
4186 (s390_expand_split_stack_prologue): Likewise.
4187 (output_asm_nops): Likewise.
4188 (s390_function_profiler): Likewise.
4189 (s390_emit_call): Likewise.
4190 (s390_conditional_register_usage): Likewise.
4191 (s390_optimize_prologue): Likewise.
4192 (s390_reorg): Remove TARGET_CPU_ZARCH and
4193 split_branches_pending_p usages.
4194 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
4196 (s390_output_indirect_thunk_function): Likewise.
4197 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
4198 (TARGET_CPU_ZARCH_P): Removed.
4199 (struct machine_function): Remove split_branches_pending_p.
4200 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
4202 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
4204 * common/config/s390/s390-common.c (processor_flags_table):
4206 * config.gcc: Remove with_arch/with_tune support.
4207 * config/s390/2064.md: Remove cpu attribute comparisons.
4208 * config/s390/driver-native.c (s390_host_detect_local_cpu):
4210 * config/s390/linux.h (ASM_SPEC):
4211 Remove -march support.
4212 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
4213 Use a table to get an arch level.
4214 * config/s390/s390-opts.h (enum processor_type):
4216 * config/s390/s390.c
4217 (processor_table): Remove entries, add arch_level values.
4218 (s390_issue_rate): Remove cases.
4219 (s390_option_override): Adjust
4220 s390_option_override_internal() call.
4221 (s390_option_override_internal): Remove deprecation warning.
4222 (s390_valid_target_attribute_tree): Adjust
4223 s390_option_override_internal() call.
4224 * config/s390/s390.h (struct s390_processor):
4225 Share with s390-c.c, add arch_level field.
4226 * config/s390/s390.md:
4227 Remove occurrences in cpu attribute.
4228 * config/s390/s390.opt: Remove -march/-mtune support.
4229 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
4230 * doc/invoke.texi: Remove deprecation warning.
4232 2018-08-08 Luis Machado <luis.machado@linaro.org>
4234 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
4236 (qdf24xx_tunings): Set vector cost structure to
4237 qdf24xx_vector_cost.
4239 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
4240 <register_sextend>: Set to 3.
4242 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
4245 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
4246 * config/aarch64/aarch64-simd.md
4247 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
4248 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
4249 (aarch64_frecpx<mode>): ...this new pattern.
4250 * config/aarch64/aarch64-simd-builtins.def: Remove comment
4251 about aarch64_frecp<FRECP:frecp_suffix><mode>.
4253 2018-08-07 Martin Liska <mliska@suse.cz>
4256 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
4257 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
4258 * predict.def (PRED_MALLOC_NONNULL): New predictor.
4259 * doc/extend.texi: Document that malloc attribute adds
4262 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
4265 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4266 Define to speculation_safe_value_not_needed.
4268 2018-08-06 Jeff Law <law@redhat.com>
4270 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
4271 the vr_values instance to cprop_into_stmt.
4272 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
4273 (cprop_operand): Also query EVRP to determine if OP is a constant.
4275 2018-08-06 Nathan Sidwell <nathan@acm.org>
4277 * diagnostic.c (diagnostic_report_current_module): Reroll
4278 included-at loop. Translate text.
4280 2018-08-06 David Malcolm <dmalcolm@redhat.com>
4282 * function-tests.c (selftest::test_expansion_to_rtl): Call
4283 free_after_compilation.
4285 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4287 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
4289 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
4291 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
4292 loops with memory block operations from getting unrolled.
4294 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
4297 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4298 Define to speculation_safe_value_not_needed.
4300 2018-08-06 Jeff Law <law@redhat.com>
4302 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
4305 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4308 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
4309 with all enabled __intN types.
4311 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
4313 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4315 * alias.c (record_set): Check for clobber high.
4316 * cfgexpand.c (expand_gimple_stmt): Likewise.
4317 * combine-stack-adj.c (single_set_for_csa): Likewise.
4318 * combine.c (find_single_use_1): Likewise.
4319 (set_nonzero_bits_and_sign_copies): Likewise.
4320 (get_combine_src_dest): Likewise.
4321 (is_parallel_of_n_reg_sets): Likewise.
4322 (try_combine): Likewise.
4323 (record_dead_and_set_regs_1): Likewise.
4324 (reg_dead_at_p_1): Likewise.
4325 (reg_dead_at_p): Likewise.
4326 * dce.c (deletable_insn_p): Likewise.
4327 (mark_nonreg_stores_1): Likewise.
4328 (mark_nonreg_stores_2): Likewise.
4329 * df-scan.c (df_find_hard_reg_defs): Likewise.
4330 (df_uses_record): Likewise.
4331 (df_get_call_refs): Likewise.
4332 * dwarf2out.c (mem_loc_descriptor): Likewise.
4333 * haifa-sched.c (haifa_classify_rtx): Likewise.
4334 * ira-build.c (create_insn_allocnos): Likewise.
4335 * ira-costs.c (scan_one_insn): Likewise.
4336 * ira.c (equiv_init_movable_p): Likewise.
4337 (rtx_moveable_p): Likewise.
4338 (interesting_dest_for_shprep): Likewise.
4339 * jump.c (mark_jump_label_1): Likewise.
4340 * postreload-gcse.c (record_opr_changes): Likewise.
4341 * postreload.c (reload_cse_simplify): Likewise.
4342 (struct reg_use): Add source expr.
4343 (reload_combine): Check for clobber high.
4344 (reload_combine_note_use): Likewise.
4345 (reload_cse_move2add): Likewise.
4346 (move2add_note_store): Likewise.
4347 * print-rtl.c (print_pattern): Likewise.
4348 * recog.c (decode_asm_operands): Likewise.
4349 (store_data_bypass_p): Likewise.
4350 (if_test_bypass_p): Likewise.
4351 * regcprop.c (kill_clobbered_value): Likewise.
4352 (kill_set_value): Likewise.
4353 * reginfo.c (reg_scan_mark_refs): Likewise.
4354 * reload1.c (maybe_fix_stack_asms): Likewise.
4355 (eliminate_regs_1): Likewise.
4356 (elimination_effects): Likewise.
4357 (mark_not_eliminable): Likewise.
4358 (scan_paradoxical_subregs): Likewise.
4359 (forget_old_reloads_1): Likewise.
4360 * reorg.c (find_end_label): Likewise.
4361 (try_merge_delay_insns): Likewise.
4362 (redundant_insn): Likewise.
4363 (own_thread_p): Likewise.
4364 (fill_simple_delay_slots): Likewise.
4365 (fill_slots_from_thread): Likewise.
4366 (dbr_schedule): Likewise.
4367 * resource.c (update_live_status): Likewise.
4368 (mark_referenced_resources): Likewise.
4369 (mark_set_resources): Likewise.
4370 * rtl.c (copy_rtx): Likewise.
4371 * rtlanal.c (reg_referenced_p): Likewise.
4372 (single_set_2): Likewise.
4373 (noop_move_p): Likewise.
4374 (note_stores): Likewise.
4375 * sched-deps.c (sched_analyze_reg): Likewise.
4376 (sched_analyze_insn): Likewise.
4378 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4380 * cse.c (invalidate_reg): New function extracted from...
4381 (invalidate): ...here.
4382 (canonicalize_insn): Check for clobber high.
4383 (invalidate_from_clobbers): invalidate clobber highs.
4384 (invalidate_from_sets_and_clobbers): Likewise.
4385 (count_reg_usage): Check for clobber high.
4386 (insn_live_p): Likewise.
4387 * cselib.c (cselib_expand_value_rtx_1):Likewise.
4388 (cselib_invalidate_regno): Check for clobber in setter.
4389 (cselib_invalidate_rtx): Pass through setter.
4390 (cselib_invalidate_rtx_note_stores):
4391 (cselib_process_insn): Check for clobber high.
4392 * cselib.h (cselib_invalidate_rtx): Add operand.
4394 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4396 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
4397 (mark_not_eliminable): Likewise.
4398 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
4399 * lra-lives.c (process_bb_lives): Check for clobber high.
4400 * lra.c (new_insn_reg): Remember clobber highs.
4401 (collect_non_operand_hard_regs): Check for clobber high.
4402 (lra_set_insn_recog_data): Likewise.
4403 (add_regs_to_insn_regno_info): Likewise.
4404 (lra_update_insn_regno_info): Likewise.
4406 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4408 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
4409 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
4411 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4413 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
4414 (copy_insn_1): Likewise.
4415 (gen_hard_reg_clobber_high): New gen function.
4416 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
4417 * genemit.c (gen_exp): Likewise.
4418 (gen_emit_seq): Pass through info.
4419 (gen_insn): Check for CLOBBER_HIGH.
4420 (gen_expand): Pass through info.
4421 (gen_split): Likewise.
4422 (output_add_clobbers): Likewise.
4423 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
4424 (remove_clobbers): Likewise.
4425 * rtl.h (gen_hard_reg_clobber_high): New declaration.
4427 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4429 * doc/rtl.texi (clobber_high): Add.
4430 (parallel): Add in clobber high
4431 * rtl.c (rtl_check_failed_code3): Add function.
4432 * rtl.def (CLOBBER_HIGH): Add expression.
4433 * rtl.h (RTL_CHECKC3): Add macro.
4434 (rtl_check_failed_code3): Add declaration.
4435 (XC3EXP): Add macro.
4437 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
4440 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
4441 cfun->machine->max_used_stack_alignment if needed.
4443 2018-08-04 Martin Sebor <msebor@redhat.com>
4445 PR tree-optimization/86571
4446 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
4449 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
4452 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4455 2018-08-03 Jeff Law <law@redhat.com>
4458 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4459 Define to speculation_safe_value_not_needed.
4461 2018-08-03 David Malcolm <dmalcolm@redhat.com>
4463 * doc/gcov.texi (-x): Remove duplicate "to".
4464 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
4465 (-Wif-not-aligned): Remove duplicate "is".
4466 (-flto): Remove duplicate "the".
4467 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
4468 duplicate "v5.00.b".
4469 (MSP430 Options): Remove duplicate "and" from the description
4470 of "-mgprel-sec=regexp".
4471 (x86 Options): Remove duplicate copies of "vmldLog102" and
4472 vmlsLog104 from description of "-mveclibabi=type".
4474 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
4476 * internal-fn.h (first_commutative_argument): Declare.
4477 * internal-fn.c (first_commutative_argument): New function.
4478 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
4479 restrictions for pattern statements. Use first_commutative_argument
4480 to look for commutative operands in calls to internal functions.
4482 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
4484 * Makefile.in (wide-int-range.o): New.
4485 * tree-vrp.c: Move all the wide_int_* functions to...
4486 * wide-int-range.cc: ...here.
4487 * tree-vrp.h: Move all the wide_int_* prototypes to...
4488 * wide-int-range.h: ...here.
4490 2018-08-03 Tom de Vries <tdevries@suse.de>
4492 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
4494 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
4495 * except.c (output_function_exception_table): Do early exit if
4496 targetm_common.except_unwind_info (&global_options) == UI_NONE.
4498 2018-08-03 Martin Liska <mliska@suse.cz>
4500 * predict.c (dump_prediction): Change to 2 digits
4503 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
4505 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
4507 (falkor_am_1_gtov_gtov): ... a new insn reservation.
4509 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
4511 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
4512 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
4513 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
4514 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
4515 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
4517 2018-08-02 David Malcolm <dmalcolm@redhat.com>
4519 * diagnostic-show-locus.c (diagnostic_show_locus): Use
4520 pp_take_prefix when saving the existing prefix.
4521 * diagnostic.c (diagnostic_append_note): Likewise.
4522 * langhooks.c (lhd_print_error_function): Likewise.
4523 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
4524 param's type. Free the existing prefix.
4525 (pp_take_prefix): New function.
4526 (pretty_printer::pretty_printer): Drop the prefix parameter.
4527 Rename the length parameter to match the comment.
4528 (pretty_printer::~pretty_printer): Free the prefix.
4529 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
4531 (struct pretty_printer): Drop the "const" from "prefix" field's
4532 type and clarify memory management.
4533 (pp_set_prefix): Drop the "const" from the 2nd param.
4534 (pp_take_prefix): New decl.
4536 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4538 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
4539 for word_mode_ok here instead of passing as argument.
4540 (expand_block_compare): Change select_block_compare_mode() call.
4541 (expand_strncmp_gpr_sequence): New function.
4542 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
4544 2018-08-02 Jeff Law <law@redhat.com>
4547 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4548 Define to speculation_safe_value_not_needed.
4551 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4552 Define to speculation_safe_value_not_needed.
4554 2018-08-02 Tom de Vries <tdevries@suse.de>
4557 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
4558 function. Return UI_TARGET unconditionally.
4559 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
4560 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
4562 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4564 * genemit.c (print_overload_test): Fix typo.
4566 2018-08-02 Richard Biener <rguenther@suse.de>
4568 PR tree-optimization/86816
4569 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
4570 which checks for value availability before querying it.
4571 (gvn_uses_equal): Use it.
4572 (same_succ_hash): Likewise.
4573 (gimple_equal_p): Likewise.
4575 2018-08-02 Nick Clifton <nickc@redhat.com>
4578 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4579 Define to speculation_safe_value_not_needed.
4582 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4583 Define to speculation_safe_value_not_needed.
4586 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4587 Define to speculation_safe_value_not_needed.
4590 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4591 Define to speculation_safe_value_not_needed.
4594 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4595 Define to speculation_safe_value_not_needed.
4598 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4599 Define to speculation_safe_value_not_needed.
4602 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4603 Define to speculation_safe_value_not_needed.
4606 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4607 Define to speculation_safe_value_not_needed.
4610 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4611 speculation_safe_value_not_needed.
4614 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
4615 to speculation_safe_value_not_needed.
4617 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4619 * doc/md.texi: Expand the documentation of instruction names
4620 to mention port-local uses. Document '@' in pattern names.
4621 * read-md.h (overloaded_instance, overloaded_name): New structs.
4623 (md_reader::handle_overloaded_name): New member function.
4624 (md_reader::get_overloads): Likewise.
4625 (md_reader::m_first_overload): New member variable.
4626 (md_reader::m_next_overload_ptr): Likewise.
4627 (md_reader::m_overloads_htab): Likewise.
4628 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
4629 m_next_overload_ptr and m_overloads_htab.
4630 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
4631 (get_mode_token, get_code_token, get_int_token): New functions.
4632 (map_attr_string): Add an optional argument that passes back
4633 the associated iterator.
4634 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
4635 (md_reader::handle_overloaded_name, add_overload_instance): New
4637 (apply_iterators): Handle '@' names. Report an error if '@'
4638 is used without iterators.
4639 (initialize_iterators): Initialize the new iterator_group fields.
4640 * genopinit.c (handle_overloaded_code_for)
4641 (handle_overloaded_gen): New functions.
4642 (main): Use them to print declarations of maybe_code_for_* and
4643 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
4644 * genemit.c (print_overload_arguments, print_overload_test)
4645 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
4646 (main): Use it to print definitions of maybe_code_for_* and
4647 maybe_gen_* functions.
4648 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
4649 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
4650 instead of explicit mode checks.
4651 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
4652 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
4653 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
4654 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
4655 (aarch64_expand_compare_and_swap): Likewise
4656 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
4657 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
4658 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
4659 (aarch64_constant_pool_reload_icode): Delete.
4660 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
4661 instead of aarch64_constant_pool_reload_icode. Use
4662 code_for_aarch64_reload_mov instead of explicit mode checks.
4663 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
4664 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
4665 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
4666 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
4667 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
4668 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
4669 (aarch64_atomic_load_op_code): Delete.
4670 (aarch64_emit_atomic_load_op): Likewise.
4671 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
4672 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
4673 instead of aarch64_emit_atomic_load_op.
4674 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
4675 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
4676 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
4677 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
4678 character before the pattern name.
4679 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
4680 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
4681 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
4682 (aarch64_frecps<mode>): Likewise.
4683 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
4684 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
4685 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
4686 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
4687 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
4689 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4691 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
4692 Allow HFmode constants if TARGET_FP_F16INST.
4694 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
4697 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
4698 No longer check last store for clobber of address register.
4700 2018-08-02 Martin Liska <mliska@suse.cz>
4702 PR gcov-profile/86817
4703 * gcov.c (process_all_functions): New function.
4705 (process_file): Move functions processing to
4706 process_all_functions.
4708 2018-08-02 David Malcolm <dmalcolm@redhat.com>
4710 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
4711 "const" to the "gimple *" and "rtx_insn *" parameters.
4712 * dumpfile.h (dump_user_location_t::dump_user_location_t):
4714 (dump_location_t::dump_location_t): Likewise.
4716 2018-08-01 Martin Sebor <msebor@redhat.com>
4718 PR tree-optimization/86650
4719 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
4720 rather than a "gcall *". Directly pass the data of interest
4721 to percent_K_format, rather than building a temporary CALL_EXPR
4723 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
4724 (gimple_fold_builtin_strncat): Adjust.
4725 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
4726 gcall* argument with gimple*.
4727 * gimple-ssa-warn-restrict.c (check_call): Same.
4728 (wrestrict_dom_walker::before_dom_children): Same.
4729 (builtin_access::builtin_access): Same.
4730 (check_bounds_or_overlap): Same
4731 (maybe_diag_overlap): Same.
4732 (maybe_diag_offset_bounds): Same.
4733 * tree-diagnostic.c (default_tree_printer): Move usage of
4734 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
4736 * tree-pretty-print.c (percent_K_format): Add argument.
4737 * tree-pretty-print.h: Add argument.
4738 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
4739 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
4740 (maybe_diag_stxncpy_trunc): Same.
4741 (handle_builtin_stxncpy): Same.
4742 (handle_builtin_strcat): Same.
4744 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4746 * match.pd: Optimise pointer range checks.
4748 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4750 PR tree-optimization/86758
4751 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
4752 to remove pattern statements.
4754 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4756 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
4757 result of dfs_enumerate_from when constructing stmt_vec_infos,
4758 instead of additionally calling get_loop_body.
4760 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4762 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
4764 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4765 When creating an iv, assert that the step is not known to be zero.
4766 (vect_setup_realignment): Update call accordingly.
4767 * tree-vect-stmts.c (vectorizable_store): Likewise.
4768 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
4770 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4772 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4773 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4774 (vectorizable_reduction): Likewise.
4775 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4776 (vect_detect_hybrid_slp_stmts): Likewise.
4777 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4779 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
4781 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4782 (wide_int_set_zero_nonzero_bits): ...this.
4783 (zero_nonzero_bits_from_vr): Rename to...
4784 (vrp_set_zero_nonzero_bits): ...this.
4785 (extract_range_from_multiplicative_op_1): Abstract wide int
4787 (wide_int_range_multiplicative_op): ...here.
4788 (extract_range_from_binary_expr_1): Extract wide int binary
4789 operations into their own functions.
4790 (wide_int_range_lshift): New.
4791 (wide_int_range_can_optimize_bit_op): New.
4792 (wide_int_range_shift_undefined_p): New.
4793 (wide_int_range_bit_xor): New.
4794 (wide_int_range_bit_ior): New.
4795 (wide_int_range_bit_and): New.
4796 (wide_int_range_trunc_mod): New.
4797 (extract_range_into_wide_ints): New.
4798 (vrp_shift_undefined_p): New.
4799 (extract_range_from_multiplicative_op): New.
4800 (vrp_can_optimize_bit_op): New.
4801 * tree-vrp.h (value_range::dump): New.
4802 (wide_int_range_multiplicative_op): New.
4803 (wide_int_range_lshift):New.
4804 (wide_int_range_shift_undefined_p): New.
4805 (wide_int_range_bit_xor): New.
4806 (wide_int_range_bit_ior): New.
4807 (wide_int_range_bit_and): New.
4808 (wide_int_range_trunc_mod): New.
4809 (zero_nonzero_bits_from_bounds): Rename to...
4810 (wide_int_set_zero_nonzero_bits): ...this.
4811 (zero_nonzero_bits_from_vr): Rename to...
4812 (vrp_set_zero_nonzero_bits): ...this.
4813 (range_easy_mask_min_max): Rename to...
4814 (wide_int_range_can_optimize_bit_op): this.
4815 * vr-values.c (simplify_bit_ops_using_ranges): Rename
4816 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
4818 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4820 * tree-vectorizer.h (vect_orig_stmt): New function.
4821 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
4822 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
4823 (vect_create_epilog_for_reduction): Likewise.
4824 (vectorizable_live_operation): Likewise.
4825 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4826 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
4827 * tree-vect-stmts.c (vectorizable_call): Likewise.
4828 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
4830 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4832 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
4834 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
4835 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
4836 (vect_transform_loop): Likewise.
4837 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4839 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4841 * tree-vectorizer.h (vect_schedule_slp): Return void.
4842 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
4843 (vect_schedule_slp): Likewise.
4845 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4847 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
4849 (vect_transform_loop): Update calls accordingly. Schedule SLP
4850 instances before the main loop, if any exist.
4852 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
4854 PR tree-optimization/86749
4855 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
4856 If the lhs is used in a COND_EXPR, check that it is being used
4857 as the "then" or "else" value.
4859 2018-08-01 Tom de Vries <tdevries@suse.de>
4862 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4863 speculation_safe_value_not_needed.
4865 2018-08-01 Richard Biener <rguenther@suse.de>
4867 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
4870 2018-08-01 Martin Liska <mliska@suse.cz>
4872 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
4873 format how successful transformation is dumped.
4874 (gimple_mod_pow2_value_transform): Likewise.
4875 (gimple_mod_subtract_transform): Likewise.
4876 (gimple_stringops_transform): Likewise.
4878 2018-08-01 Martin Liska <mliska@suse.cz>
4881 * value-prof.c (interesting_stringop_to_profile_p):
4882 Simplify the code and add BUILT_IN_MEMMOVE.
4883 (gimple_stringops_transform): Likewise.
4885 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
4887 * config/aarch64/aarch64-simd.md
4888 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
4889 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
4890 use GPI iterator instead of SI mode.
4892 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4894 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
4895 rs6000_speculation_barrier.
4896 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
4897 new barrier pattern name.
4899 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
4901 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
4902 (speculation_barrier): New insn.
4904 2018-08-01 Richard Biener <rguenther@suse.de>
4907 * graphite.h: Include isl/id.h and isl/space.h to allow build
4910 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
4913 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4914 mode for COFF targets.
4915 * defaults.h (TARGET_COFF): Define.
4916 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
4917 TARGET_COFF): Define.
4918 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
4919 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
4920 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
4922 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4924 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
4925 * gimple-streamer-out.c (output_bb): Save it.
4926 * lto-streamer-in.c (input_struct_function_base): Restore
4927 instance discriminator if available. Create map on demand.
4928 * lto-streamer-out.c (output_struct_function_base): Save it if
4930 * final.c (decl_to_instance_map): Document LTO strategy.
4932 2018-07-31 Alexandre Oliva <oliva@adacore.com>
4933 Olivier Hainque <hainque@adacore.com>
4935 * debug.h (decl_to_instance_map_t): New type.
4936 (decl_to_instance_map): Declare.
4937 (maybe_create_decl_to_instance_map): New inline function.
4938 * final.c (bb_discriminator, last_bb_discriminator): New statics,
4939 to track basic block discriminators.
4940 (final_start_function_1): Initialize them.
4941 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
4943 (decl_to_instance_map): New variable.
4944 (map_decl_to_instance, maybe_set_discriminator): New functions.
4945 (notice_source_line): Set discriminator.
4947 2018-07-31 Ian Lance Taylor <iant@golang.org>
4949 * targhooks.c (default_have_speculation_safe_value): Add
4952 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4954 * dump-context.h: Include "pretty-print.h".
4955 (dump_context::refresh_dumps_are_enabled): New decl.
4956 (dump_context::emit_item): New decl.
4957 (class dump_context): Add fields "m_test_pp" and
4959 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
4960 (temp_dump_context::get_dumped_text): New decl.
4961 (class temp_dump_context): Add field "m_pp".
4962 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
4963 (dump_context::refresh_dumps_are_enabled): ...and add a test for
4965 (set_dump_file): Update for above change.
4966 (set_alt_dump_file): Likewise.
4967 (dump_loc): New overload, taking a pretty_printer *.
4968 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
4969 to any test pretty-printer.
4970 (make_item_for_dump_gimple_stmt): New function, adapted from
4971 optinfo::add_gimple_stmt.
4972 (dump_context::dump_gimple_stmt): Call it, and use the result,
4973 eliminating the direct usage of dump_file and alt_dump_file in
4974 favor of indirectly using them via emit_item.
4975 (make_item_for_dump_gimple_expr): New function, adapted from
4976 optinfo::add_gimple_expr.
4977 (dump_context::dump_gimple_expr): Call it, and use the result,
4978 eliminating the direct usage of dump_file and alt_dump_file in
4979 favor of indirectly using them via emit_item.
4980 (make_item_for_dump_generic_expr): New function, adapted from
4982 (dump_context::dump_generic_expr): Call it, and use the result,
4983 eliminating the direct usage of dump_file and alt_dump_file in
4984 favor of indirectly using them via emit_item.
4985 (make_item_for_dump_printf_va): New function, adapted from
4986 optinfo::add_printf_va.
4987 (make_item_for_dump_printf): New function.
4988 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
4989 and use the result, eliminating the direct usage of dump_file and
4990 alt_dump_file in favor of indirectly using them via emit_item.
4991 (make_item_for_dump_dec): New function.
4992 (dump_context::dump_dec): Call it, and use the result,
4993 eliminating the direct usage of dump_file and alt_dump_file in
4994 favor of indirectly using them via emit_item.
4995 (make_item_for_dump_symtab_node): New function, adapted from
4996 optinfo::add_symtab_node.
4997 (dump_context::dump_symtab_node): Call it, and use the result,
4998 eliminating the direct usage of dump_file and alt_dump_file in
4999 favor of indirectly using them via emit_item.
5000 (dump_context::begin_scope): Reimplement, avoiding direct usage
5001 of dump_file and alt_dump_file in favor of indirectly using them
5003 (dump_context::emit_item): New member function.
5004 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5005 Set up test pretty-printer on the underlying context. Call
5006 refresh_dumps_are_enabled.
5007 (temp_dump_context::~temp_dump_context): Call
5008 refresh_dumps_are_enabled.
5009 (temp_dump_context::get_dumped_text): New member function.
5010 (selftest::verify_dumped_text): New function.
5011 (ASSERT_DUMPED_TEXT_EQ): New macro.
5012 (selftest::test_capture_of_dump_calls): Run all tests twice, with
5013 and then without optinfo enabled. Add uses of
5014 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
5015 * dumpfile.h: Update comment for the dump_* API.
5016 * optinfo-emit-json.cc
5017 (selftest::test_building_json_from_dump_calls): Update for new
5018 param for temp_dump_context ctor.
5019 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
5020 and "m_owned" field.
5021 (optinfo_item::~optinfo_item): Likewise.
5022 (optinfo::add_item): New member function.
5023 (optinfo::emit): Update comment.
5024 (optinfo::add_string): Delete.
5025 (optinfo::add_printf): Delete.
5026 (optinfo::add_printf_va): Delete.
5027 (optinfo::add_gimple_stmt): Delete.
5028 (optinfo::add_gimple_expr): Delete.
5029 (optinfo::add_tree): Delete.
5030 (optinfo::add_symtab_node): Delete.
5031 (optinfo::add_dec): Delete.
5032 * optinfo.h (class dump_context): New forward decl.
5033 (optinfo::add_item): New decl.
5034 (optinfo::add_string): Delete.
5035 (optinfo::add_printf): Delete.
5036 (optinfo::add_printf_va): Delete.
5037 (optinfo::add_gimple_stmt): Delete.
5038 (optinfo::add_gimple_expr): Delete.
5039 (optinfo::add_tree): Delete.
5040 (optinfo::add_symtab_node): Delete.
5041 (optinfo::add_dec): Delete.
5042 (optinfo::add_poly_int): Delete.
5043 (optinfo_item::optinfo_item): Remove "owned" param.
5044 (class optinfo_item): Remove field "m_owned".
5046 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5049 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
5050 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
5051 requested variable alignment.
5052 (expand_one_ssa_partition): Likewise.
5053 (expand_one_var): Likewise.
5055 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5057 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5058 to speculation_safe_value_not_needed.
5060 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5062 * targhooks.h (speculation_safe_value_not_needed): New prototype.
5063 * targhooks.c (speculation_safe_value_not_needed): New function.
5064 * target.def (have_speculation_safe_value): Update documentation.
5065 * doc/tm.texi: Regenerated.
5067 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5069 * config/aarch64/iterators.md (ALLI_TI): New iterator.
5070 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
5072 (despeculate_copy<ALLI:mode>_insn): New insn.
5073 (despeculate_copyti_insn): New insn.
5074 (despeculate_simple<ALLI:mode>): New insn
5075 (despeculate_simpleti): New insn.
5076 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
5078 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
5079 aarch64_speculation_safe_value.
5080 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
5082 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5084 * config/aarch64/aarch64-speculation.cc: New file.
5085 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
5086 before pass_reorder_blocks.
5087 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
5089 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
5090 X14 and X15 when tracking speculation.
5091 * config/aarch64/aarch64.md (register name constants): Add
5092 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
5093 (unspec): Add UNSPEC_SPECULATION_TRACKER.
5094 (speculation_barrier): New insn attribute.
5095 (cmp<mode>): Allow SP in comparisons.
5096 (speculation_tracker): New insn.
5097 (speculation_barrier): Add speculation_barrier attribute.
5098 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
5099 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
5100 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
5102 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5104 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
5105 aarch64_track_speculation is true.
5106 (tb<optab><mode>1): Likewise.
5107 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
5108 generate CB[N]Z when tracking speculation.
5109 (aarch64_split_compare_and_swap): Likewise.
5110 (aarch64_split_atomic_op): Likewise.
5112 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5114 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
5116 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5118 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
5119 (speculation_barrier): New insn.
5121 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5123 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
5124 * config/arm/arm.md (speculation_barrier): New expand.
5125 (speculation_barrier_insn): New pattern.
5127 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5129 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
5130 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
5131 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
5132 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
5134 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
5135 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
5136 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
5137 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
5138 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
5139 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
5140 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
5141 * builtins.c (expand_speculation_safe_value): New function.
5142 (expand_builtin): Call it.
5143 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
5144 * doc/extend.texi: Document __builtin_speculation_safe_value.
5145 * doc/md.texi: Document "speculation_barrier" pattern.
5146 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
5147 TARGET_HAVE_SPECULATION_SAFE_VALUE.
5148 * doc/tm.texi: Regenerated.
5149 * target.def (have_speculation_safe_value, speculation_safe_value): New
5151 * targhooks.c (default_have_speculation_safe_value): New function.
5152 (default_speculation_safe_value): New function.
5153 * targhooks.h (default_have_speculation_safe_value): Add prototype.
5154 (default_speculation_safe_value): Add prototype.
5156 2018-07-31 David Malcolm <dmalcolm@redhat.com>
5158 * dump-context.h (dump_context::dump_loc): New decl.
5159 * dumpfile.c (dump_context::dump_loc): New member function.
5160 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
5161 and dump_gimple_stmt.
5162 (dump_context::dump_gimple_expr_loc): Likewise, using
5164 (dump_context::dump_generic_expr_loc): Likewise, using
5166 (dump_context::dump_printf_loc_va): Likewise, using
5168 (dump_context::begin_scope): Explicitly using the global function
5169 "dump_loc", rather than the member function.
5171 2018-07-31 Martin Sebor <msebor@redhat.com>
5173 PR tree-optimization/86741
5174 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
5176 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
5178 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
5179 depend on whether prefetch instructions will be emitted or not.
5180 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
5181 will be emitted or not.
5182 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
5183 (TARGET_SETMEM_PFD): New macros.
5185 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5187 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
5188 (NULL_STMT_VEC_INFO): Delete.
5189 (stmt_vec_info::operator*): Likewise.
5190 (stmt_vec_info::operator gimple *): Likewise.
5191 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
5192 of NULL_STMT_VEC_INFO.
5193 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5194 (vect_reassociating_reduction_p): Likewise.
5195 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5196 (vectorizable_store): Likewise.
5197 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
5198 (vec_info::free_stmt_vec_infos): Likewise.
5200 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5202 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
5203 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
5205 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5207 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
5208 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5209 (vec_info::free_stmt_vec_info): New private member functions.
5210 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
5211 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
5212 * tree-parloops.c (gather_scalar_reductions): Remove calls to
5213 set_stmt_vec_info_vec and free_stmt_vec_infos.
5214 * tree-vect-loop.c (_loop_vec_info): Remove call to
5215 set_stmt_vec_info_vec.
5216 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
5217 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
5218 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
5219 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5220 (vec_info::free_stmt_vec_info): ...these new functions. Remove
5221 assignments in {vec_info::,}new_stmt_vec_info that are redundant
5222 with the clearing in the xcalloc.
5223 (stmt_vec_info_vec): Delete.
5224 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
5225 (vectorize_loops): Likewise.
5226 (vec_info::~vec_info): Remove argument from call to
5227 free_stmt_vec_infos.
5228 (vec_info::add_stmt): Remove vinfo argument from call to
5231 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5233 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
5234 rather than a gimple stmt.
5235 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
5236 information for pattern statements when passed the original
5237 statement; instead wait to be passed the pattern statement itself.
5238 Don't call set_vinfo_for_stmt here.
5239 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
5240 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
5241 stmt_vec_infos here.
5242 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
5243 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
5244 stmt_vec_infos entry.
5246 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5248 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
5249 * tree-vectorizer.c (vec_info::replace_stmt): New function.
5250 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
5251 * tree-vect-stmts.c (vectorizable_call): Likewise.
5252 (vectorizable_simd_clone_call): Likewise.
5254 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5256 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
5257 * tree-vectorizer.c (vec_info::remove_stmt): New function.
5258 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
5259 * tree-vect-loop.c (vect_transform_loop): Likewise.
5260 * tree-vect-slp.c (vect_schedule_slp): Likewise.
5261 * tree-vect-stmts.c (vect_remove_stores): Likewise.
5263 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5265 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
5266 (vect_dr_stmt): Delete.
5267 * tree-vectorizer.c (vec_info::lookup_dr): New function.
5268 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
5270 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
5271 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
5272 (vect_verify_datarefs_alignment, vect_peeling_supportable)
5273 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5274 (vect_analyze_data_refs): Likewise.
5275 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
5277 (vect_find_same_alignment_drs): Likewise.
5278 (vect_slp_analyze_node_dependences): Update calls accordingly.
5279 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
5280 instead of DR_VECT_AUX.
5281 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
5282 of a vector data references. Use vec_info::lookup_dr instead of
5284 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
5285 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
5286 instead of DR_VECT_AUX.
5288 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5290 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
5292 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5294 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5295 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5296 (vect_gen_prolog_loop_niters): Likewise.
5298 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5300 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
5301 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
5302 (vect_known_alignment_in_bytes, vect_dr_behavior)
5303 (vect_get_scalar_dr_size): Take references as dr_vec_infos
5304 instead of data_references. Update calls to other routines for
5305 which the same change has been made.
5306 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
5307 dr_vec_infos instead of stmt_vec_infos.
5308 (vect_analyze_data_ref_dependence): Update call accordingly.
5309 (vect_slp_analyze_data_ref_dependence)
5310 (vect_record_base_alignments): Use DR_VECT_AUX.
5311 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
5312 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
5313 (vector_alignment_reachable_p, vect_get_data_access_cost)
5314 (vect_peeling_supportable, vect_analyze_group_access_1)
5315 (vect_analyze_group_access, vect_analyze_data_ref_access)
5316 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
5317 (vect_compile_time_alias, vect_small_gap_p)
5318 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
5319 (vect_supportable_dr_alignment): Take references as dr_vec_infos
5320 instead of data_references. Update calls to other routines for
5321 which the same change has been made.
5322 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
5323 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5324 (vect_slp_analyze_and_verify_node_alignment)
5325 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5326 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5327 (vect_setup_realignment): Use dr_vec_infos. Update calls after
5329 (_vect_peel_info::dr): Replace with...
5330 (_vect_peel_info::dr_info): ...this new field.
5331 (vect_peeling_hash_get_most_frequent)
5332 (vect_peeling_hash_choose_best_peeling): Update accordingly.
5333 (vect_peeling_hash_get_lowest_cost):
5334 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
5335 routines for which the same change has been made.
5336 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
5338 * tree-vect-loop-manip.c (get_misalign_in_elems)
5339 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
5341 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5342 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
5343 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
5344 (get_group_load_store_type, get_negative_load_store_type)
5345 (vect_get_data_ptr_increment, vectorizable_store)
5346 (vectorizable_load): Likewise.
5347 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
5348 Update calls to other routines for which the same change has been made.
5350 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5352 * tree-vectorizer.h (vec_info::move_dr): New member function.
5353 (dataref_aux): Rename to...
5354 (dr_vec_info): ...this and add "dr" and "stmt" fields.
5355 (_stmt_vec_info::dr_aux): Update accordingly.
5356 (_stmt_vec_info::data_ref_info): Delete.
5357 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
5358 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
5359 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
5360 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
5362 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
5363 (STMT_VINFO_DR_INFO): New macro.
5364 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
5365 (set_dr_misalignment): Update after rename of dataref_aux.
5366 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
5367 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
5368 initialization of STMT_VINFO_DATA_REF.
5369 * tree-vectorizer.c (vec_info::move_dr): New function.
5370 * tree-vect-patterns.c (vect_recog_bool_pattern)
5371 (vect_recog_mask_conversion_pattern)
5372 (vect_recog_gather_scatter_pattern): Use it.
5373 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
5374 the "dr" and "stmt" fields of dr_vec_info instead of
5375 STMT_VINFO_DATA_REF.
5377 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5379 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
5380 (is_pattern_stmt_p): Use it.
5381 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
5382 on pattern statements.
5384 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5386 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
5387 original stmt as a stmt_vec_info rather than a gimple stmt.
5388 (vect_pattern_recog_1): Take the statement directly as a
5389 stmt_vec_info, rather than via a gimple_stmt_iterator.
5390 Update call to vect_mark_pattern_stmts.
5391 (vect_pattern_recog): Update calls accordingly.
5393 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5395 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
5396 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
5397 a vect_def_type for the first argument.
5398 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
5399 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
5400 operand if it isn't defined by a vectorized statement.
5401 (vect_build_gather_load_calls): Remove the mask_dt argument and
5402 update calls to vect_get_vec_def_for_stmt_copy.
5403 (vectorizable_bswap): Likewise the dt argument.
5404 (vectorizable_call): Update calls to vectorizable_bswap and
5405 vect_get_vec_def_for_stmt_copy.
5406 (vectorizable_simd_clone_call, vectorizable_assignment)
5407 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
5408 (vectorizable_comparison): Update calls to
5409 vect_get_vec_def_for_stmt_copy.
5410 (vectorizable_store): Likewise. Remove now-unnecessary calls to
5412 (vect_get_loop_based_defs): Remove dt argument and update call
5413 to vect_get_vec_def_for_stmt_copy.
5414 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
5415 and vect_get_vec_def_for_stmt_copy.
5416 (vectorizable_load): Update calls to vect_build_gather_load_calls
5417 and vect_get_vec_def_for_stmt_copy.
5418 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5419 (vectorizable_reduction, vectorizable_live_operation): Update calls
5420 to vect_get_vec_def_for_stmt_copy.
5422 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5424 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
5425 and gimple stmt arguments with a stmt_vec_info.
5426 (vect_record_base_alignments): Update calls accordingly.
5427 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
5428 and gimple stmt arguments with a stmt_vec_info.
5429 (vect_build_slp_tree_1): Remove vinfo argument and update call
5430 to vect_record_max_nunits.
5431 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
5432 and vect_record_max_nunits.
5434 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5436 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
5437 file and take a stmt_vec_info instead of a gimple stmt.
5438 (supportable_widening_operation, vect_finish_replace_stmt)
5439 (vect_finish_stmt_generation, vect_get_store_rhs)
5440 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
5441 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
5442 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
5443 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
5444 (vect_create_data_ref_ptr, bump_vector_ptr)
5445 (vect_permute_store_chain, vect_setup_realignment)
5446 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
5447 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
5448 (vectorizable_reduction, vectorizable_induction)
5449 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
5450 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
5451 than gimple stmts as arguments.
5452 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
5453 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
5454 (can_group_stmts_p, vect_check_gather_scatter)
5455 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5456 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
5457 (vect_permute_load_chain, vect_shift_permute_load_chain)
5458 (vect_transform_grouped_load)
5459 (vect_record_grouped_load_vectors): Likewise.
5460 * tree-vect-loop.c (vect_fixup_reduc_chain)
5461 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
5462 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
5463 (vectorizable_reduction, vectorizable_induction)
5464 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
5465 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
5466 (vect_get_load_store_mask): Likewise.
5467 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
5468 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
5469 * tree-vect-stmts.c (vect_mark_relevant)
5470 (is_simple_and_all_uses_invariant)
5471 (exist_non_indexing_operands_for_use_p, process_use)
5472 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
5473 (vect_get_vec_def_for_operand, vect_get_vec_defs)
5474 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
5475 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
5476 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
5477 (get_negative_load_store_type, get_load_store_type)
5478 (vect_check_load_store_mask, vect_check_store_rhs)
5479 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
5480 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5481 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5482 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5483 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
5484 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
5485 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
5486 (supportable_widening_operation): Likewise.
5488 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5490 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
5491 a stmt_vec_info instead of a gcall.
5492 (vect_check_gather_scatter): Update call accordingly.
5493 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
5495 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
5496 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
5497 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
5498 instead of a gimple stmt.
5499 (vect_transform_loop): Update calls accordingly.
5500 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
5501 stmt_vec_infos instead of gimple stmts.
5502 (vect_analyze_slp_instance): Update use accordingly.
5503 * tree-vect-stmts.c (read_vector_array, write_vector_array)
5504 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
5505 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
5506 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
5507 (vect_gen_widened_results_half, vect_get_loop_based_defs)
5508 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
5509 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
5510 down to subroutines.
5512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5514 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
5515 of the worklist from a vector of gimple stmts to a vector of
5517 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5518 (vect_mark_stmts_to_be_vectorized): Likewise
5520 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5522 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
5523 statement before passing it to vect_analyze_stmt.
5524 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
5525 the chain of phi vector definitions. Track the exit phi via its
5527 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
5528 STMT_VINFO_REDUC_DEF.
5529 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
5530 stmt_vec_infos to handle the statement chains.
5531 (vect_get_slp_defs): Record the first statement in the node
5532 using a stmt_vec_info.
5533 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
5534 statements here and pass their stmt_vec_info down to subroutines.
5535 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
5536 down to vect_finish_stmt_generation.
5537 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
5538 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
5539 stmt_vec_infos to subroutines.
5540 (vect_remove_stores): Use stmt_vec_infos to handle the statement
5543 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5545 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
5546 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
5547 (vect_permute_store_chain, vect_setup_realignment)
5548 (vect_permute_load_chain, vect_shift_permute_load_chain)
5549 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
5550 stmts internally, and when passing values to other vectorizer routines.
5551 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5552 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
5553 (vect_analyze_loop_operations, get_initial_def_for_reduction)
5554 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
5555 (vectorizable_reduction, vectorizable_induction)
5556 (vectorizable_live_operation, vect_transform_loop_stmt)
5557 (vect_transform_loop): Likewise.
5558 * tree-vect-patterns.c (vect_reassociating_reduction_p)
5559 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
5560 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
5561 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5562 (vect_slp_analyze_node_operations_1): Likewise.
5563 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5564 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
5565 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
5566 (vect_finish_stmt_generation_1, get_group_load_store_type)
5567 (get_load_store_type, vect_build_gather_load_calls)
5568 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5569 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5570 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5571 (vectorizable_store, vectorizable_load, vectorizable_condition)
5572 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
5573 (supportable_widening_operation): Likewise.
5574 (vect_get_vector_types_for_stmt): Likewise.
5575 * tree-vectorizer.h (vect_dr_behavior): Likewise.
5577 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5579 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5580 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
5581 (vect_permute_store_chain, vect_permute_load_chain)
5582 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
5583 repeated stmt_vec_info lookups.
5584 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5585 (vect_update_ivs_after_vectorizer): Likewise.
5586 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
5587 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5588 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
5589 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5590 (vect_bb_slp_scalar_cost): Likewise.
5591 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
5593 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5595 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
5596 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
5597 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
5598 to get gassigns and gcalls, rather than operating on generc gimple
5600 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
5601 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
5602 (vectorizable_load, vect_analyze_stmt): Likewise.
5603 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
5605 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5607 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
5608 return stmt_vec_infos rather than gimple stmts. Do not accept
5610 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
5612 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5613 Update use of get_later_stmt.
5614 (vect_get_constant_vectors): Update call accordingly.
5615 (vect_schedule_slp_instance): Likewise
5616 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
5617 (vect_slp_analyze_instance_dependence): Likewise.
5618 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
5620 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5622 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
5623 (stmt_info_for_cost::stmt_info): ...this new field.
5624 (add_stmt_costs): Update accordingly.
5625 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5626 (vect_get_known_peeling_cost): Likewise.
5627 (vect_estimate_min_profitable_iters): Likewise.
5628 * tree-vect-stmts.c (record_stmt_cost): Likewise.
5630 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5632 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
5633 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5634 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5636 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
5638 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5640 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
5641 a gimple stmt to a stmt_vec_info.
5642 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5644 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5646 * tree-vectorizer.h (vec_info::grouped_stores): Change from
5647 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5648 (_loop_vec_info::reduction_chains): Likewise.
5649 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
5651 * tree-vect-slp.c (vect_analyze_slp): Likewise.
5653 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5655 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
5656 a gimple stmt to a stmt_vec_info.
5657 (_stmt_vec_info::next_element): Likewise.
5658 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
5659 (vect_slp_analyze_and_verify_node_alignment)
5660 (vect_analyze_group_access_1, vect_analyze_group_access)
5661 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
5662 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
5663 (vect_supportable_dr_alignment): Update accordingly.
5664 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
5665 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
5666 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
5667 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5668 * tree-vect-slp.c (vect_build_slp_tree_1)
5669 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
5670 (vect_split_slp_store_group, vect_analyze_slp_instance)
5671 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
5672 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
5673 (get_group_load_store_type, get_load_store_type)
5674 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
5675 (vect_transform_stmt, vect_remove_stores): Likewise.
5677 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5679 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
5681 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5682 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
5683 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
5684 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
5685 (vector_alignment_reachable_p, vect_get_data_access_cost)
5686 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
5687 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
5688 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5689 (vect_analyze_group_access_1, vect_analyze_group_access)
5690 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
5691 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
5692 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
5693 result of vect_dr_stmt and use the stmt_vec_info instead of
5694 the associated gimple stmt.
5695 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5696 (vect_gen_prolog_loop_niters): Likewise.
5697 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5699 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5701 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
5702 to a vec<stmt_vec_info>.
5703 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
5704 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
5706 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
5707 to a vec<stmt_vec_info>.
5708 (bst_traits::value_type, bst_traits::value_type): Likewise.
5709 (bst_traits::hash): Update accordingly.
5710 (vect_get_and_check_slp_defs): Change the stmts parameter from
5711 a vec<gimple *> to a vec<stmt_vec_info>.
5712 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
5713 (vect_build_slp_tree): Likewise.
5714 (vect_build_slp_tree_2): Likewise. Update uses of
5715 SLP_TREE_SCALAR_STMTS.
5716 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
5717 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
5718 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
5719 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
5720 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
5721 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
5722 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
5723 (vect_get_constant_vectors, vect_get_slp_defs)
5724 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
5725 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
5726 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
5727 instead of gimple stmts.
5728 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
5729 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
5730 (vect_slp_analyze_instance_dependence): Update uses of
5731 SLP_TREE_SCALAR_STMTS.
5732 (vect_slp_analyze_and_verify_node_alignment): Likewise.
5733 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
5734 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
5735 (get_initial_defs_for_reduction): Likewise.
5736 (vect_create_epilog_for_reduction): Likewise.
5737 (vectorize_fold_left_reduction): Likewise.
5738 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
5739 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
5740 (can_vectorize_live_stmts): Likewise.
5742 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5744 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
5745 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5746 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
5748 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
5750 (gather_scalar_reductions): Update after above interface changes.
5751 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
5752 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
5754 (vect_force_simple_reduction): Likewise.
5755 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
5756 LOOP_VINFO_REDUCTIONS.
5757 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5759 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5761 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
5762 a gimple stmt to a stmt_vec_info.
5763 * tree-vect-loop.c (vect_active_double_reduction_p)
5764 (vect_force_simple_reduction, vectorizable_reduction): Update
5767 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5769 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5770 vec<gimple *> to a vec<stmt_vec_info>.
5771 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5772 the reduction_phis argument from a vec<gimple *> to a
5774 (vectorizable_reduction): Likewise the phis local variable that
5775 is passed to vect_create_epilog_for_reduction. Update for new type
5776 of SLP_TREE_VEC_STMTS.
5777 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5778 (vectorizable_live_operation): Likewise.
5779 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5780 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5782 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5784 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5785 a gimple stmt to a stmt_vec_info.
5786 (vectorizable_condition, vectorizable_live_operation)
5787 (vectorizable_reduction, vectorizable_induction): Pass back the
5788 vectorized statement as a stmt_vec_info.
5789 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5790 use of STMT_VINFO_VEC_STMT.
5791 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5792 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5793 as stmt_vec_infos rather than gimple stmts.
5794 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5796 (vectorizable_live_operation): Likewise.
5797 (vectorizable_reduction, vectorizable_induction): Likewise,
5798 updating use of STMT_VINFO_VEC_STMT.
5799 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
5800 of STMT_VINFO_VEC_STMT.
5801 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
5802 (vectorizable_simd_clone_call, vectorizable_conversion)
5803 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5804 (vectorizable_store, vectorizable_load, vectorizable_condition)
5805 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
5806 from a gimple stmt to a stmt_vec_info.
5807 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
5808 pointer to a stmt_vec_info to the vectorizable_* routines.
5810 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5812 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
5813 a gimple stmt to a stmt_vec_info.
5814 (is_pattern_stmt_p): Update accordingly.
5815 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
5816 (vect_record_grouped_load_vectors): Likewise.
5817 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5818 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
5819 (vect_model_reduction_cost): Likewise.
5820 (vect_create_epilog_for_reduction): Likewise.
5821 (vectorizable_reduction, vectorizable_induction): Likewise.
5822 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5823 Return the stmt_vec_info for the pattern statement.
5824 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
5825 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
5826 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5827 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
5828 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5829 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
5830 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
5831 (free_stmt_vec_info, vect_is_simple_use): Likewise.
5833 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5835 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
5836 (vect_finish_stmt_generation): Likewise.
5837 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5838 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
5839 (vect_build_gather_load_calls): Use the return value of the above
5840 functions instead of a separate call to vinfo_for_stmt. Use narrow
5841 scopes for the input gimple stmt and wider scopes for the associated
5842 stmt_vec_info. Use vec_info::lookup_def when setting these
5843 stmt_vec_infos from an SSA_NAME definition.
5844 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5845 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5846 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5847 (vectorizable_store, vectorizable_load, vectorizable_condition)
5848 (vectorizable_comparison): Likewise.
5849 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
5850 (vectorizable_reduction): Likewise.
5852 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5854 * tree-vectorizer.h (vect_is_simple_use): Add an optional
5855 stmt_vec_info * parameter before the optional gimple **.
5856 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5857 (process_use, vect_get_vec_def_for_operand_1): Update callers.
5858 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
5859 * tree-vect-loop.c (vectorizable_reduction): Likewise.
5860 (vectorizable_live_operation): Likewise.
5861 * tree-vect-patterns.c (type_conversion_p): Likewise.
5862 (vect_look_through_possible_promotion): Likewise.
5863 (vect_recog_rotate_pattern): Likewise.
5864 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5866 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5868 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
5869 a typedef to a wrapper class.
5870 (NULL_STMT_VEC_INFO): New macro.
5871 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
5872 (stmt_vec_info::operator*): New function.
5873 (stmt_vec_info::operator gimple *): Likewise.
5874 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
5875 (add_stmt_costs): Likewise.
5876 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
5877 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5878 (vect_get_known_peeling_cost): Likewise.
5879 (vect_estimate_min_profitable_iters): Likewise.
5880 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5881 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
5882 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5883 (vectorizable_store, free_stmt_vec_infos): Likewise.
5884 (new_stmt_vec_info): Change return type of xcalloc to
5887 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5889 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
5890 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
5891 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
5892 a single_imm_use-based sequence.
5893 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
5895 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5897 * tree-vectorizer.h (vec_info::lookup_def): Declare.
5898 * tree-vectorizer.c (vec_info::lookup_def): New function.
5899 * tree-vect-patterns.c (vect_get_internal_def): Use it.
5900 (vect_widened_op_tree): Likewise.
5901 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5902 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
5903 (vectorizable_reduction): Likewise.
5904 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
5906 (vect_is_slp_reduction): Update calls accordingly. Use
5907 vec_info::lookup_def.
5908 (vect_is_simple_reduction): Likewise
5909 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
5911 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5913 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
5914 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
5915 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
5917 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
5918 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
5919 (vect_update_vf_for_slp, vect_analyze_loop_operations)
5920 (vect_is_slp_reduction, vectorizable_induction)
5921 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
5922 * tree-vect-patterns.c (vect_init_pattern_stmt):
5923 (vect_determine_min_output_precision_1, vect_determine_precisions)
5924 (vect_pattern_recog): Likewise.
5925 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
5926 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
5927 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
5928 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
5929 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
5930 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
5931 info field from a loop to a loop_vec_info.
5933 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5935 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
5936 (vec_info::add_stmt): Declare.
5937 * tree-vectorizer.c (vec_info::add_stmt): New function.
5938 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
5939 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
5940 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5941 (vectorizable_induction): Likewise.
5942 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
5943 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5944 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
5945 (vectorizable_load): Likewise.
5946 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5947 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
5948 (vect_recog_gather_scatter_pattern): Likewise.
5949 (append_pattern_def_seq): Likewise. Remove a check that is
5950 performed by add_stmt itself.
5952 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5954 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
5955 which make_ssa_name was called with new_stmt before new_stmt
5958 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5960 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
5962 (vect_is_slp_reduction): ...here...
5963 (vect_is_simple_reduction): ...and here. Remove repetition of tests
5964 that are already known to be false.
5966 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5968 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
5969 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
5970 STMT_VINFO_NUM_SLP_USES when it's true.
5971 (vect_free_slp_instance): Add a final_p parameter and pass it to
5973 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
5974 (vect_analyze_slp_instance): Likewise.
5975 (vect_slp_analyze_operations): Likewise.
5976 (vect_slp_analyze_bb_1): Likewise.
5977 * tree-vectorizer.c (vec_info): Likewise.
5978 * tree-vect-loop.c (vect_transform_loop): Likewise.
5980 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5982 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5983 function is not called for second and subsequent members of
5986 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5988 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
5989 cases for nested loops from here to ...
5990 (vect_create_epilog_for_reduction): ...here. Only call
5991 vect_is_simple_use for inner-loop reductions.
5993 2018-07-31 Martin Liska <mliska@suse.cz>
5995 PR gcov-profile/85338
5996 PR gcov-profile/85350
5997 PR gcov-profile/85372
5998 * profile.c (struct location_triplet): New.
5999 (struct location_triplet_hash): Likewise.
6000 (output_location): Do not output a BB that
6001 is already recorded for a line.
6002 (branch_prob): Use streamed_locations.
6004 2018-07-31 Martin Liska <mliska@suse.cz>
6006 PR gcov-profile/85370
6007 * coverage.c (coverage_begin_function): Do not mark target
6008 clones as artificial functions.
6010 2018-07-31 Martin Liska <mliska@suse.cz>
6012 PR gcov-profile/83813
6013 PR gcov-profile/84758
6014 PR gcov-profile/85217
6015 PR gcov-profile/85332
6016 * profile.c (branch_prob): Do not record GOTO expressions
6017 for GIMPLE statements which locations are already streamed.
6019 2018-07-31 Olivier Hainque <hainque@adacore.com>
6021 * gcc.c (handle_spec_function): Accept a soft_matched_part
6022 argument, as do_spec_1. Pass it down to ...
6023 (eval_spec_function): Accept a soft_matched_part argument,
6024 and pass it down to ...
6025 (do_spec_2): Accept a soft_matched_part argument, and pass
6026 it down to do_spec_1.
6027 (do_spec_1): Pass soft_matched_part to handle_spec_function.
6028 (handle_braces): Update call to handle_spec_function.
6029 (driver::set_up_specs): Update calls to do_spec_2.
6030 (compare_debug_dump_opt_spec_function): Likewise.
6031 (compare_debug_self_opt_spec_function): Likewise.
6033 2018-07-31 Olivier Hainque <hainque@adacore.com>
6035 * common.opt (nolibc): New option.
6036 * doc/invoke.texi (Link Options): Document it.
6037 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
6038 * config/alpha/linux.h: Likewise.
6039 * config/arc/elf.h: Likewise.
6040 * config/arm/uclinux-elf.h: Likewise.
6041 * config/arm/unknown-elf.h: Likewise.
6042 * config/avr/avrlibc.h: Likewise.
6043 * config/bfin/bfin.h: Likewise.
6044 * config/bfin/linux.h: Likewise.
6045 * config/bfin/uclinux.h: Likewise.
6046 * config/darwin.h: Likewise.
6047 * config/darwin10.h: Likewise.
6048 * config/darwin12.h: Likewise.
6049 * config/gnu-user.h: Likewise.
6050 * config/lm32/uclinux-elf.h: Likewise.
6051 * config/pa/pa-hpux11.h: Likewise.
6052 * config/pa/pa64-hpux.h: Likewise.
6053 * config/sparc/sparc.h: Likewise.
6055 2018-07-31 Olivier Hainque <hainque@adacore.com>
6057 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
6058 undefined variables.
6060 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
6063 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
6066 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6068 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6069 terminated string literal.
6071 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
6073 PR rtl-optimization/85160
6074 * combine.c (is_just_move): New function.
6075 (try_combine): Allow combining two instructions into two if neither of
6076 the original instructions was a move.
6078 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
6081 * doc/extend.texi (Global Register Variables): Discourage use of type
6083 (Local Register Variables): Likewise.
6085 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
6087 PR tree-optimization/86506
6088 * hwint.h (ceil_log2): Resync with hwint.c implementation.
6090 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
6093 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
6094 hard_regno, make sure no insn between `from` and `to` clobbers it.
6096 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
6097 Tom de Vries <tdevries@suse.de>
6099 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
6100 (PTX_DEFAULT_RUNTIME_DIM): ... this.
6101 (nvptx_goacc_validate_dims): Set default worker and gang dims to
6102 PTX_DEFAULT_RUNTIME_DIM.
6103 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
6105 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
6107 * config/pa/pa.c (pa_output_addr_vec): Align address table.
6108 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
6109 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
6111 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
6113 * config/rs6000/constraints.md (wG constraint): Delete, no longer
6115 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
6116 predicate to reflect toc fusion has been deleted.
6117 (toc_fusion_mem_raw): Delete, no longer used.
6118 (toc_fusion_mem_wrapped): Likewise.
6119 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
6121 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
6122 Delete, no longer used.
6123 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
6124 meant to be used for toc fusion.
6125 (rs6000_debug_print_mode): Delete toc fusion debugging.
6126 (rs6000_debug_reg_global): Likewise.
6127 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
6128 fusion and secondary reload support that were never used.
6129 (rs6000_option_override_internal): Delete TOC fusion, that was only
6130 partially defined, and it did not work unless you also used the
6132 (rs6000_legitimate_address_p): Delete TOC fusion support.
6133 (rs6000_opt_masks): Likewise.
6134 (fusion_wrap_memory_address): Delete function, no longer used.
6135 (fusion_split_address); Delete TOC fusion support.
6136 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
6137 longer used with toc fusion being deleted.
6138 (TARGET_TOC_FUSION_FP): Likewise.
6139 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
6141 (toc fusion spliter): Delete TOC fusion support.
6142 (toc_fusionload_<mode>): Likewise.
6143 (toc_fusionload_di): Likewise.
6144 (fusion_gpr_load_<mode>): Delete generator function, this insn no
6145 longer needs to be named. Rename predicate to delete TOC fusion.
6146 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6147 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6148 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6149 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6150 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
6152 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
6154 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
6155 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
6156 __int128 in built-in function prototypes.
6157 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
6158 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
6160 2018-07-27 Martin Sebor <msebor@redhat.com>
6162 PR tree-optimization/86696
6163 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
6164 types, including enums.
6165 (handle_char_store): Be prepared for the above function to fail.
6167 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
6169 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
6170 when optimization level is lower than 2 or optimize for size.
6172 2018-07-26 Martin Sebor <msebor@redhat.com>
6174 PR tree-optimization/86043
6175 PR tree-optimization/86042
6176 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
6177 (get_string_cst_length): Rename...
6178 (get_min_string_length): ...to this. Add argument.
6179 (handle_char_store): Extend to handle multi-character stores by
6181 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
6182 * tree.h (initializer_zerop): Add argument.
6184 2018-07-26 Jakub Jelinek <jakub@redhat.com>
6187 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
6188 declare target to variables if they have always,{to,from,tofrom} map
6191 2018-07-26 Martin Liska <mliska@suse.cz>
6194 * lto-wrapper.c: Add linker_output as prefix
6195 for ltrans_output_file.
6197 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
6199 PR rtl-optimization/85805
6200 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
6201 value for hard registers if that was written in the same mode.
6203 2018-07-26 Martin Liska <mliska@suse.cz>
6205 PR gcov-profile/86536
6206 * gcov.c (format_gcov): Use printf format %.*f directly
6207 and do not handle special values.
6209 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6211 * common/config/arc/arc-common.c (arc_option_optimization_table):
6212 Update default optimizations for size.
6214 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6216 * config/arc/arc.md (movsf_insn): Add short instruction selection.
6217 * config/arc/constraints.md (CfZ): New constraint.
6218 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
6219 (subsf3_fpu): Likewise.
6220 (cmpsf_fpu): Likewise.
6221 (cmpsf_fpu_uneq): Likewise.
6223 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6225 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
6227 (arc_is_uncached_mem_p): uncached applies to both the variable and
6230 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6232 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
6235 2018-07-25 David Malcolm <dmalcolm@redhat.com>
6237 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
6238 field "m_scopes" from vec to auto_vec.
6240 2018-07-25 Martin Liska <mliska@suse.cz>
6242 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
6245 2018-07-25 Richard Biener <rguenther@suse.de>
6248 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
6249 special wrt context_die late.
6250 (gen_subprogram_die): Re-use DIEs in local scope.
6252 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
6254 PR tree-optimization/86644
6255 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
6257 2018-07-25 Martin Liska <mliska@suse.cz>
6260 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
6261 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
6263 2018-07-25 Martin Liska <mliska@suse.cz>
6266 * params.def: Explain ASan abbreviation and provide
6267 a documentation link.
6269 2018-07-24 Martin Sebor <msebor@redhat.com>
6271 PR tree-optimization/86622
6272 PR tree-optimization/86532
6273 * builtins.h (string_length): Declare.
6274 * builtins.c (c_strlen): Correct handling of non-constant offsets.
6275 (check_access): Be prepared for non-constant length ranges.
6276 (string_length): Make extern.
6277 * expr.c (string_constant): Only handle the minor non-constant
6278 array index. Use string_constant to compute the length of
6279 a generic string constant.
6281 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
6283 PR tree-optimization/86618
6284 * tree-vect-stmts.c (vectorizable_call): Don't take the address
6285 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
6287 2018-07-24 David Malcolm <dmalcolm@redhat.com>
6289 PR tree-optimization/86636
6290 * json.cc (json::object::set): Fix comment. Add assertions.
6291 (json::array::append): Move here from json.h. Add comment and an
6293 (json::string::string): Likewise.
6294 * json.h (json::array::append): Move to json.cc.
6295 (json::string::string): Likewise.
6296 * optinfo-emit-json.cc
6297 (optrecord_json_writer::impl_location_to_json): Assert that we
6298 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
6299 wrapper around it. Expand the location once, rather than three
6301 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
6302 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
6304 (optrecord_json_writer::optinfo_to_json): Likewise, in four
6305 places. Fix some overlong lines.
6307 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
6309 * config/aarch64/aarch64-simd.md
6310 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
6311 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
6312 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
6313 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
6314 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
6315 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
6316 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
6317 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
6318 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
6320 2018-07-24 Jakub Jelinek <jakub@redhat.com>
6323 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
6324 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
6325 and abs_d == d, do the power of two handling if profitable.
6327 2018-07-24 Richard Biener <rguenther@suse.de>
6329 * match.pd: Add BIT_FIELD_REF canonicalizations.
6331 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6334 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
6336 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6338 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
6339 terminated STRING_CST object.
6341 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6343 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
6345 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6347 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
6348 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
6349 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
6350 the elements into a register.
6351 (rs6000_split_v4si_init_di_reg): Delete.
6352 (rs6000_split_v4si_init): Delete.
6353 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
6354 (vsx_init_v4si): Rewrite as a define_expand.
6356 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6358 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
6359 zero_extend argument from memory): New.
6361 2018-07-22 Martin Sebor <msebor@redhat.com>
6364 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
6365 diagnosing calls with unknown arguments unless -Walloca-larger-than
6366 is restricted to less than PTRDIFF_MAX bytes.
6368 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
6370 * doc/gcov.texi (Invoking Gcov): Editorial changes.
6372 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6374 * pretty-print.c (text_info::set_location): Remove redundant
6375 "line_table" parameter from call to rich_location::set_range.
6377 2018-07-20 Martin Sebor <msebor@redhat.com>
6380 * builtins.c (expand_builtin_alloca): Adjust.
6381 * calls.c (alloc_max_size): Simplify.
6382 * cgraphunit.c (cgraph_node::expand): Adjust.
6383 * common.opt (larger_than_size, warn_frame_larger_than): Remove
6385 (frame_larger_than_size): Same.
6386 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
6387 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
6389 * doc/invoke.texi (GCC Command Options): Document option arguments.
6390 Explain byte-size arguments and suffixes.
6391 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
6392 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
6393 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
6394 * doc/options.texi (UInteger): Expand.
6395 (Host_Wide_Int, ByteSize): Document new properties.
6396 * final.c (final_start_function_1): Include sizes in an error message.
6397 * function.c (frame_offset_overflow): Same.
6398 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
6399 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
6400 Diagnose unbounded alloca calls only for limits of less than
6402 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
6403 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
6405 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
6406 only for limits of less than PTRDIFF_MAX.
6407 * langhooks-def.h (lhd_handle_option): Change function argument
6409 * langhooks.c (lhd_handle_option): Same.
6410 * langhooks.h (handle_option): Same.
6411 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
6413 (var_type, var_type_struct): Same.
6414 (var_set): Handle ByteSize flag.
6415 * optc-gen.awk: Add comments to output to ease debugging. Make
6416 use of HOST_WIDE_INT where appropriate.
6417 * opts-gen-save.awk: Use %lx to format unsigned long.
6418 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
6419 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
6420 arguments. Parse bytes-size suffixes.
6421 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
6422 (enum_value_to_arg): Same.
6423 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
6424 (handle_option): Adjust.
6425 (generate_option): Change function argument to HOST_WIDE_INT.
6426 (cmdline_handle_error): Adjust.
6427 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
6428 (set_option): Change function argument to HOST_WIDE_INT.
6429 (option_enabled): Handle cl_host_wide_int.
6430 (get_option_state): Handle CLVC_SIZE.
6431 (control_warning_option): Same.
6432 * opts.c (common_handle_option): Change function argument to
6433 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
6434 OPT_Wvla_larger_than_.
6435 * opts.h (enum cl_var_type): Add an enumerator.
6436 * stor-layout.c (layout_decl): Print a more meaningful warning.
6437 * toplev.c (output_stack_usage): Adjust.
6439 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
6441 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
6442 call to inline_expand_builtin_string_cmp.
6443 (expand_builtin_strcmp): Likewise.
6444 (expand_builtin_strncmp): Likewise.
6445 (inline_string_cmp): Delete the last parameter, change char_type_node
6446 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
6448 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
6449 the inlining expansion on target where the type of the call has same or
6450 narrower precision than unsigned char.
6452 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6454 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
6455 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
6456 * common.opt (fsave-optimization-record): New option.
6457 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
6458 * doc/invoke.texi (-fsave-optimization-record): New option.
6459 * dumpfile.c: Include "optinfo-emit-json.h".
6460 (struct kv_pair): Move to coretypes.h.
6461 (optgroup_options): Make non-static.
6462 (dump_context::end_scope): Call
6463 optimization_records_maybe_pop_dump_scope.
6464 * dumpfile.h (optgroup_options): New decl.
6465 * json.cc: New file.
6467 * optinfo-emit-json.cc: New file.
6468 * optinfo-emit-json.h: New file.
6469 * optinfo.cc: Include "optinfo-emit-json.h".
6470 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
6471 (optinfo_enabled_p): Check optimization_records_enabled_p.
6472 (optinfo_wants_inlining_info_p): Likewise.
6473 * optinfo.h: Update comment.
6474 * profile-count.c (profile_quality_as_string): New function.
6475 * profile-count.h (profile_quality_as_string): New decl.
6476 (profile_count::quality): New accessor.
6477 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
6478 and optinfo_emit_json_cc_tests.
6479 * selftest.h (selftest::json_cc_tests): New decl.
6480 (selftest::optinfo_emit_json_cc_tests): New decl.
6481 * toplev.c: Include "optinfo-emit-json.h".
6482 (compile_file): Call optimization_records_finish.
6483 (do_compile): Call optimization_records_start.
6484 * tree-ssa-live.c: Include optinfo.h.
6485 (remove_unused_scope_block_p): Retain inlining information if
6486 optinfo_wants_inlining_info_p returns true.
6488 2018-07-20 Richard Biener <rguenther@suse.de>
6491 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
6492 to cover -flto-partition=none.
6494 2018-07-20 Martin Liska <mliska@suse.cz>
6496 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
6497 (get_decl_source_range): Remove unused function.
6499 2018-07-20 Richard Biener <rguenther@suse.de>
6501 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
6502 (struct vn_phi_s): Likewise.
6503 (struct vn_reference_s): Likewise.
6504 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
6505 for searching the slot of an entry known to be in the hash itself.
6506 (vn_phi_hasher::equal): Likewise.
6507 (vn_reference_hasher::equal): Likewise.
6508 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
6510 (optimistic_info, current_info): Remove, keeping only valid_info.
6511 (vn_reference_lookup_1): Remove fallback lookup.
6512 (vn_reference_lookup_2): Likewise.
6513 (vn_nary_op_lookup_1): Likewise.
6514 (vn_phi_lookup): Likewise.
6515 (vn_nary_build_or_lookup_1): Make sure to not chain the built
6517 (vn_reference_insert): Adjust, chain the inserted hash element
6518 at last_inserted_ref.
6519 (vn_reference_insert_pieces): Likewise.
6520 (visit_reference_op_call): Likewise.
6521 (vn_nary_op_insert_into): Chain the inserted hash element at
6523 (vn_nary_op_insert_pieces): Adjust.
6524 (vn_nary_op_insert): Likewise.
6525 (vn_nary_op_insert_stmt): Likewise.
6526 (vn_phi_insert): Adjust, chain the inserted hash element at
6528 (process_scc): Remove clearing and copying the optimistic
6529 table. Instead remove elements inserted during an optimistic
6530 iteration from the single table we maintain.
6531 (init_scc_vn): Adjust.
6532 (free_scc_vn): Likewise.
6533 (sccvn_dom_walker::record_cond): Likewise.
6534 (sccvn_dom_walker::after_dom_children): Likewise.
6536 2018-07-19 Martin Sebor <msebor@redhat.com>
6538 PR tree-optimization/84047
6539 PR tree-optimization/83776
6540 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
6541 (check_array_bounds): Call it.
6543 2018-07-19 Martin Sebor <msebor@redhat.com>
6545 * align.h (align_flags): Use member initialization.
6547 2018-07-19 David Malcolm <dmalcolm@redhat.com>
6549 * Makefile.in (OBJS): Add optinfo.o.
6550 * coretypes.h (class symtab_node): New forward decl.
6551 (struct cgraph_node): New forward decl.
6552 (class varpool_node): New forward decl.
6553 * dump-context.h: New file.
6554 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
6556 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
6557 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
6558 (set_alt_dump_file): Likewise.
6559 (dump_context::~dump_context): New dtor.
6560 (dump_gimple_stmt): Move implementation to...
6561 (dump_context::dump_gimple_stmt): ...this new member function.
6562 Add the stmt to any pending optinfo, creating one if need be.
6563 (dump_gimple_stmt_loc): Move implementation to...
6564 (dump_context::dump_gimple_stmt_loc): ...this new member function.
6565 Start a new optinfo and add the stmt to it.
6566 (dump_gimple_expr): Move implementation to...
6567 (dump_context::dump_gimple_expr): ...this new member function.
6568 Add the stmt to any pending optinfo, creating one if need be.
6569 (dump_gimple_expr_loc): Move implementation to...
6570 (dump_context::dump_gimple_expr_loc): ...this new member function.
6571 Start a new optinfo and add the stmt to it.
6572 (dump_generic_expr): Move implementation to...
6573 (dump_context::dump_generic_expr): ...this new member function.
6574 Add the tree to any pending optinfo, creating one if need be.
6575 (dump_generic_expr_loc): Move implementation to...
6576 (dump_context::dump_generic_expr_loc): ...this new member
6577 function. Add the tree to any pending optinfo, creating one if
6579 (dump_printf): Move implementation to...
6580 (dump_context::dump_printf_va): ...this new member function. Add
6581 the text to any pending optinfo, creating one if need be.
6582 (dump_printf_loc): Move implementation to...
6583 (dump_context::dump_printf_loc_va): ...this new member function.
6584 Start a new optinfo and add the stmt to it.
6585 (dump_dec): Move implementation to...
6586 (dump_context::dump_dec): ...this new member function. Add the
6587 value to any pending optinfo, creating one if need be.
6588 (dump_context::dump_symtab_node): New member function.
6589 (dump_context::get_scope_depth): New member function.
6590 (dump_context::begin_scope): New member function.
6591 (dump_context::end_scope): New member function.
6592 (dump_context::ensure_pending_optinfo): New member function.
6593 (dump_context::begin_next_optinfo): New member function.
6594 (dump_context::end_any_optinfo): New member function.
6595 (dump_context::s_current): New global.
6596 (dump_context::s_default): New global.
6597 (dump_scope_depth): Delete global.
6598 (dumpfile_ensure_any_optinfo_are_flushed): New function.
6599 (dump_symtab_node): New function.
6600 (get_dump_scope_depth): Reimplement in terms of dump_context.
6601 (dump_begin_scope): Likewise.
6602 (dump_end_scope): Likewise.
6603 (selftest::temp_dump_context::temp_dump_context): New ctor.
6604 (selftest::temp_dump_context::~temp_dump_context): New dtor.
6605 (selftest::verify_item): New function.
6606 (ASSERT_IS_TEXT): New macro.
6607 (ASSERT_IS_TREE): New macro.
6608 (ASSERT_IS_GIMPLE): New macro.
6609 (selftest::test_capture_of_dump_calls): New test.
6610 (selftest::dumpfile_c_tests): Call it.
6611 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
6612 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
6613 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
6614 descriptive comment.
6615 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
6616 (dump_node, dump_bb): Move these unrelated decls.
6617 (class dump_manager): Add leading comment.
6618 * optinfo.cc: New file.
6619 * optinfo.h: New file.
6621 2018-07-19 Michael Collison <michael.collison@arm.com>
6622 Richard Henderson <rth@redhat.com>
6624 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
6625 (subti): Handle op1 zero.
6626 (subvti4, usub4ti4): New.
6627 (*sub<GPI>3_compare1_imm): New.
6628 (sub<GPI>3_carryinCV): New.
6629 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
6630 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
6632 2018-07-19 Michael Collison <michael.collison@arm.com>
6633 Richard Henderson <rth@redhat.com>
6635 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
6636 (addti3): Create simpler code if low part is already known to be 0.
6637 (addvti4, uaddvti4): New.
6638 (*add<GPI>3_compareC_cconly_imm): New.
6639 (*add<GPI>3_compareC_cconly): New.
6640 (*add<GPI>3_compareC_imm): New.
6641 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
6642 handle constants within this pattern..
6643 (*add<GPI>3_compareV_cconly_imm): New.
6644 (*add<GPI>3_compareV_cconly): New.
6645 (*add<GPI>3_compareV_imm): New.
6646 (add<GPI>3_compareV): New.
6647 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
6648 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
6649 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
6650 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
6651 with 'comparison' operator.
6652 (*add<GPI>3_compareV_cconly_imm): Ditto.
6653 (*add<GPI>3_compareV_cconly): Ditto.
6654 (*add<GPI>3_compareV_imm): Ditto.
6655 (add<GPI>3_compareV): Ditto.
6656 (add<mode>3_carryinC): Ditto.
6657 (*add<mode>3_carryinC_zero): Ditto.
6658 (*add<mode>3_carryinC): Ditto.
6659 (add<mode>3_carryinV): Ditto.
6660 (*add<mode>3_carryinV_zero): Ditto.
6661 (*add<mode>3_carryinV): Ditto.
6663 2018-07-19 Michael Collison <michael.collison@arm.com>
6664 Richard Henderson <rth@redhat.com>
6666 * config/aarch64/aarch64-modes.def (CC_V): New.
6667 * config/aarch64/aarch64-protos.h
6668 (aarch64_addti_scratch_regs): Declare
6669 (aarch64_subvti_scratch_regs): Declare.
6670 (aarch64_expand_subvti): Declare.
6671 (aarch64_gen_unlikely_cbranch): Declare
6672 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
6673 for signed overflow using CC_Vmode.
6674 (aarch64_get_condition_code_1): Handle CC_Vmode.
6675 (aarch64_gen_unlikely_cbranch): New function.
6676 (aarch64_addti_scratch_regs): New function.
6677 (aarch64_subvti_scratch_regs): New function.
6678 (aarch64_expand_subvti): New function.
6680 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6682 * config/aarch64/aarch64-option-extensions.def: New entry for profile
6684 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
6685 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
6688 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6691 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6692 address check not strict.
6694 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6696 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
6698 (store_pair_lanes<mode>): Likewise.
6699 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
6700 enum value 'ADDR_QUERY_LDP_STP_N'.
6701 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
6702 (aarch64_print_address_internal): Add declaration.
6703 (aarch64_print_ldpstp_address): Remove.
6704 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
6705 (aarch64_print_operand): Change printing of 'y'.
6706 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
6707 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
6708 'true' rather than '1'.
6709 * config/aarch64/constraints.md (Uml): Likewise.
6710 (Uml): Rename to Umn.
6713 2018-07-19 Richard Biener <rguenther@suse.de>
6715 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
6717 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
6718 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
6719 (vn_reference_hasher): Likewise.
6720 (struct vn_tables_s): Remove obstack and alloc-pool members.
6721 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
6722 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
6723 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
6724 (vn_reference_insert_pieces): Likewise.
6725 (alloc_vn_nary_op_noinit): Adjust.
6726 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
6727 (vn_phi_eq): Adjust.
6728 (shared_lookup_phiargs): Remove.
6729 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
6730 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
6731 (visit_reference_op_call): Likewise.
6732 (copy_nary, copy_phi, copy_reference): Remove.
6733 (process_scc): Rewind the obstack when iterating. Do not
6734 copy the elements to valid_info but just move them from one
6735 hashtable to the other.
6736 (allocate_vn_table): Adjust.
6737 (free_vn_table): Likewise.
6738 (init_scc_vn): Likewise.
6739 (free_scc_vn): Likewise.
6741 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
6744 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
6745 indirect_return as function type attribute.
6746 (ix86_attribute_table): Change indirect_return to function
6748 * doc/extend.texi: Update indirect_return attribute.
6750 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
6752 * wide-int.h (widest2_int): New.
6753 * gimple-fold.c (arith_overflowed_p): Use it.
6754 * tree.h (widest2_int_cst): New.
6755 * tree-vrp.c (wide_int_binop_overflow): Rename from
6756 vrp_int_const_binop.
6757 Rewrite to work on trees.
6758 (extract_range_from_multiplicative_op_1): Abstract code to...
6759 (wide_int_range_min_max): ...here.
6760 (wide_int_range_cross_product): ...and here.
6761 (extract_range_from_binary_expr_1): Abstract overflow code to...
6762 (wide_int_range_mult_wrapping): ...here.
6763 * tree-vrp.h (wide_int_range_cross_product): New.
6764 (wide_int_range_mult_wrapping): New.
6766 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
6767 Julia Koval <julia.koval@intel.com>
6769 * config/i386/x86-tune-costs.h (skylake_memcpy,
6770 skylake_memset): Replace rep_prefix with unrolling for size 512.
6772 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
6775 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6776 comparision with EQ_EXPR in last stmt.
6778 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
6780 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6781 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6782 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6783 previously known as "PowerPC AltiVec Built-in Functions". Move
6784 some material to new subsubsections "PowerPC AltiVec Built-in
6785 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6787 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6788 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6789 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6790 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6792 2018-07-18 Richard Biener <rguenther@suse.de>
6794 PR tree-optimization/86557
6795 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
6798 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
6800 * config/s390/s390.c (s390_function_profiler): Generate CFI.
6802 2018-07-17 Jeff Law <law@redhat.com>
6804 * config/arm/arm.c (get_label_padding): Update for recent
6805 changes to label_to_alignment.
6807 PR tree-optimization/86010
6808 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6810 * config/mips/mips.c (vr4130_align_insns): Update for recent
6811 changes to label_to_alignment.
6813 * config/frv/frv.c (frv_label_align): Update for recent changes
6814 to label_to_alignment.
6816 * config/nios2/nios2.c (nios2_label_align): Update for recent
6817 changes which dropped ALIGN_LABELS_LOG.
6819 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
6821 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
6822 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
6824 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
6826 * config/arc/arc.c (arc_label_align): Use align_labels instead of
6827 deprecated align_labels_log.
6829 2018-07-17 Richard Biener <rguenther@suse.de>
6832 * dwarf2out.c (init_sections_and_labels): Always generate
6833 a debug_line_str_section for early LTO debug.
6834 (dwarf2out_finish): Reset debug_line_str_hash output early.
6835 Bump counter for extra dwarf5 .debug_loc labels to not conflict
6837 (dwarf2out_early_finish): Output debug_line_str.
6839 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6841 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
6842 index register on z196 or later.
6844 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
6846 * config/s390/s390.c (s390_default_align): Set default function
6848 (s390_override_options_after_change): Call s390_default align.
6849 (s390_option_override_internal): Call s390_default_align.
6850 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
6852 2018-07-17 Jakub Jelinek <jakub@redhat.com>
6855 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
6856 to _looptemp_ clauses, other than the first two.
6858 2018-07-17 Martin Liska <mliska@suse.cz>
6860 * opts.c: Do not enable OPT_falign_* for -Os.
6862 2018-07-17 Martin Liska <mliska@suse.cz>
6864 * align.h (MAX_CODE_ALIGN): New.
6865 (MAX_CODE_ALIGN_VALUE): New.
6866 * common/config/i386/i386-common.c (ix86_handle_option):
6867 (MAX_CODE_ALIGN): Moved to align.h.
6868 * final.c (MAX_CODE_ALIGN): Likewise.
6869 * opts.c (parse_and_check_align_values):
6870 (MAX_CODE_ALIGN): Likewise.
6871 (MAX_CODE_ALIGN_VALUE): Likewise.
6873 2018-07-17 Martin Liska <mliska@suse.cz>
6875 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
6876 in order to fulfil coding style.
6877 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
6878 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6879 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6880 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
6881 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
6882 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
6883 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
6884 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
6885 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6887 2018-07-17 Martin Liska <mliska@suse.cz>
6889 * align.h: New file.
6890 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
6892 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
6893 align_flags of label_to_alignment.
6894 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
6896 * config/m68k/m68k.c: Do not use removed align_labels_value and
6898 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
6899 (LOOP_ALIGN): Likewise.
6900 (LABEL_ALIGN): Likewise.
6901 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6902 Remove not used macro.
6903 (rs6000_loop_align): Change return type to align_flags.
6904 (rs6000_loop_align_max_skip): Remove.
6905 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
6906 Change return type to align_flags.
6907 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
6908 Remove not used macro.
6909 (rs6000_loop_align): Change return type to align_flags.
6910 (rs6000_loop_align_max_skip): Remove.
6911 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
6912 * config/rx/rx-protos.h (rx_align_for_label): Make it
6914 * config/rx/rx.c (rx_align_for_label): Change return type
6916 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
6918 into align_flags class.
6919 (LABEL_ALIGN): Likewise.
6920 (LOOP_ALIGN): Likewise.
6921 * config/s390/s390.c (s390_label_align): Use align_flags
6923 (s390_asm_output_function_label): Likewise.
6924 * config/sh/sh.c (sh_override_options_after_change):
6925 Use align_flags class directly without macros.
6926 (find_barrier): Likewise.
6927 (barrier_align): Likewise.
6928 (sh_loop_align): Likewise.
6929 * config/spu/spu.c (spu_option_override):
6930 Use align_flags_tuple::get_value instead of removed macros.
6931 (spu_sched_init): Likewise.
6932 * config/spu/spu.h (GTY): Likewise.
6933 * config/visium/visium.c (visium_option_override):
6934 Set "8" as default secondary alignment.
6935 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
6936 in order to guarantee secondary alignment of 8.
6937 * coretypes.h: Include align.h header file.
6938 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
6939 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6940 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
6941 * doc/tm.texi.in: Likewise.
6942 * final.c (struct label_alignment): Remove not used structure.
6943 (LABEL_ALIGN): Change type to align_flags.
6944 (LOOP_ALIGN): Likewise.
6945 (JUMP_ALIGN): Likewise.
6946 (default_loop_align_max_skip): Remove.
6947 (default_label_align_max_skip): Likewise.
6948 (default_jump_align_max_skip): Likewise.
6949 (default_label_align_after_barrier_max_skip):
6950 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
6951 (LABEL_TO_MAX_SKIP): Remove.
6952 (label_to_alignment): Return align_flags type instead of integer.
6953 (label_to_max_skip): Remove.
6954 (align_fuzz): Use align_flags type.
6955 (compute_alignments): Use align_flags type and use align_flags::max
6956 to combine multiple alignments.
6957 (grow_label_align): Grow vec instead of C array.
6958 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
6959 (shorten_branches): Use align_flags type and use align_flags::max
6960 to combine multiple alignments.
6961 (final_scan_insn_1): Remove usage of secondary alignment that comes
6962 from label alignment, but instead use proper secondary alignment
6963 which is computed in grow_label_align.
6964 * flags.h (struct align_flags_tuple): Move to align.h.
6965 (struct align_flags): Likewise.
6966 (state_align_loops): Rename to align_loops.
6967 (state_align_jumps): Rename to align_jumps.
6968 (state_align_labels): Rename to align_labels.
6969 (state_align_functions): Rename to align_functions.
6970 (align_loops_log): Remove.
6971 (align_jumps_log): Remove.
6972 (align_labels_log): Remove.
6973 (align_functions_log): Remove.
6974 (align_loops_max_skip): Remove.
6975 (align_jumps_max_skip): Remove.
6976 (align_labels_max_skip): Remove.
6977 (align_functions_max_skip): Remove.
6978 (align_loops_value): Remove.
6979 (align_jumps_value): Remove.
6980 (align_labels_value): Remove.
6981 (align_functions_value): Remove.
6982 * output.h (label_to_alignment): Change return type to align_flags.
6983 (label_to_max_skip): Remove.
6984 * target.def: Remove loop_align_max_skip, label_align_max_skip,
6985 jump_align_max_skip macros.
6986 * targhooks.h (default_loop_align_max_skip): Remove.
6987 (default_label_align_max_skip): Likewise.
6988 (default_jump_align_max_skip): Likewise.
6989 (default_label_align_after_barrier_max_skip): Remove.
6990 * toplev.c (read_log_maxskip): Use ::normalize function.
6991 (parse_N_M): Remove not used argument and also call ::normalize.
6992 (parse_alignment_opts): Do not pass unused arguments.
6993 * varasm.c (assemble_start_function): Use directly align_functions
6994 instead of removed macros.
6995 * system.h: Do not poison removed macros.
6997 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7000 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
7001 and cond temporaries don't have reference type if iterator has
7002 pointer type. For init use &for_pre_body instead of pre_p if
7003 for_pre_body is non-empty.
7005 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7007 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
7008 double-double modes to SFmode directly directly.
7009 (trunc<mode>sf2_fprs): Delete.
7011 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7013 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
7014 for conversions between IFmode and the decimal floating point modes.
7015 (init_float128_ieee): Use the correct names for conversions between
7016 KFmode and the decimal floating point modes.
7018 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7020 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
7021 for the conversions between TDmode and IFmode.
7022 (init_float128_ieee): Use more correct names for the conversions
7023 between TDmode and KFmode.
7025 2018-07-16 Jakub Jelinek <jakub@redhat.com>
7027 PR tree-optimization/86526
7028 * builtins.c (expand_builtin_memcmp): Formatting fixes.
7029 (inline_expand_builtin_string_cmp): Likewise.
7030 (inline_string_cmp): Likewise. Use c_readstr instead of
7031 builtin_memcpy_read_str. Add unit_mode temporary.
7033 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
7036 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
7037 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
7039 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7041 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
7042 Alphabetize prototypes of built-in functions, separating out
7043 built-in functions that are listed in this section but should be
7044 described elsewhere.
7046 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
7049 * expmed.c (emit_store_flag): Do not emit setcc followed by a
7050 conditional move when trapping comparison was split to a
7051 non-trapping one (and vice versa).
7053 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7055 * config/s390/s390.c (s390_function_profiler): Generate nops
7056 instead of profiler call sequences.
7057 * config/s390/s390.opt: Add the new option.
7059 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7061 * config/s390/s390.c (s390_function_profiler): Generate
7062 __mcount_loc section.
7063 * config/s390/s390.opt: Add the new option.
7065 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7067 * common.opt: Add the new warning.
7068 * config/s390/s390.c (s390_function_profiler): Emit "brasl
7069 %r0,__fentry__" when -mfentry is specified.
7070 (s390_option_override_internal): Disallow -mfentry for 31-bit
7072 * config/s390/s390.opt: Add the new option.
7074 2018-07-16 Richard Biener <rguenther@suse.de>
7077 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
7078 for function-local FUNCTION_DECL and RESULT_DECL immediately.
7080 2018-07-16 Martin Liska <mliska@suse.cz>
7083 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
7086 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
7088 * config/arc/arcHS.md: Update ARCHS scheduling rules.
7090 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
7092 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
7094 * config/arc/arc-cpus.def (hs4x): New cpu.
7096 * config/arc/arc-tables.opt: Regenerate.
7097 * config/arc/arc.c (arc_sched_issue_rate): New function.
7098 (TARGET_SCHED_ISSUE_RATE): Define.
7099 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
7100 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
7102 (attr tune): Add ARCHS4x tune values.
7103 (attr tune_dspmpy): Define.
7104 (*tst): Correct instruction type.
7105 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
7106 * config/arc/arcHS4x.md: New file.
7107 * config/arc/fpu.md: Update instruction type attributes.
7108 * config/arc/t-multilib: Regenerate.
7110 2018-07-16 Tom de Vries <tdevries@suse.de>
7113 * var-tracking.c (vt_initialize): Fix pre_dec handling.
7115 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7117 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
7120 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
7122 PR tree-optimization/86514
7123 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
7124 conversion to a boolean type from a type with greater precision.
7126 2018-07-16 Tom de Vries <tdevries@suse.de>
7128 * var-tracking.c (vt_initialize): Print adjusted insn slim if
7129 dump_flags request TDF_SLIM.
7131 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
7133 * fold-const.c (int_const_binop_1): Abstract...
7134 (wide_int_binop): ...wide int code here.
7135 (poly_int_binop): ...poly int code here.
7136 Abstract the rest of int_const_binop_1 into int_const_binop.
7137 * fold-const.h (wide_int_binop): New.
7138 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
7139 Remove useless PLUS/MINUS_EXPR case.
7140 (zero_nonzero_bits_from_vr): Move wide int code...
7141 (zero_nonzero_bits_from_bounds): ...here.
7142 (extract_range_from_binary_expr_1): Move mask optimization code...
7143 (range_easy_mask_min_max): ...here.
7144 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
7145 (range_easy_mask_min_max): New.
7147 2018-07-15 Jeff Law <law@redhat.com>
7150 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
7153 2018-07-14 Jim Wilson <jimw@sifive.com>
7155 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
7157 2018-07-14 Paul Koning <ni1d@arrl.net>
7159 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
7161 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
7163 * lto-streamer-out.c (copy_function_or_variable): Dump info about
7166 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
7167 Steve Munroe <munroesj52@gmail.com>
7169 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
7170 (_mm_andnot_si128): Likewise.
7171 (_mm_or_si128): Likewise.
7172 (_mm_xor_si128): Likewise.
7174 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
7177 * builtins.c (expand_builtin_memcmp): Inline the calls first
7178 when result_eq is false.
7179 (expand_builtin_strcmp): Inline the calls first.
7180 (expand_builtin_strncmp): Likewise.
7181 (inline_string_cmp): New routine. Expand a string compare
7182 call by using a sequence of char comparison.
7183 (inline_expand_builtin_string_cmp): New routine. Inline expansion
7184 a call to str(n)cmp/memcmp.
7185 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
7187 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
7189 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
7191 * config/arm/driver-arm.c: Include arm-native.h.
7192 (host_detect_local_cpu): Use auto-generated data tables.
7193 (vendors, arm_cpu_table): Delete. Move part information to ...
7194 * config/arm/arm-cpus.in: ... here.
7195 * config/arm/parsecpu.awk (gen_native): New function.
7196 (vendor, part): New CPU fields.
7197 (END): Add support for building the native CPU detection tables.
7198 * config/arm/t-arm (arm-native.h): Add build rule.
7199 (driver-arm.o): Add dependency on arm-native.h.
7201 2018-07-13 Richard Biener <rguenther@suse.de>
7204 * match.pd (addr1 - addr2): Allow either of the operand to
7207 2018-07-13 Tom de Vries <tdevries@suse.de>
7209 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
7212 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
7214 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
7215 arrays instead of numbered variables.
7217 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
7219 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
7220 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
7222 2018-07-13 Richard Biener <rguenther@suse.de>
7225 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
7226 instead of get_context_die.
7228 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
7229 Richard Biener <rguenther@suse.de>
7232 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
7233 that the loop latch destination where phi is defined.
7235 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
7237 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
7238 (riscv_expand_epilogue): Add assertion to check interrupt mode.
7239 (riscv_set_current_function): Extract getting interrupt type to new
7241 (riscv_get_interrupt_type): New function.
7242 (riscv_merge_decl_attributes): New function, checking interrupt type is
7244 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
7246 2018-07-12 Paul Koning <ni1d@arrl.net>
7248 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
7251 2018-07-12 Paul Koning <ni1d@arrl.net>
7253 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
7254 zero reference, add doloop_end instead.
7255 * doc/md.texi (decrement_and_branch_until_zero): Remove.
7256 (Looping patterns): Remove decrement_and_branch_until_zero. Add
7257 detail for doloop_end.
7259 2018-07-12 Martin Sebor <msebor@redhat.com>
7262 * attribs.c (decl_attributes): Reject conflicting attributes before
7263 calling attribute handlers.
7265 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7267 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
7269 (gcc::dump_manager::get_dump_file_name): likewise.
7270 (dump_begin): Likewise.
7271 * dumpfile.h (dump_begin): Update prototype.
7272 (gcc::dump_manager::get_dump_file_name,
7273 gcc::dump_manager::get_dump_file_name): Update prototype.
7275 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7277 * internal-fn.h (vectorizable_internal_fn_p): New function.
7278 * tree-vect-slp.c (compatible_calls_p): Likewise.
7279 (vect_build_slp_tree_1): Remove nops argument. Handle calls
7280 to internal functions.
7281 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
7283 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7285 * fold-const.h (inverse_conditions_p): Declare.
7286 * fold-const.c (inverse_conditions_p): New function.
7287 * match.pd: Use inverse_conditions_p. Add folds of view_converts
7288 that test the inverse condition of a conditional internal function.
7289 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
7290 * internal-fn.c (internal_fn_mask_index): Handle conditional
7292 (vectorized_internal_fn_supported_p): New function.
7293 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
7294 (any_pred_load_store): Replace with...
7295 (need_to_predicate): ...this new variable.
7296 (redundant_ssa_names): New variable.
7297 (ifcvt_can_use_mask_load_store): Move initial checks to...
7298 (ifcvt_can_predicate): ...this new function. Handle tree codes
7299 for which a conditional internal function exists.
7300 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
7301 instead of ifcvt_can_use_mask_load_store. Update after variable
7303 (predicate_load_or_store): New function, split out from
7304 predicate_mem_writes.
7305 (check_redundant_cond_expr): New function.
7306 (value_available_p): Likewise.
7307 (predicate_rhs_code): Likewise.
7308 (predicate_mem_writes): Rename to...
7309 (predicate_statements): ...this. Use predicate_load_or_store
7310 and predicate_rhs_code.
7311 (combine_blocks, tree_if_conversion): Update after above name changes.
7312 (ifcvt_local_dce): Handle redundant_ssa_names.
7313 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
7314 general conditional functions.
7315 * tree-vect-stmts.c (vectorizable_call): Likewise.
7317 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7318 Alan Hayward <alan.hayward@arm.com>
7319 David Sherwood <david.sherwood@arm.com>
7321 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
7322 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
7323 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
7324 plus and minus and convert them into IFN_COND_FMA-based sequences.
7325 (convert_mult_to_fma): Handle conditional plus and minus.
7327 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7329 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
7330 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
7331 (cond_fnms_optab): New optabs.
7332 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
7334 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
7335 * internal-fn.h (get_conditional_internal_fn): Declare.
7336 (get_unconditional_internal_fn): Likewise.
7337 * internal-fn.c (cond_ternary_direct): New macro.
7338 (expand_cond_ternary_optab_fn): Likewise.
7339 (direct_cond_ternary_optab_supported_p): Likewise.
7340 (FOR_EACH_COND_FN_PAIR): Likewise.
7341 (get_conditional_internal_fn): New function.
7342 (get_unconditional_internal_fn): Likewise.
7343 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
7344 (gimple_match_op::gimple_match_op): Add a new overload for 5
7346 (gimple_match_op::set_op): Likewise.
7347 (gimple_resimplify5): Declare.
7348 * genmatch.c (decision_tree::gen): Generate simplifications for
7350 * gimple-match-head.c (gimple_simplify): Define an overload for
7351 5 operands. Handle calls with 5 arguments in the top-level overload.
7352 (convert_conditional_op): Handle conversions from unconditional
7353 internal functions to conditional ones.
7354 (gimple_resimplify5): New function.
7355 (build_call_internal): Pass a fifth operand.
7356 (maybe_push_res_to_seq): Likewise.
7357 (try_conditional_simplification): Try converting conditional
7358 internal functions to unconditional internal functions.
7359 Handle 3-operand unconditional forms.
7360 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
7361 Define ternary equivalents of the current rules for binary conditional
7363 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
7365 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
7366 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
7367 (optab): Handle them.
7368 (SVE_COND_FP_TERNARY): New int iterator.
7369 (sve_fmla_op, sve_fmad_op): New int attributes.
7370 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
7371 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
7372 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
7374 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7376 * target.def (preferred_else_value): New target hook.
7377 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
7378 * doc/tm.texi: Regenerate.
7379 * targhooks.h (default_preferred_else_value): Declare.
7380 * targhooks.c (default_preferred_else_value): New function.
7381 * internal-fn.h (conditional_internal_fn_code): Declare.
7382 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
7383 (get_conditional_internal_fn): Use it.
7384 (conditional_internal_fn_code): New function.
7385 * gimple-match.h (gimple_match_cond): New struct.
7386 (gimple_match_op): Add a cond member function.
7387 (gimple_match_op::gimple_match_op): Update all forms to take a
7389 * genmatch.c (expr::gen_transform): Use the same condition as res_op
7390 for the suboperation, but don't specify a particular else_value.
7391 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
7392 (visit_nary_op, visit_reference_op_load): Pass
7393 gimple_match_cond::UNCOND to the gimple_match_op constructor.
7394 * gimple-match-head.c: Include tree-eh.h
7395 (convert_conditional_op): New function.
7396 (maybe_resimplify_conditional_op): Likewise.
7397 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
7398 (gimple_resimplify2): Likewise.
7399 (gimple_resimplify3): Likewise.
7400 (gimple_resimplify4): Likewise.
7401 (maybe_push_res_to_seq): Return null for conditional operations.
7402 (try_conditional_simplification): New function.
7403 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
7405 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
7407 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
7409 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
7411 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7413 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7415 (hash_tree): Do not hash DECL_FCONTEXT
7416 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
7417 Do not stream DECL_FCONTEXT.
7418 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
7419 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
7421 2018-07-12 Richard Biener <rguenther@suse.de>
7424 * dwarf2out.c (gen_block_die): Only output blocks when they have
7425 at least one !DECL_IGNORED_P variable.
7427 2018-07-12 Richard Biener <rguenther@suse.de>
7430 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
7431 Remove -mieee-fp handling.
7433 2018-07-12 Richard Biener <rguenther@suse.de>
7435 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
7436 left-over from last patch.
7438 2018-07-12 Jakub Jelinek <jakub@redhat.com>
7440 PR tree-optimization/86492
7441 * gimple-ssa-store-merging.c
7442 (imm_store_chain_info::coalesce_immediate_stores): Call
7443 check_no_overlap even for the merge_overlapping case. Formatting fix.
7445 2018-07-12 Richard Biener <rguenther@suse.de>
7448 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
7449 move possibly trapping operations into the conditional.
7451 2018-07-12 Richard Biener <rguenther@suse.de>
7453 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
7454 (vn_lookup_simplify_result): Remove recursion limit applied
7456 (vn_nary_build_or_lookup_1): Adjust.
7457 (try_to_simplify): Likewise.
7458 * gimple-match-head.c (gimple_resimplify1): Instead apply one
7460 (gimple_resimplify2): Likewise.
7461 (gimple_resimplify3): Likewise.
7462 (gimple_resimplify4): Likewise.
7464 2018-07-11 Jakub Jelinek <jakub@redhat.com>
7466 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
7467 Use __mmask64 type instead of __mmask8 for __M argument.
7468 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
7469 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
7471 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
7472 __mmask16 for __M argument.
7473 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
7474 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
7475 to __mmask16 instead of __mmask8.
7476 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
7477 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
7478 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
7479 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
7480 instead of __mmask16 for __U argument.
7481 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
7482 __mmask16 instead of __mmask8 for __U argument.
7483 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
7485 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
7487 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
7489 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
7491 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
7493 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
7495 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
7497 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
7499 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
7501 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
7503 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
7504 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
7505 return type as well as __M argument type and all casts from __mmask8
7507 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
7508 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
7509 return type as well as __M argument type and all casts from __mmask8
7511 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
7512 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
7513 return type as well as __M argument type and all casts from __mmask8
7515 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
7516 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
7517 return type as well as __M argument type and all casts from __mmask8
7519 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
7520 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
7523 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
7525 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
7526 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
7529 2018-07-11 Paul Koning <ni1d@arrl.net>
7531 * doc/md.texi (define_subst): Document how multiple occurrences of
7532 the same argument in the replacement pattern are handled.
7534 2018-07-11 Paul Koning <ni1d@arrl.net>
7536 * doc/extend.texi (Common Variable Attributes): Move "mode" into
7538 (Common Type Attributes): Add "mode" attribute.
7540 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
7542 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
7543 stream DECL_ORIGINAL_TYPE.
7544 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
7545 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
7546 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
7547 Do not walk original type.
7548 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
7550 (write_ts_decl_non_common_tree_pointers): Do not stream
7552 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
7553 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
7555 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7557 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
7558 thread twice from the same starting edge.
7560 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7562 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
7563 * gimple.c (gimple_call_nonnull_result_p): ...here...
7564 (gimple_call_nonnull_arg): ...and here.
7565 * gimple.h (gimple_call_nonnull_result_p): New.
7566 (gimple_call_nonnull_arg): New.
7568 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
7570 * config/arm/arm-cpus.in: Move information from fpu field of each
7571 cpu definition to the isa field.
7572 * config/arm/parsecpu.awk (fpu): Delete match rule.
7573 (gen_comm_data): Don't add bits from the CPU's FPU entry.
7575 2018-07-11 Richard Biener <rguenther@suse.de>
7578 * dwarf2out.c (init_sections_and_labels): Use
7579 output_asm_line_debug_info consistently.
7580 (dwarf2out_early_finish): Likewise.
7581 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
7584 2018-07-11 Richard Biener <rguenther@suse.de>
7586 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
7587 Rework father_bb setting in a way to avoid propagating constants
7588 multiple times on a loop body.
7590 2018-07-10 Mark Wielaard <mark@klomp.org>
7593 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
7595 2018-07-10 Richard Biener <rguenther@suse.de>
7597 * hash-map.h (hash_map::iterator::operator*): Return
7598 references to key and value.
7600 2018-07-10 Jakub Jelinek <jakub@redhat.com>
7603 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
7604 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
7605 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
7606 (gimplify_omp_for): For composite loops, move outer
7607 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
7608 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
7609 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
7610 TREE_LIST for both the original class iterator and the "last" helper
7611 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
7612 loop, remember has_decl_expr from outer composite loops for the
7613 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
7615 2018-07-09 Martin Sebor <msebor@redhat.com>
7619 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
7620 accessing implicitly initialized array elements.
7621 * expr.c (string_constant): Handle string initializers of
7622 character arrays within aggregates.
7623 * gimple-fold.c (fold_array_ctor_reference): Add argument.
7624 Store element offset. As a special case, handle zero size.
7625 (fold_nonarray_ctor_reference): Same.
7626 (fold_ctor_reference): Add argument. Store subobject offset.
7627 * gimple-fold.h (fold_ctor_reference): Add argument.
7629 2018-07-09 Paul Koning <ni1d@arrl.net>
7631 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
7632 (pdp11_insn_cost): New function.
7633 (pdp11_md_asm_adjust): New function.
7634 (TARGET_INVALID_WITHIN_DOLOOP): Define.
7635 (pdp11_rtx_costs): Update to match machine better.
7636 (output_addr_const_pdp11): Correct format mismatch warnings.
7637 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
7638 * config/pdp11/pdp11.md: General change to add base_cost and/or
7639 length attributes for use by new pdp11_insn_cost function.
7640 (MIN_BRANCH): Correct definition.
7642 (doloop_end): Use standard pattern name for looping pattern.
7643 (doloop_end_nocc): New.
7644 (movsf): Add another constraint alternative.
7645 (zero_extendqihi2): Add constraint alternatives for not in place
7647 (zero_extendhisi2): Remove.
7648 (shift patterns): Add CC handling variants.
7652 (define_peephole2): New peephole to recognize mov that sets CC for
7655 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7657 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
7658 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
7661 2018-07-09 Jakub Jelinek <jakub@redhat.com>
7664 * real.c (real_nextafter): Return true if result is denormal.
7666 2018-07-09 Martin Liska <mliska@suse.cz>
7668 * common.opt: Add back wrongly removed attribute.
7670 2018-07-09 Richard Biener <rguenther@suse.de>
7673 * dwarf2out.c (gen_block_die): For an early generated DIE
7674 always output high/low PC attributes.
7676 2018-07-09 Tom de Vries <tdevries@suse.de>
7678 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
7679 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
7680 onto VAR_DECL with abstract origin.
7682 2018-07-07 Jim Wilson <jimw@sifive.com>
7684 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
7686 2018-07-07 Tom de Vries <tdevries@suse.de>
7688 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
7690 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
7692 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
7694 (combine_bound): Use wide-int overflow calculation instead of
7696 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
7698 * fold-const.c (int_const_binop_2): Same.
7699 (extract_muldiv_1): Same.
7700 (fold_div_compare): Same.
7701 (fold_abs_const): Same.
7703 * poly-int.h (add): Same.
7707 * predict.c (predict_iv_comparison): Same.
7708 * profile-count.c (slow_safe_scale_64bit): Same.
7709 * simplify-rtx.c (simplify_const_binary_operation): Same.
7710 * tree-chrec.c (tree_fold_binomial): Same.
7711 * tree-data-ref.c (split_constant_offset_1): Same.
7712 * tree-if-conv.c (idx_within_array_bound): Same.
7713 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
7714 * tree-ssa-phiopt.c (minmax_replacement): Same.
7715 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
7716 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
7717 * vr-values.c (vr_values::adjust_range_with_scev): Same.
7718 * wide-int.cc (wi::add_large): Same.
7719 (wi::mul_internal): Same.
7720 (wi::sub_large): Same.
7721 (wi::divmod_internal): Same.
7722 * wide-int.h: Change overflow type to overflow_type for neg, add,
7723 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
7724 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
7725 mul_internal, divmod_internal.
7726 (overflow_type): New enum.
7727 (accumulate_overflow): New.
7729 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
7731 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
7732 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
7734 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
7736 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
7737 argument is checked for zero before entering loop, avoid checking again.
7739 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
7741 * gimplify.h (generic_expr_could_trap_p): Set as global function.
7742 * gimplify.h (generic_expr_could_trap_p): Likwise.
7743 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
7745 2018-07-06 Jakub Jelinek <jakub@redhat.com>
7747 PR tree-optimization/86401
7748 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
7749 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
7750 (fold_bit_and_mask): ... here. New helper function for match.pd.
7751 * fold-const.h (fold_bit_and_mask): Declare.
7752 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
7754 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
7757 * target.def (translate_mode_attribute): New hook.
7758 * targhooks.h (default_translate_mode_attribute): Declare.
7759 * targhooks.c (default_translate_mode_attribute): New function.
7760 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
7761 * doc/tm.texi: Regenerate.
7762 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
7763 (rs6000_translate_mode_attribute): New function.
7765 2018-07-06 Paul Koning <ni1d@arrl.net>
7767 * doc/md.texi (define_split): Document DONE and FAIL.
7768 (define_peephole2): Ditto.
7770 2018-07-05 Jeff Law <law@redhat.com>
7772 PR tree-optimization/86010
7773 * tree-ssa-dse.c (compute_trims): More aggressively trim at
7774 both the head and tail of mem* and str* calls.
7776 2018-07-05 Jim Wilson <jimw@sifive.com>
7778 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7780 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
7782 * config/aarch64/aarch64-simd.md: correct flags text for
7783 MIN_EXPR replacement.
7785 2018-07-05 James Clarke <jrtc27@jrtc27.com>
7787 * configure: Regenerated.
7789 2018-07-05 Carl Love <cel@us.ibm.com>
7791 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7792 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7793 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7794 VSX_BUILTIN_DOUBLEL_V4SF.
7796 2018-07-05 Martin Sebor <msebor@redhat.com>
7799 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
7800 than its domain to compute its the upper bound of a char array.
7802 2018-07-05 Nathan Sidwell <nathan@acm.org>
7804 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
7805 * doc/cpp.texi: Update comment.
7806 * doc/tm.texi: Rebuilt.
7807 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
7808 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
7809 * doc/extend.texi (Backwards Compatibility): Clarify it is system
7810 headers affected by extern "C".
7811 * system.h: Poison NO_IMPLICIT_EXTERN_C.
7812 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
7813 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
7814 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
7815 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
7816 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
7817 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
7818 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
7819 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
7820 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
7821 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
7822 config/sparc/sp64-elf.h, config/spu/spu.h,
7823 config/stormy16/stormy16.h, config/v850/v850.h,
7824 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
7825 define NO_IMPLICIT_EXTERN_C.
7826 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
7828 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7831 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7832 instead of GET_MODE_SIZE when comparing Units.
7834 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7837 * rtlanal.c (set_noop_p): Constrain on mode change,
7838 include hard-reg-set.h
7840 2018-07-05 Tamar Christina <tamar.christina@arm.com>
7842 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
7844 2018-07-05 Jakub Jelinek <jakub@redhat.com>
7847 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7850 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7852 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7854 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
7857 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
7859 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
7861 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
7864 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
7866 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
7868 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
7869 neg_*_op* variables.
7871 2018-07-04 Martin Liska <mliska@suse.cz>
7873 * tree-switch-conversion.c: Define
7874 max_ratio_for_speed and max_ratio_for_size constants.
7876 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7877 Martin Liska <mliska@suse.cz>
7882 * common.opt: Rename align options with 'str_' prefix.
7883 * common/config/i386/i386-common.c (set_malign_value): New
7885 (ix86_handle_option): Use it to set -falign-* options/
7886 * config/aarch64/aarch64-protos.h (struct tune_params): Change
7887 type from int to string.
7888 * config/aarch64/aarch64.c: Update default values from int
7890 * config/alpha/alpha.c (alpha_override_options_after_change):
7892 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
7893 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7894 max skip conditionally.
7895 * config/i386/freebsd.h (SUBALIGN_LOG): New.
7896 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7897 max skip conditionally.
7898 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7899 max skip conditionally.
7900 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
7901 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7902 max skip conditionally.
7903 * config/i386/i386.c (struct ptt): Change type from int to
7905 (ix86_default_align): Set default values.
7906 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
7907 max skip conditionally.
7908 * config/i386/iamcu.h (SUBALIGN_LOG): New.
7909 (ASM_OUTPUT_MAX_SKIP_ALIGN):
7910 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
7911 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7912 max skip conditionally.
7913 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
7914 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
7915 * config/i386/x86-64.h (SUBALIGN_LOG): New.
7916 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
7917 max skip conditionally.
7918 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7919 * config/ia64/ia64.c (ia64_option_override): Set default values
7920 for alignment options.
7921 * config/m68k/m68k.c: Handle new str_align_* options.
7922 * config/mips/mips.c (mips_set_compression_mode): Change
7924 (mips_option_override): Set default values for options.
7925 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7927 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7929 * config/rx/rx.c (rx_option_override): Likewise.
7930 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
7931 (LABEL_ALIGN): Use align_labels_log.
7932 (LOOP_ALIGN): Use align_loops_align.
7933 * config/s390/s390.c (s390_asm_output_function_label): Use new
7935 * config/sh/sh.c (sh_override_options_after_change):
7936 Change type of constants.
7937 * config/spu/spu.c (spu_sched_init): Likewise.
7938 * config/sparc/sparc.c (sparc_option_override): Set default
7940 * config/visium/visium.c (visium_option_override): Likewise.
7941 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
7942 emit p2align format with last argument if it's not needed.
7943 * doc/invoke.texi: Document extended format of -falign-*.
7944 * final.c: Use align_labels alignment.
7945 * flags.h (struct target_flag_state): Change type to use
7947 (struct align_flags_tuple): New.
7948 (struct align_flags): Likewise.
7949 (align_loops_log): Redefine macro to use new types.
7950 (align_loops_max_skip): Redefine macro to use new types.
7951 (align_jumps_log): Redefine macro to use new types.
7952 (align_jumps_max_skip): Redefine macro to use new types.
7953 (align_labels_log): Redefine macro to use new types.
7954 (align_labels_max_skip): Redefine macro to use new types.
7955 (align_functions_log): Redefine macro to use new types.
7956 (align_loops): Redefine macro to use new types.
7957 (align_jumps): Redefine macro to use new types.
7958 (align_labels): Redefine macro to use new types.
7959 (align_functions): Redefine macro to use new types.
7960 (align_functions_max_skip): Redefine macro to use new types.
7961 (align_loops_value): New macro.
7962 (align_jumps_value): New macro.
7963 (align_labels_value): New macro.
7964 (align_functions_value): New macro.
7965 * function.c (invoke_set_current_function_hook): Propagate
7966 alignment values from flags to global variables default in
7968 * ipa-icf.c (sem_function::equals_wpa): Use
7969 cl_optimization_option_eq instead of memcmp.
7970 * lto-streamer.h (cl_optimization_stream_out): Support streaming
7972 (cl_optimization_stream_in): Likewise.
7973 * optc-save-gen.awk: Support strings in cl_optimization.
7974 * opth-gen.awk: Likewise.
7975 * opts.c (finish_options): Remove error checking of invalid
7977 (MAX_CODE_ALIGN): Remove.
7978 (MAX_CODE_ALIGN_VALUE): Likewise.
7979 (parse_and_check_align_values): New function.
7980 (check_alignment_argument): Likewise.
7981 (common_handle_option): Use check_alignment_argument.
7982 * opts.h (parse_and_check_align_values): Declare.
7983 * toplev.c (init_alignments): Remove.
7984 (read_log_maxskip): New.
7985 (parse_N_M): Likewise.
7986 (parse_alignment_opts): Likewise.
7987 (backend_init_target): Remove usage of init_alignments.
7988 * toplev.h (parse_alignment_opts): Declare.
7989 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
7991 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
7992 * tree.c (cl_option_hasher::equal): New.
7993 * varasm.c: Use new global macros.
7995 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
7997 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
7998 Use a simpler align directive also if MAXSKIP = ALIGN-1.
7999 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8000 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8001 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8002 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8003 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
8004 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
8005 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
8006 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8007 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8008 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8009 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8011 2018-07-04 Martin Liska <mliska@suse.cz>
8012 Jonathan Wakely <jwakely@redhat.com>
8014 * coverage.c: Use correct type.
8015 * doc/invoke.texi: Language correction.
8017 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
8020 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
8021 ENDBRANCH for non-tail call which may return via indirect branch.
8022 * doc/extend.texi: Document indirect_return attribute.
8024 2018-07-03 Martin Sebor <msebor@redhat.com>
8026 PR tree-optimization/86274
8027 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
8029 (format_floating): Correct handling of infinities and NaNs.
8031 2018-07-03 Martin Sebor <msebor@redhat.com>
8033 * print-tree.c (print_real_cst): New function.
8034 (print_node_brief): Call it.
8035 (print_node): Ditto.
8037 2018-07-03 Jeff Law <law@redhat.com>
8039 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
8040 into a single pattern.
8042 * config/h8300/h8300.md (ors code_iterator): New.
8043 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
8044 a single pattern and single splitter.
8045 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
8046 (iorqi3_1, xorqi3_1): Likewise.
8047 (iorqi3, xorqi3 expanders): Similarly.
8049 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
8050 (movmd_internal) into a single pattern using the P mode iterator.
8051 (movmd splitters): Similarly.
8052 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
8053 (movsd splitters): Similarly.
8055 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
8056 ADDB, ADDW and ADDL into a single ADD attribute which selects the
8057 right table based on the size of the operand.
8058 * config/h8300/h8300.md (length_table): Corresponding changes. All
8059 references to "addb", "addw" and "addl" changed to "add".
8060 (btst patterns): Merge two variants into a single pattern.
8061 (tstqi, tsthi): Likewise.
8062 (addhi3_incdec, addsi3_incdec): Likewise.
8063 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
8064 (mulhi3, mulsi3): Likewise.
8065 (udivhi3, udivsi3): Likewise.
8066 (divhi3, divsi3): Likewise.
8067 (andorqi3, andorhi3, andorsi3): Likewise.
8069 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
8072 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
8073 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
8075 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8077 PR tree-optimization/85694
8078 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
8079 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
8081 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
8082 (<u>avg<mode>3_ceil): New patterns.
8084 2018-07-03 David Malcolm <dmalcolm@redhat.com>
8086 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
8087 scan-tree-dump directive.
8088 * gcc.dg/vect/slp-perm-2.c: Likewise.
8089 * gcc.dg/vect/slp-perm-3.c: Likewise.
8090 * gcc.dg/vect/slp-perm-5.c: Likewise.
8091 * gcc.dg/vect/slp-perm-6.c: Likewise.
8092 * gcc.dg/vect/slp-perm-7.c: Likewise.
8093 * gcc.dg/vect/slp-perm-8.c: Likewise.
8095 2018-07-03 Marek Polacek <polacek@redhat.com>
8098 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
8100 2018-07-03 Richard Biener <rguenther@suse.de>
8103 * tree-ssa-structalias.c (find_func_clobbers): Properly
8104 handle indirect calls.
8106 2018-07-03 Jeff Law <law@redhat.com>
8108 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
8109 (shifts): New code iterator.
8110 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
8111 expander. Fix HImode handling on H8/SX.
8112 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
8113 (subqi3, subhi3, subsi3 expanders): Likewise.
8114 (andqi3, andhi3, andsi3 expanders): Likewise.
8115 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
8116 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
8117 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
8118 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
8119 (zero_extendqihi2, zero_extendqisi2): Likewise.
8120 (extendqihi2, extendqisi2): Likewise.
8121 (rotlqi3, rotlhi3, rotlsi3): Likewise.
8122 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
8123 (rotlqi3_1, rotlhi3_1): Likewise.
8124 (logicalhi3_sn, logicalsi3_sn): Likewise.
8125 (logicalhi3, logicalsi3): Likewise.
8127 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8129 * tree-vect-patterns.c (vect_recog_rotate_pattern)
8130 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8131 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
8132 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
8133 type to append_pattern_def_seq instead of creating a stmt_vec_info
8135 (build_mask_conversion): Likewise. Remove vinfo argument.
8136 (vect_add_conversion_to_patterm): Likewise, renaming to...
8137 (vect_add_conversion_to_pattern): ...this.
8138 (vect_recog_mask_conversion_pattern): Update call to
8139 build_mask_conversion. Pass the vector type to
8140 append_pattern_def_seq here too.
8141 (vect_recog_gather_scatter_pattern): Update call to
8142 vect_add_conversion_to_pattern.
8144 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8146 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
8147 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8148 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
8149 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
8150 STMT_VINFO_PATTERN_DEF_SEQ to null here.
8151 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
8152 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
8153 append_pattern_def_seq instead of new_pattern_def_seq.
8154 (vect_recog_divmod_pattern): Do both of the above.
8155 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
8158 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8160 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
8161 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
8162 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
8163 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
8164 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
8165 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8166 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
8167 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8168 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8169 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
8170 parameter with a single stmt_vec_info.
8171 (vect_recog_func_ptr): Likewise.
8172 (vect_recog_gather_scatter_pattern): Likewise, folding in...
8173 (vect_try_gather_scatter_pattern): ...this.
8174 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
8175 the stmt_vec_info of the statement to be matched. Don't clear
8176 STMT_VINFO_RELATED_STMT.
8177 (vect_pattern_recog): Update call accordingly.
8179 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8181 PR tree-optimization/85694
8182 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
8183 (uavgM3_ceil): Document new optabs.
8184 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
8185 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
8187 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
8188 (savg_ceil_optab): New optabs.
8189 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
8190 (vect_vect_recog_func_ptrs): Add it.
8191 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
8192 constant directly from the associated lhs.
8194 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8196 * tree-vect-patterns.c (vect_split_statement): New function.
8197 (vect_convert_input): Use it to try to split an existing cast.
8199 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8201 * poly-int.h (print_hex): New function.
8202 * dumpfile.h (dump_dec, dump_hex): Declare.
8203 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
8204 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
8205 min_input_precision, operation_precision and operation_sign.
8206 * tree-vect-patterns.c (vect_get_range_info): New function.
8207 (vect_same_loop_or_bb_p, vect_single_imm_use)
8208 (vect_operation_fits_smaller_type): Delete.
8209 (vect_look_through_possible_promotion): Add an optional
8210 single_use_p parameter.
8211 (vect_recog_over_widening_pattern): Rewrite to use new
8212 stmt_vec_info infomration. Handle one operation at a time.
8213 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
8214 (vect_truncatable_operation_p, vect_set_operation_type)
8215 (vect_set_min_input_precision): New functions.
8216 (vect_determine_min_output_precision_1): Likewise.
8217 (vect_determine_min_output_precision): Likewise.
8218 (vect_determine_precisions_from_range): Likewise.
8219 (vect_determine_precisions_from_users): Likewise.
8220 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
8221 (vect_vect_recog_func_ptrs): Put over_widening first.
8223 (vect_pattern_recog): Call vect_determine_precisions.
8225 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8227 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
8228 statements that have been replaced by further pattern statements.
8229 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
8231 2018-07-03 Richard Biener <rguenther@suse.de>
8233 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
8234 always set *dt. Dump vectype in vectype overload.
8235 * dumpfile.h (dump_gimple_expr): New function.
8236 (dump_gimple_expr_loc): Likewise.
8237 * dumpfile.c (dump_gimple_expr): New function.
8238 (dump_gimple_expr_loc): Likewise.
8240 2018-07-02 Jeff Law <law@redhat.com>
8242 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
8243 the H8/300, H8/300H and H8/S variants into a single pattern.
8244 (movhi_h8300, movqi_h8300hs): Similarly.
8245 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
8246 (QHI mode iterator): New.
8248 * config/h8300/h8300.md: Remove trailing whitespace.
8250 2018-07-02 Jim Wilson <jimw@sifive.com>
8252 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
8253 instead of emit_insn for interrupt returns.
8254 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
8255 (riscv_sret, riscv_uret): Likewise.
8257 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8259 * pretty-print.c (selftest::test_pp_format): Move save and restore
8260 of quotes to class auto_fix_quotes, and add an instance.
8261 * selftest.c: Include "intl.h".
8262 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
8263 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
8264 * selftest.h (selftest::auto_fix_quotes): New class.
8266 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
8268 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
8269 (aarch64_sve_prepare_conditional_op): Remove.
8270 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
8271 Allow aarch64_simd_reg_or_zero as select operand; remove
8272 the aarch64_sve_prepare_conditional_op call.
8273 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
8274 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8275 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
8276 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
8277 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
8278 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
8279 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
8280 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
8281 and a splitters to match all of the *_any patterns.
8282 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
8284 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
8285 (SVE_COND_FP_BINARY_REV): Remove.
8286 (sve_int_op_rev, sve_fp_op_rev): New.
8287 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
8288 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
8289 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
8290 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
8291 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
8292 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
8293 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
8294 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
8295 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
8297 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
8298 Remove match_dup 1 from the inner unspec.
8299 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8301 * config/aarch64/aarch64.md (movprfx): New attr.
8302 (length): Default movprfx to 8.
8303 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
8304 (*madd<SVE_I>, *msub<SVE_I): Likewise.
8305 (*<su>mul<SVE_I>3_highpart): Likewise.
8306 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
8307 (*v<ASHIFT><SVE_I>3): Likewise.
8308 (*<su><MAXMIN><SVE_I>3): Likewise.
8309 (*<su><MAXMIN><SVE_F>3): Likewise.
8310 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
8311 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
8312 (*div<SVE_F>4): Likewise.
8314 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
8316 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
8319 2018-07-02 Richard Biener <rguenther@suse.de>
8321 PR tree-optimization/86363
8322 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
8323 memset argument refers to a non-variable address.
8325 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8327 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
8328 {PLUS,MINUS}_EXPR code to...
8329 (adjust_symbolic_bound): ...here,
8330 (combine_bound): ...here,
8331 (set_value_range_with_overflow): ...and here.
8333 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8335 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
8337 (extract_range_from_abs_expr): ...here.
8339 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
8341 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
8342 -fno-omit-frame-pointer when not optimizing.
8344 2018-07-02 Martin Liska <mliska@suse.cz>
8347 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
8348 (propagate_nothrow): Likewise.
8350 2018-07-02 Martin Liska <mliska@suse.cz>
8353 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
8355 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8357 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
8358 function in r262149, changing "loc" param from source_location to
8359 const dump_location_t &.
8360 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
8361 declaration, as above.
8363 2018-07-01 Paul Koning <ni1d@arrl.net>
8365 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8366 -munit-asm, -mgnu-asm, -mdec-asm.
8367 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
8368 (pdp11_output_labelref): New.
8369 (pdp11_output_def): New.
8370 (pdp11_output_addr_vec_elt): New.
8371 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
8372 %# and %@ format codes.
8373 (pdp11_option_override): New.
8374 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
8375 (pdp11_output_ident): New.
8376 (pdp11_asm_named_section): New.
8377 (pdp11_asm_init_sections): New.
8378 (pdp11_file_start): New.
8379 (pdp11_file_end): New.
8380 (output_ascii): Use .ascii/.asciz for -mdec-asm.
8381 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
8382 %o, like %c but octal.
8383 (pdp11_option_override): New.
8384 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
8386 (DATA_SECTION_ASM_OP): Ditto.
8387 (READONLY_DATA_SECTION_ASM_OP): New.
8388 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
8389 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
8390 (ASM_OUTPUT_LABELREF): Ditto.
8391 (ASM_OUTPUT_DEF): Ditto.
8392 (ASM_OUTPUT_EXTERNAL): New.
8393 (ASM_OUTPUT_SOURCE_FILENAME): New.
8394 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
8395 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
8396 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
8397 %# and %@ format codes.
8398 * config/pdp11/pdp11.opt (mgnu-asm): New.
8399 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
8400 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
8401 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
8403 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
8405 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
8406 dereferencing path[] beyond its length.
8408 (debug_all_paths): New.
8409 (rewire_first_differing_edge): New.
8410 (adjust_paths_after_duplication): New.
8411 (duplicate_thread_path): Call adjust_paths_after_duplication.
8413 (thread_through_all_blocks): Add new argument to
8414 duplicate_thread_path.
8416 2018-06-30 Jim Wilson <jimw@sifive.com>
8418 * config/riscv/predicates.md (p2m1_shift_operand): New.
8419 (high_mask_shift_operand): New.
8420 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
8421 pattern using p2m1_shift_operand.
8422 (lshsi3_zero_extend_3+2): New combiner pattern using
8423 high_mask_shift_operand.
8425 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8427 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
8429 (vect_recog_rotate_pattern): ...here.
8430 (vect_convert_input): Try to insert casts of invariants in the
8432 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
8433 preheader to be empty.
8435 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8437 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
8438 vector type. If given, install it in the new statement's
8440 (vect_element_precision): New function.
8441 (vect_unpromoted_value): New struct.
8442 (vect_unpromoted_value::vect_unpromoted_value): New function.
8443 (vect_unpromoted_value::set_op): Likewise.
8444 (vect_look_through_possible_promotion): Likewise.
8445 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
8446 (vect_widened_op_tree, vect_convert_input): Likewise.
8447 (vect_convert_inputs, vect_convert_output): Likewise.
8448 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
8449 to handle the optional cast of the multiplication result and
8450 vect_widened_op_tree to detect the widened multiplication itself.
8451 Do not require the input and output of promotion casts to have
8452 the same sign, but base the signedness of the operation on the
8453 input rather than the result. If the pattern includes two
8454 promotions, check that those promotions have the same sign.
8455 Do not restrict the MULT_EXPR handling to a double-width result;
8456 handle quadruple-width results and wider. Use vect_convert_inputs
8457 to convert the inputs to the common type.
8458 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
8459 to handle the optional cast of the ABS result. Also allow a sign
8460 change or a sign extension between the ABS and MINUS.
8461 Use vect_widened_op_tree to detect the widened subtraction and use
8462 vect_convert_inputs to convert the inputs to the common type.
8463 (vect_handle_widen_op_by_const): Delete.
8464 (vect_recog_widen_op_pattern): New function.
8465 (vect_recog_widen_mult_pattern): Use it.
8466 (vect_recog_widen_shift_pattern): Likewise.
8467 (vect_recog_widen_sum_pattern): Use
8468 vect_look_through_possible_promotion to handle the promoted
8471 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8473 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
8474 the containing gimple_seq *.
8475 * gimple-iterator.h (gsi_for_stmt): Declare it.
8476 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
8477 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
8478 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8479 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8480 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
8482 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
8484 (vect_mark_pattern_stmts): ...here. Handle cases in which the
8485 statement being replaced is part of an existing pattern
8486 definition sequence, inserting the new pattern statements before
8488 (vect_pattern_recog_1): Don't return a bool. If the statement
8489 is already part of a pattern, instead apply pattern matching
8490 to the pattern definition statements. Don't clear the
8491 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
8492 (vect_pattern_recog): Don't break after the first match;
8493 continue processing the pattern definition statements instead.
8494 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
8496 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8498 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8499 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8500 (vect_recog_widen_sum_pattern): Use it.
8502 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8504 * tree-vect-loop.c (vectorizable_reduction): Assert that the
8505 phi is not a pattern statement and has not been replaced by
8506 a pattern statement.
8507 * tree-vect-patterns.c (type_conversion_p): Don't check
8508 STMT_VINFO_IN_PATTERN_P.
8509 (vect_recog_vector_vector_shift_pattern): Likewise.
8510 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
8511 the pattern statement rather than the original statement; check
8512 directly for a WIDEN_MULT_EXPR here.
8513 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
8514 vect_is_simple_use to return the pattern statement rather
8515 than the original statement; use is_pattern_stmt_p to check
8516 for such a pattern statement.
8517 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
8518 to return the pattern statement rather than the original statement;
8519 don't do the same transformation here.
8520 (vect_is_simple_use): If the defining statement has been replaced
8521 by a pattern statement, return the pattern statement instead.
8522 Remove the corresponding (local) transformation from the vectype
8525 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8527 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
8528 end and default to null.
8529 * tree-vect-loop.c (vect_create_epilog_for_reduction)
8530 (vectorizable_reduction): Update calls accordingly, dropping the
8531 gimple ** argument if the passed-back statement isn't needed.
8532 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
8533 (vect_recog_rotate_pattern): Likewise.
8534 (vect_recog_mask_conversion_pattern): Likewise.
8535 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8536 (vect_mask_constant_operand_p): Likewise.
8537 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
8538 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
8539 (get_group_load_store_type, get_load_store_type): Likewise.
8540 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
8541 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
8542 (vectorizable_conversion, vectorizable_assignment): Likewise.
8543 (vectorizable_shift, vectorizable_operation): Likewise.
8544 (vectorizable_store, vect_is_simple_cond): Likewise.
8545 (vectorizable_condition, vectorizable_comparison): Likewise.
8546 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
8547 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
8548 and move it to the end. Cope with null def_stmt_outs.
8550 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8552 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
8554 2018-06-29 Jeff Law <law@redhat.com>
8556 * config/v850/v850.c (v850_legitimate_address_p): Handle large
8557 displacements for TARGET_V850E2V3 and newer.
8558 (TARGET_LRA_P): Remove. Defaults to LRA now.
8559 * config/v850/v850.md (sign23byte_load): Remove.
8560 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
8561 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
8563 2018-06-29 Martin Liska <mliska@suse.cz>
8566 * coverage.c (coverage_init): Mangle full path name.
8567 * doc/invoke.texi: Document the change.
8568 * gcov-io.c (mangle_path): New.
8569 * gcov-io.h (mangle_path): Likewise.
8570 * gcov.c (mangle_name): Use mangle_path for path mangling.
8572 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8574 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
8575 if starting source register is not even.
8577 2018-06-29 Martin Liska <mliska@suse.cz>
8579 PR tree-optimization/86263
8580 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
8581 Make edge redirection.
8583 2018-06-29 David Malcolm <dmalcolm@redhat.com>
8585 * dumpfile.c (dump_loc): Add indentation based on scope depth.
8586 (dump_scope_depth): New variable.
8587 (get_dump_scope_depth): New function.
8588 (dump_begin_scope): New function.
8589 (dump_end_scope): New function.
8590 * dumpfile.h (get_dump_scope_depth): New declaration.
8591 (dump_begin_scope): New declaration.
8592 (dump_end_scope): New declaration.
8593 (class auto_dump_scope): New class.
8594 (AUTO_DUMP_SCOPE): New macro.
8595 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
8598 2018-06-29 Richard Biener <rguenther@suse.de>
8600 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
8601 compute_all_dependences succeeds.
8602 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
8603 exceed --param loop-max-datarefs-for-datadeps.
8605 2018-06-29 Jakub Jelinek <jakub@redhat.com>
8607 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
8609 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
8612 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
8613 alternative 0 in preferred_for_speed attribute.
8615 2018-06-28 Paul Koning <ni1d@arrl.net>
8617 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
8618 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
8619 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
8620 * config/pdp11/pdp11.md: Correct "length" attribute calculation
8621 for shift insn patterns.
8623 2018-06-28 David Malcolm <dmalcolm@redhat.com>
8625 * cgraph.c (cgraph_node::get_body): Replace assignments to
8626 "dump_file" with calls to set_dump_file.
8627 * dumpfile.c (alt_dump_file): Make static, and group with...
8628 (alt_flags): ...this definition.
8629 (dumps_are_enabled): New variable.
8630 (refresh_dumps_are_enabled): New function.
8631 (set_dump_file): New function.
8632 (set_alt_dump_file): New function.
8633 (gcc::dump_manager::dump_start): Replace assignments to
8634 "dump_file" and "alt_dump_file" with calls to set_dump_file and
8636 (gcc::dump_manager::dump_finish): Likewise.
8637 * dumpfile.h (alt_dump_file): Delete decl.
8638 (dumps_are_enabled): New variable decl.
8639 (set_dump_file): New function decl.
8640 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
8642 * tree-nested.c (lower_nested_functions): Replace assignments to
8643 "dump_file" with calls to set_dump_file.
8645 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
8647 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
8648 goto_locus of each outgoing edge of each basic block.
8650 2018-06-28 Richard Biener <rguenther@suse.de>
8652 * dwarf2out.c (decl_scope_table): Remove.
8653 (push_decl_scope): Likewise.
8654 (pop_decl_scope): Likewise.
8655 (gen_type_die_for_member): Do not call push/pop_decl_scope.
8656 (gen_struct_or_union_type_die): Likewise.
8657 (gen_tagged_type_die): Likewise.
8658 (dwarf2out_init): Do not initialize decl_scope_table.
8659 (dwarf2out_c_finalize): Do not free it.
8661 2018-06-28 Richard Biener <rguenther@suse.de>
8663 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
8664 deciding whether to not re-use a DIE.
8666 2018-06-28 Richard Biener <rguenther@suse.de>
8668 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
8669 DW_AT_abstract_origin attribute.
8671 2018-06-28 Martin Liska <mliska@suse.cz>
8673 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8674 Use newly introduced constants.
8675 * tree-switch-conversion.h (struct jump_table_cluster):
8676 Define max_ratio_for_size and max_ratio_for_speed.
8678 2018-06-28 Martin Liska <mliska@suse.cz>
8680 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8681 Add new checking assert to catch invalid state.
8682 (jump_table_cluster::can_be_handled): Handle single case
8684 (jump_table_cluster::is_beneficial): Bail out for such case.
8685 (bit_test_cluster::find_bit_tests):
8686 Add new checking assert to catch invalid state.
8687 (bit_test_cluster::can_be_handled): Handle single case
8689 (bit_test_cluster::is_beneficial): Bail out for such case.
8690 (switch_decision_tree::analyze_switch_statement):
8693 2018-06-28 Martin Liska <mliska@suse.cz>
8695 * common.opt: Introduce -completion option.
8696 * gcc.c (driver_handle_option): Handle it.
8697 (driver::main): Print completions if completion
8699 * opt-suggestions.c (option_proposer::get_completions):
8701 (option_proposer::suggest_completion): Likewise.
8702 (option_proposer::find_param_completions): Likewise.
8703 (verify_autocompletions): Likewise.
8704 (test_completion_valid_options): Likewise.
8705 (test_completion_valid_params): Likewise.
8706 (in_completion_p): Likewise.
8707 (empty_completion_p): Likewise.
8708 (test_completion_partial_match): Likewise.
8709 (test_completion_garbage): Likewise.
8710 (opt_proposer_c_tests): Likewise.
8711 * opt-suggestions.h: Declare new functions.
8712 * opts.c (common_handle_option): Handle OPT__completion_.
8713 * selftest-run-tests.c (selftest::run_tests): Add
8714 opt_proposer_c_tests.
8715 * selftest.c (assert_str_startswith): New.
8716 * selftest.h (assert_str_startswith): Likewise.
8717 (opt_proposer_c_tests): New.
8718 (ASSERT_STR_STARTSWITH): Likewise.
8720 2018-06-28 Martin Liska <mliska@suse.cz>
8722 * Makefile.in: Add opt-suggestions.o.
8723 * gcc-main.c: Include opt-suggestions.h.
8724 * gcc.c (driver::driver): Likewise.
8725 (driver::~driver): Remove m_option_suggestions.
8726 (driver::build_option_suggestions): Moved to option_proposer.
8727 (driver::suggest_option): Likewise.
8728 (driver::handle_unrecognized_options): Use option_proposer.
8729 * gcc.h (class driver): Add new memver m_option_proposer.
8730 * opt-suggestions.c: New file.
8731 * opt-suggestions.h: New file.
8733 2018-06-28 Martin Liska <mliska@suse.cz>
8735 * vec.h (class auto_string_vec): New (moved from auto_argvec).
8736 (auto_string_vec::~auto_string_vec): Likewise.
8738 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
8740 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
8741 prevent_decl_creation_for_types fields up and add reset_location field.
8742 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
8743 statement if id->reset_location is true.
8744 (copy_edges_for_bb): Do not set goto_locus on the new edges if
8745 id->reset_location is true.
8746 (copy_phis_for_bb): Force input_location on the arguments if
8747 id->reset_location is true.
8748 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
8749 is set on the function to be inlined.
8751 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
8753 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
8755 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
8757 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
8758 registers for Pmode.
8759 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
8760 hard registers for the clobbered pseudo.
8762 2018-06-27 Paul Koning <ni1d@arrl.net>
8764 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8765 mutually exclusive options.
8766 * config/pdp11/constraints.md (h): New constraint.
8767 (O): Update definition to match shift code generation.
8768 (D): New constraint.
8769 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8771 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8773 (output_jump): Change arguments.
8774 (pdp11_fixed_cc_regs): New function.
8775 (pdp11_cc_mode): Ditto.
8776 (pdp11_expand_shift): Ditto.
8777 (pdp11_assemble_shift): Ditto.
8778 (pdp11_small_shift): Ditto.
8779 (pdp11_branch_cost): Remove.
8780 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8782 (pdp11_register_move_cost): Update for CC registers.
8783 (pdp11_rtx_costs): Add case for LSHIFTRT.
8784 (pdp11_output_jump): Add CCNZ mode conditional branches.
8785 (notice_update_cc_on_set): Remove.
8786 (pdp11_cc_mode): New function.
8787 (simple_memory_operand): Correct pre/post decrement case.
8788 (no_side_effect_operand): New function.
8789 (pdp11_regno_reg_class): Add CC_REGS class.
8790 (pdp11_fixed_cc_regs): New function.
8791 (pdp11_small_shift): New function.
8792 (pdp11_expand_shift): New function to expand shift insns.
8793 (pdp11_assemble_shift): New function to output shifts.
8794 (pdp11_branch_cost): Remove.
8795 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
8796 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
8797 (WCHAR_TYPE): Ditto.
8798 (PTRDIFF_TYPE): Ditto.
8799 (ADJUST_INSN_LENGTH): New macro.
8800 (FIXED_REGISTERS): Add CC registers.
8801 (CALL_USED_REGISTERS): Ditto.
8803 (REG_CLASS_NAMES): Ditto.
8804 (REG_CLASS_CONTENTS): Ditto.
8805 (SELECT_CC_MODE): Use new function.
8806 (TARGET_FLAGS_REGNUM): New macro.
8807 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
8808 (cc0_reg_rtx): Remove.
8809 (CC_STATUS_MDEP): Remove.
8810 (CC_STATUS_MDEFP_INIT): Remove.
8811 (CC_IN_FPU): Remove.
8812 (NOTICE_UPDATE_CC): Remove.
8813 (REGISTER_NAMES): Add CC registers.
8814 (BRANCH_COST): Change to constant 1.
8815 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
8817 * config/pdp11/pdp11.opt (mbcopy): Remove.
8818 (mbcopy-builtin): Remove.
8819 (mbranch-cheap): Remove.
8820 (mbranch-expensive): Remove.
8821 * config/pdp11/predicates.md (expand_shift_operand): Update to
8822 match shift code generation.
8823 (ccnz_operator): New predicate.
8824 * doc/invoke.texi (PDP-11 Options): Remove deleted options
8825 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
8826 Remove non-existent option -mabshi, -mno-abshi. Document mutually
8828 * doc/md.texi (PDP-11): Document new D and h constraints. Update
8829 description of O constraint.
8831 2018-06-27 Jeff Law <law@redhat.com>
8832 Austin Law <austinklaw@gmail.com>
8834 * config/v850/v850.md (addsi3_set_flags): New pattern.
8835 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
8836 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
8837 (zero_extendhisi2_v850_set_flags): Likewise.
8838 (zero_extendqisi2_v850_set_flags): Likewise.
8839 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
8840 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
8841 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
8843 * config/v850/v850-protos.h (notice_update_cc): Remove.
8844 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
8845 (v850_print_operand): Handle 'D' and "d".
8846 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
8847 Add handling of arithmetic/logical operations compared against zero.
8848 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
8849 Do not look at v850_compare_op, instead get mode from last argument.
8850 (v850_gen_compare): Remove
8851 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
8852 after reload for prologue insns.
8853 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
8855 (construct_save_jarl): Likewise.
8856 (TARGET_FLAGS_REGNUM): Define.
8857 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
8858 (NOTICE_UPDATE_CC): Remove.
8859 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
8860 than cc0. Conditionalize on reload_completed.
8861 (cmpsi_insn, setfcc_insn): Likewise.
8862 (tst1 splitter): Turn into define_and_split which sets the flags
8864 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
8865 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
8866 (cstoresf4, cstoredf4): Clobber the flags.
8867 (cmpsi, cmpsf, cmpdf): Remove expanders.
8868 (setf_insn): Remove pattern.
8869 (addsi3): Turn into define_and_split which clobbers the flags after
8870 reload and a suitable pattern (addsi3_clobber_flags) for use after
8872 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
8873 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
8874 (ashrsi3, ashrsi3_v850e2): Likewise.
8875 (bins): Clobber the flags.
8876 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
8877 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
8878 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
8879 (call_value_internal_short, call_value_internal_long): Likewise.
8880 (callt_save_interrupt, callt_return_interrupt): Likewise.
8881 (save_interrupt, return_interrupt): Likewise.
8882 (callt_save_all_interrupt, save_all_interrupt): Likewise.
8883 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
8884 (restore_all_interrupt, _restore_all_interrupt): Likewise.
8885 (All FP comparisons): Only allow after reload has completed.
8887 (divh, divhu): Tweak output template.
8888 (branch_z_normal, branch_z_invert): Remove
8889 (branch_nz_normal, branch_nz_invert): Likewise.
8890 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
8892 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
8893 * config/v850/v850.c (notice_update_cc): Remove.
8894 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
8895 (CC_NO_CARRY): Likewise.
8896 (NOTICE_UPDATE_CC): Define to nothing.
8897 * config/v850/v850.md: Remove block comment on cc0 handling
8898 Remove "cc" attribute from all patterns. Remove cc_status handling
8899 from all patterns. Minor formatting fixes.
8901 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8903 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
8904 (cortex-a76.cortex-a55): Likewise.
8905 * config/aarch64/aarch64-tune.md: Regenerate.
8906 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
8907 cortex-a76.cortex-a55.
8909 2018-06-27 Jeff Law <law@redhat.com>
8911 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
8912 (MULTILIB_DIRNAMES): Similarly.
8914 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
8916 * gimple.h (gimple_return_retbnd): Delete.
8917 (gimple_return_set_retbnd): Likewise.
8918 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
8919 gimple_return_set_retbnd.
8920 * gimple-pretty-print.c (dump_gimple_return): Remove call to
8921 gimple_return_retbnd and adjust.
8922 * tree-inline.h (struct copy_body_data): Remove retbnd field.
8923 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
8924 Explicitly return NULL in a couple more cases. Move assertion
8925 on debug statements and remove unreachable code.
8926 (reset_debug_binding): Do not test id->retbnd.
8927 (expand_call_inline): Do not set it.
8929 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8931 * configure.ac: Add --disable-gcov option.
8932 * configure: Regenerate.
8933 * Makefile.in: Honour @enable_gcov@.
8934 * doc/install.texi: Document --disable-gcov.
8936 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8938 * config/arm/arm-cpus.in (cortex-a76): New entry.
8939 (cortex-a76.cortex-a55): Likewise.
8940 * config/arm/arm-tables.opt: Regenerate.
8941 * config/arm/arm-tune.md: Likewise.
8942 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
8943 * doc/invoke.texi (ARM Options): Document cortex-a76 and
8944 cortex-a76.cortex-a55.
8946 2018-06-27 Tamar Christina <tamar.christina@arm.com>
8949 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
8951 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
8953 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
8955 (EPILOGUE_USES): Likewise.
8957 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
8959 * tree-inline.c (remap_location): New function extracted from...
8960 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
8961 (copy_phis_for_bb): ...here. Call remap_location.
8962 (copy_cfg_body): Adjust call to copy_edges_for_bb.
8964 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
8966 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
8967 unaligned vsx for 16B memset.
8969 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
8972 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
8973 ieee128_float_type_node to long_double_type_node unless
8974 TARGET_LONG_DOUBLE_128 is set.
8976 2018-06-26 David Malcolm <dmalcolm@redhat.com>
8978 * cfgloop.c (get_loop_location): Convert return type from
8979 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
8980 by implicit construction from rtx_insn *, and using
8981 dump_user_location_t::from_function_decl for the fallback case.
8982 * cfgloop.h (get_loop_location): Convert return type from
8983 location_t to dump_user_location_t.
8984 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
8985 dump_printf_loc to pass in a dump_location_t rather than a
8986 location_t, via the gimple stmt.
8987 * coverage.c (get_coverage_counts): Update calls to
8988 dump_printf_loc to pass in dump_location_t rather than a
8990 * doc/optinfo.texi (Dump types): Convert example of
8991 dump_printf_loc from taking "locus" to taking "insn". Update
8992 description of the "_loc" calls to cover dump_location_t.
8993 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
8995 (dump_user_location_t::dump_user_location_t): New constructors,
8996 from gimple *stmt and rtx_insn *.
8997 (dump_user_location_t::from_function_decl): New function.
8998 (dump_loc): Make static.
8999 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
9000 const dump_location_t &.
9001 (dump_generic_expr_loc): Delete.
9002 (dump_printf_loc): Convert param "loc" from location_t to
9003 const dump_location_t &.
9004 (selftest::test_impl_location): New function.
9005 (selftest::dumpfile_c_tests): New function.
9006 * dumpfile.h: Include "profile-count.h".
9007 (class dump_user_location_t): New class.
9008 (struct dump_impl_location_t): New struct.
9009 (class dump_location_t): New class.
9010 (dump_printf_loc): Convert 2nd param from source_location to
9011 const dump_location_t &.
9012 (dump_generic_expr_loc): Delete.
9013 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
9014 const dump_location_t &.
9015 * gimple-fold.c (fold_gimple_assign): Update call to
9016 dump_printf_loc to pass in a dump_location_t rather than a
9017 location_t, via the gimple stmt.
9018 (gimple_fold_call): Likewise.
9019 * gimple-loop-interchange.cc
9020 (loop_cand::analyze_iloop_reduction_var): Update for change to
9021 check_reduction_path.
9022 (tree_loop_interchange::interchange): Update for change to
9024 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
9025 change in return-type of find_loop_location.
9026 (graphite_regenerate_ast_isl): Likewise.
9027 * graphite-optimize-isl.c (optimize_isl): Likewise.
9028 * graphite.c (graphite_transform_loops): Likewise.
9029 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
9030 pass in a dump_location_t rather than a location_t, via the
9032 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9033 * ipa.c (walk_polymorphic_call_targets): Likewise.
9034 * loop-unroll.c (report_unroll): Convert "locus" param from
9035 location_t to dump_location_t.
9036 (decide_unrolling): Update for change to get_loop_location's
9038 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
9039 location_t to dump_user_location_t.
9040 (grid_find_single_omp_among_assignments_1): Updates calls to
9041 dump_printf_loc to pass in a dump_location_t rather than a
9042 location_t, via the gimple stmt.
9043 (grid_parallel_clauses_gridifiable): Convert "tloc" from
9044 location_t to dump_location_t. Updates calls to dump_printf_loc
9045 to pass in a dump_location_t rather than a location_t, via the
9047 (grid_inner_loop_gridifiable_p): Likewise.
9048 (grid_dist_follows_simple_pattern): Likewise.
9049 (grid_gfor_follows_tiling_pattern): Likewise.
9050 (grid_target_follows_gridifiable_pattern): Likewise.
9051 (grid_attempt_target_gridification): Convert initialization
9052 of local "grid" from memset to zero-initialization; FIXME: does
9053 this require C++11? Update call to dump_printf_loc to pass in a
9054 optinfo_location rather than a location_t, via the gimple stmt.
9055 * profile.c (read_profile_edge_counts): Updates call to
9056 dump_printf_loc to pass in a dump_location_t rather than a
9058 (compute_branch_probabilities): Likewise.
9059 * selftest-run-tests.c (selftest::run_tests): Call
9061 * selftest.h (dumpfile_c_tests): New decl.
9062 * tree-loop-distribution.c (pass_loop_distribution::execute):
9063 Update for change in return type of find_loop_location.
9064 * tree-parloops.c (parallelize_loops): Likewise.
9065 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
9066 "locus" from location_t to dump_user_location_t.
9067 (canonicalize_loop_induction_variables): Likewise.
9068 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
9069 for change in return type of find_loop_location.
9070 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
9071 to dump_printf_loc to pass in a dump_location_t rather than a
9072 location_t, via the stmt.
9073 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
9075 * tree-vect-loop-manip.c (find_loop_location): Convert return
9076 type from source_location to dump_user_location_t.
9077 (vect_do_peeling): Update for above change.
9078 (vect_loop_versioning): Update for change in type of
9080 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
9081 from location_t to dump_user_location_t.
9082 (vect_estimate_min_profitable_iters): Update for change in type
9084 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
9085 location_t to dump_location_t.
9086 (vect_slp_bb): Update for change in type of vect_location.
9087 * tree-vectorizer.c (vect_location): Convert from source_location
9088 to dump_user_location_t.
9089 (try_vectorize_loop_1): Update for change in vect_location's type.
9090 (vectorize_loops): Likewise.
9091 (increase_alignment): Likewise.
9092 * tree-vectorizer.h (vect_location): Convert from source_location
9093 to dump_user_location_t.
9094 (find_loop_location): Convert return type from source_location to
9095 dump_user_location_t.
9096 (check_reduction_path): Convert 1st param from location_t to
9097 dump_user_location_t.
9098 * value-prof.c (check_counter): Update call to dump_printf_loc to
9099 pass in a dump_user_location_t rather than a location_t; update
9100 call to error_at for change in type of "locus".
9101 (check_ic_target): Update call to dump_printf_loc to
9102 pass in a dump_user_location_t rather than a location_t, via the
9105 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
9107 * config/s390/s390.h (enum processor_flags): Do not use
9108 default tune parameter when -march was specified.
9110 2018-06-26 Jakub Jelinek <jakub@redhat.com>
9113 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
9114 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
9117 2018-06-26 Richard Biener <rguenther@suse.de>
9119 PR tree-optimization/86287
9121 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
9122 (vect_analyze_loop): Initialize n_stmts.
9124 2018-06-26 Richard Biener <rguenther@suse.de>
9127 * fold-const.c (fold_convertible_p): Pointer extension
9130 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
9133 * dwarf2out.c (loc_list_has_views): Adjust comments.
9134 (dw_loc_list): Split single cross-partition range with
9137 2018-06-25 Jeff Law <law@redhat.com>
9139 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
9140 on -mbig-switch by default.
9142 * config/v850/predicates.md (const_float_1_operand): Fix match_code
9144 (const_float_0_operand): Remove unused predicate.
9145 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
9146 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
9147 (recipsf2): New expander. Original pattern now called
9149 (recipdf2, recipdf2_insn): Similarly.
9150 (rsqrtsf2, rsqrtsf2_insn): Similarly
9151 (rsqrtdf2, rsqrtdf2_insn): Similarly
9153 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
9155 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
9156 Simplify logic for FreeBSD (twice).
9158 2018-06-25 Martin Sebor <msebor@redhat.com>
9160 PR tree-optimization/86204
9161 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
9162 a strnlen result if it's less than the length of the string.
9164 2018-06-25 Martin Sebor <msebor@redhat.com>
9166 PR tree-optimization/85700
9167 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
9168 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
9169 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
9171 2018-06-25 Martin Sebor <msebor@redhat.com>
9173 * doc/extend.texi (Zero-length arrays): Update and clarify.
9175 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
9177 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
9178 added IEEE/IBM long double multilib support on PowerPC little
9179 endian Linux systems.
9180 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
9181 (MULTILIB_DEFAULTS): Likewise.
9182 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9184 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
9185 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
9186 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
9188 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
9191 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
9192 (REORDER_45): Likewise.
9194 2018-06-25 Jeff Law <law@redhat.com>
9196 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
9197 dividend to 32 bits. Adjust length.
9198 (udivmodhi4): Cleanup output template. Fix length.
9200 2018-06-25 Carl Love <cel@us.ibm.com>
9202 * config/rs6000/vsx.md: Change word selector to prefered location.
9204 2018-06-25 Richard Biener <rguenther@suse.de>
9206 PR tree-optimization/86304
9207 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
9208 epilogue-if-converted loops as well.
9210 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
9212 * lto-section-out.c (lto_begin_section): Do not print section
9213 name for noaddr and unnumbered dumps.
9215 2018-06-25 Richard Biener <rguenther@suse.de>
9217 * tree-vectorizer.h (struct vec_info_shared): New structure
9218 with parts split out from struct vec_info and loop_nest from
9219 struct _loop_vec_info.
9220 (struct vec_info): Adjust accordingly.
9221 (struct _loop_vec_info): Likewise.
9222 (LOOP_VINFO_LOOP_NEST): Adjust.
9223 (LOOP_VINFO_DATAREFS): Likewise.
9224 (LOOP_VINFO_DDRS): Likewise.
9225 (struct _bb_vec_info): Likewise.
9226 (BB_VINFO_DATAREFS): Likewise.
9227 (BB_VINFO_DDRS): Likewise.
9228 (struct _stmt_vec_info): Add dr_aux member.
9229 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
9230 (DR_MISALIGNMENT_UNINITIALIZED): New.
9231 (set_dr_misalignment): Adjust.
9232 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
9233 (vect_analyze_loop): Adjust prototype.
9234 (vect_analyze_loop_form): Likewise.
9235 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9236 Compute dependences lazily.
9237 (vect_record_base_alignments): Use shared datarefs/ddrs.
9238 (vect_verify_datarefs_alignment): Likewise.
9239 (vect_analyze_data_refs_alignment): Likewise.
9240 (vect_analyze_data_ref_accesses): Likewise.
9241 (vect_analyze_data_refs): Likewise.
9242 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
9243 constructor parameter for shared part.
9244 (vect_analyze_loop_form): Pass in shared part and adjust.
9245 (vect_analyze_loop_2): Pass in storage for the number of
9246 stmts. Move loop nest finding to the caller. Compute
9248 (vect_analyze_loop): Pass in shared part.
9249 (vect_transform_loop): Verify shared datarefs are unchanged.
9250 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
9251 constructor parameter for shared part.
9252 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
9253 (vect_slp_bb): Verify shared datarefs are unchanged before
9255 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
9257 (new_stmt_vec_info): Initialize DR_AUX misalignment to
9258 DR_MISALIGNMENT_UNINITIALIZED.
9259 * tree-vectorizer.c (vec_info::vec_info): Add constructor
9260 parameter for shared part.
9261 (vec_info::~vec_info): Adjust.
9262 (vec_info_shared::vec_info_shared): New.
9263 (vec_info_shared::~vec_info_shared): Likewise.
9264 (vec_info_shared::save_datarefs): Likewise.
9265 (vec_info_shared::check_datarefs): Likewise.
9266 (try_vectorize_loop_1): Construct shared part live for analyses
9267 of a single loop for multiple vector sizes.
9268 * tree-parloops.c (gather_scalar_reductions): Adjust.
9270 2018-06-25 Richard Biener <rguenther@suse.de>
9272 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
9273 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
9274 (vect_analyze_data_refs): Remove similar code from here and
9275 simplify accordingly.
9277 2018-06-25 Richard Biener <rguenther@suse.de>
9279 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
9280 for reverse storage order accesses rather than asserting
9281 they cannot happen here.
9283 2018-06-25 Tom de Vries <tdevries@suse.de>
9286 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
9287 Use data16 instead of .byte for insn prefix.
9289 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
9292 * parser.c (make_char_string_pack): Pass this literal chars
9293 through cpp_interpret_string.
9294 (cp_parser_userdef_numeric_literal): Check the result of
9295 make_char_string_pack.
9297 2018-06-24 Maya Rashish <coypu@sdf.org>
9299 * ginclude/stddef.h: Simplify conditions around avoiding
9300 re-definition of __size_t.
9302 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9304 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
9305 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
9307 2018-06-22 Maya Rashish <coypu@sdf.org>
9309 * doc/invoke.texi (mno-fancy-math-387): Update for changes
9310 made to OpenBSD and NetBSD through the years.
9312 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9314 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9315 behavior of vec_pack (vector double, vector double) to match
9316 behavior of vec_float2 (vector double, vector double).
9318 2018-06-22 Olivier Hainque <hainque@adacore.com>
9320 * gimplify.c (gimplify_function_tree): Prevent creation
9321 of a trampoline for the address of the current function
9322 passed to entry/exit instrumentation hooks.
9324 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
9327 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
9330 2018-06-22 Martin Liska <mliska@suse.cz>
9332 PR tree-optimization/86263
9333 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9334 Bail out if is_enabled is false.
9335 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
9337 (jump_table_cluster::is_enabled): New function.
9339 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9341 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9342 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
9343 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
9344 (lto_input_ts_binfo_tree_pointers): Likewise.
9345 * tree-streamer-out.c (streamer_write_tree_bitfields,
9346 write_ts_binfo_tree_pointers): Likewise.
9347 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
9349 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9351 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
9353 2018-06-22 Martin Liska <mliska@suse.cz>
9355 * symbol-summary.h (get): Make it pure and inline move
9356 functionality from ::get function.
9357 (get): Remove and inline into ::get and ::get_create.
9358 (get_create): Move code from ::get function.
9360 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9363 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
9364 -x assembler-with-cpp.
9366 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9368 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
9369 _FILE_OFFSET_BITS=64 for C++.
9371 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
9373 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
9374 conversion insn that shows up when pr85657-3.c is compiled using
9375 IEEE 128-bit long double.
9376 (neg<mode>2_internal): Use the correct mode to check whether the
9377 mode is IBM extended.
9378 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
9379 multiply and divide external functions from being created more
9382 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9384 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
9386 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
9387 the edge can be forwarded.
9388 (cfg_layout_merge_blocks): Likewise.
9390 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9392 * except.c (finish_eh_generation): Commit edge insertions only after
9393 the EH edges have been redirected from post-landing to landing pads.
9395 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9397 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
9398 create_tmp_var_for to create the FRAME decl.
9399 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
9401 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9403 * tree-inline.c (copy_edges_for_bb): Minor tweak.
9404 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
9405 debug statement when resetting its value.
9406 (expand_call_inline): Copy the locus of the call onto the assignment
9407 of the return value, if any. Use local variable in more cases.
9409 2018-06-21 Martin Liska <mliska@suse.cz>
9411 * ipa-pure-const.c (propagate_nothrow): Use
9412 funct_state_summaries->get.
9413 (dump_malloc_lattice): Likewise.
9414 (propagate_malloc): Likewise.
9416 2018-06-21 Richard Biener <rguenther@suse.de>
9418 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
9419 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
9420 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
9422 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
9423 BLOCK_ABSTRACT_ORIGIN unconditionally.
9425 2018-06-21 David Malcolm <dmalcolm@redhat.com>
9427 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
9429 * ipa-reference.c (ipa_reference_c_finalize): Delete
9430 ipa_ref_opt_sum_summaries and set it to NULL.
9432 2018-06-21 Tom de Vries <tdevries@suse.de>
9434 PR tree-optimization/85859
9435 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
9436 test with comment from bb_no_side_effects_p.
9438 2018-06-21 Richard Biener <rguenther@suse.de>
9440 PR tree-optimization/86232
9441 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
9442 max for constant niter.
9444 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9446 * config/aarch64/aarch64-simd.md
9447 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
9449 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9451 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
9452 Make opernads of the unspec commutative.
9454 2018-06-21 Richard Biener <rguenther@suse.de>
9456 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
9457 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
9458 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
9459 (vect_analyze_data_ref_dependence): Re-order checks to deal with
9461 (vect_record_base_alignments): Do not record base alignment
9462 for gathers or scatters.
9463 (vect_compute_data_ref_alignment): Drop return value that is always
9464 true. Bail out early for gathers or scatters.
9465 (vect_enhance_data_refs_alignment): Bail out early for gathers
9467 (vect_find_same_alignment_drs): Likewise.
9468 (vect_analyze_data_refs_alignment): Remove dead code.
9469 (vect_slp_analyze_and_verify_node_alignment): Likewise.
9470 (vect_analyze_data_refs): For possible gathers or scatters do
9471 not create an alternate DR, just check their possible validity
9472 and mark them. Adjust DECL_NONALIASED handling to not rely
9474 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
9475 update inits of gathers or scatters.
9476 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
9477 Also copy gather/scatter flag to pattern vinfo.
9479 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
9481 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9482 behavior of vec_packsu (vector unsigned long long, vector unsigned
9483 long long) to match behavior of vec_packs with same signature.
9485 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
9486 Thomas Schwinge <thomas@codesourcery.com>
9487 Cesar Philippidis <cesar@codesourcery.com>
9489 * gimplify.c (gimplify_scan_omp_clauses): Add support for
9490 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9491 (gimplify_adjust_omp_clauses): Likewise.
9492 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
9493 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
9494 (gimplify_omp_target_update): Update handling of acc update and
9496 * omp-low.c (install_var_field): Remove unused parameter
9497 base_pointers_restrict.
9498 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
9499 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
9501 (omp_target_base_pointers_restrict_p): Delete.
9502 (scan_omp_target): Update call to scan_sharing_clauses.
9503 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
9505 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
9506 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9507 (convert_local_omp_clauses): Likewise.
9508 * tree-pretty-print.c (dump_omp_clause): Likewise.
9509 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
9511 (omp_clause_code_name): Likewise.
9513 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9516 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
9519 PR tree-optimization/86231
9520 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
9521 anti-range don't overwrite *vr0min before using it to compute *vr0max.
9523 2018-06-20 Tom de Vries <tdevries@suse.de>
9525 PR tree-optimization/86097
9526 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
9527 iv type if signedness of iv type is not the same as that of *nit.
9529 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9531 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
9532 EDGE_EH edges, verify they are all EDGE_EH.
9534 2018-06-20 Maya Rashish <coypu@sdf.org>
9536 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
9538 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9540 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
9541 * config/aarch64/aarch64.c (xgene1_tunings): Add
9542 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
9543 (aarch64_mode_valid_for_sched_fusion_p):
9544 Allow 16-byte modes.
9545 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
9546 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
9548 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
9550 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
9551 * config/aarch64/iterators.md (VQ2): New mode iterator.
9553 2018-06-20 Martin Liska <mliska@suse.cz>
9555 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9556 Change default ratio from 10 to 8.
9558 2018-06-20 Martin Liska <mliska@suse.cz>
9560 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9562 (bit_test_cluster::find_bit_tests): Likewise.
9563 (switch_decision_tree::analyze_switch_statement): Find clusters.
9564 * tree-switch-conversion.h (struct jump_table_cluster): Document
9567 2018-06-20 Martin Liska <mliska@suse.cz>
9569 * tree-switch-conversion.c (switch_conversion::collect):
9570 Record m_uniq property.
9571 (switch_conversion::expand): Bail out for special conditions.
9572 (group_cluster::~group_cluster): New.
9573 (group_cluster::group_cluster): Likewise.
9574 (group_cluster::dump): Likewise.
9575 (jump_table_cluster::emit): New.
9576 (switch_decision_tree::fix_phi_operands_for_edges): New.
9577 (struct case_node): Remove struct.
9578 (jump_table_cluster::can_be_handled): New.
9579 (case_values_threshold): Moved to header.
9580 (reset_out_edges_aux): Likewise.
9581 (jump_table_cluster::is_beneficial): New.
9582 (bit_test_cluster::can_be_handled): Likewise.
9583 (add_case_node): Remove.
9584 (bit_test_cluster::is_beneficial): New.
9585 (case_bit_test::cmp): New.
9586 (bit_test_cluster::emit): New.
9587 (expand_switch_as_decision_tree_p): Remove.
9588 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
9589 (fix_phi_operands_for_edge): Likewise.
9590 (switch_decision_tree::analyze_switch_statement): New.
9591 (compute_cases_per_edge): Move ...
9592 (switch_decision_tree::compute_cases_per_edge): ... here.
9593 (try_switch_expansion): Likewise.
9594 (switch_decision_tree::try_switch_expansion): Likewise.
9595 (record_phi_operand_mapping): Likewise.
9596 (switch_decision_tree::record_phi_operand_mapping): Likewise.
9597 (emit_case_decision_tree): Likewise.
9598 (switch_decision_tree::emit): Likewise.
9599 (balance_case_nodes): Likewise.
9600 (switch_decision_tree::balance_case_nodes): Likewise.
9601 (dump_case_nodes): Likewise.
9602 (switch_decision_tree::dump_case_nodes): Likewise.
9603 (emit_jump): Likewise.
9604 (switch_decision_tree::emit_jump): Likewise.
9605 (emit_cmp_and_jump_insns): Likewise.
9606 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
9607 (emit_case_nodes): Likewise.
9608 (switch_decision_tree::emit_case_nodes): Likewise.
9609 (conditional_probability): Remove.
9610 * tree-switch-conversion.h (enum cluster_type): New.
9612 (struct cluster): Likewise.
9613 (cluster::cluster): Likewise.
9614 (struct simple_cluster): Likewise.
9615 (simple_cluster::simple_cluster): Likewise.
9616 (struct group_cluster): Likewise.
9617 (struct jump_table_cluster): Likewise.
9618 (struct bit_test_cluster): Likewise.
9619 (struct min_cluster_item): Likewise.
9620 (struct case_tree_node): Likewise.
9621 (case_tree_node::case_tree_node): Likewise.
9622 (jump_table_cluster::case_values_threshold): Likewise.
9623 (struct case_bit_test): Likewise.
9624 (struct switch_decision_tree): Likewise.
9625 (struct switch_conversion): Likewise.
9626 (switch_decision_tree::reset_out_edges_aux): Likewise.
9628 2018-06-20 Martin Liska <mliska@suse.cz>
9630 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
9631 (hoist_edge_and_branch_if_true): Likewise.
9632 (expand_switch_using_bit_tests_p): Likewise.
9633 (struct case_bit_test): Likewise.
9634 (case_bit_test_cmp): Likewise.
9635 (emit_case_bit_tests): Likewise.
9636 (switch_conversion::switch_conversion): New class.
9637 (struct switch_conv_info): Remove old struct.
9638 (collect_switch_conv_info): More to ...
9639 (switch_conversion::collect): ... this.
9640 (check_range): Likewise.
9641 (switch_conversion::check_range): Likewise.
9642 (check_all_empty_except_final): Likewise.
9643 (switch_conversion::check_all_empty_except_final): Likewise.
9644 (check_final_bb): Likewise.
9645 (switch_conversion::check_final_bb): Likewise.
9646 (create_temp_arrays): Likewise.
9647 (switch_conversion::create_temp_arrays): Likewise.
9648 (free_temp_arrays): Likewise.
9649 (gather_default_values): Likewise.
9650 (switch_conversion::gather_default_values): Likewise.
9651 (build_constructors): Likewise.
9652 (switch_conversion::build_constructors): Likewise.
9653 (constructor_contains_same_values_p): Likewise.
9654 (switch_conversion::contains_same_values_p): Likewise.
9655 (array_value_type): Likewise.
9656 (switch_conversion::array_value_type): Likewise.
9657 (build_one_array): Likewise.
9658 (switch_conversion::build_one_array): Likewise.
9659 (build_arrays): Likewise.
9660 (switch_conversion::build_arrays): Likewise.
9661 (gen_def_assigns): Likewise.
9662 (switch_conversion::gen_def_assigns): Likewise.
9663 (prune_bbs): Likewise.
9664 (switch_conversion::prune_bbs): Likewise.
9665 (fix_phi_nodes): Likewise.
9666 (switch_conversion::fix_phi_nodes): Likewise.
9667 (gen_inbound_check): Likewise.
9668 (switch_conversion::gen_inbound_check): Likewise.
9669 (process_switch): Use the newly created class.
9670 (switch_conversion::expand): New.
9671 (switch_conversion::~switch_conversion): New.
9672 * tree-switch-conversion.h: New file.
9674 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9676 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
9677 tree-vect-patterns.c.
9678 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
9679 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
9680 (vect_recog_sad_pattern): Likewise.
9681 (vect_recog_widen_sum_pattern): Likewise.
9682 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
9683 (vect_recog_widen_shift_pattern): Remove the type_in argument.
9684 (vect_recog_rotate_pattern): Likewise.
9685 (vect_recog_mult_pattern): Likewise.
9686 (vect_recog_vector_vector_shift_pattern): Likewise.
9687 (vect_recog_divmod_pattern): Likewise.
9688 (vect_recog_mixed_size_cond_pattern): Likewise.
9689 (vect_recog_bool_pattern): Likewise.
9690 (vect_recog_mask_conversion_pattern): Likewise.
9691 (vect_try_gather_scatter_pattern): Likewise.
9692 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
9693 (vect_recog_over_widening_pattern): Likewise.
9694 (vect_recog_gather_scatter_pattern): Likewise.
9695 (vect_recog_func_ptr): Move from tree-vectorizer.h
9696 (vect_vect_recog_func_ptrs): Move further down the file.
9697 (vect_recog_func): Likewise. Remove the third argument.
9698 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
9699 (vect_pattern_recog_1): Expect the pattern function to do any
9700 necessary target tests. Also expect it to provide a vector type.
9701 Remove the type_in handling.
9703 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9705 * tree-vect-patterns.c (vect_pattern_detected): New function.
9706 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
9707 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
9708 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
9709 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
9710 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9711 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9712 (vect_recog_mask_conversion_pattern)
9713 (vect_try_gather_scatter_pattern): Likewise.
9715 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9717 * tree-vect-patterns.c (vect_get_internal_def): New function.
9718 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9719 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
9720 (search_type_for_mask_1): Use it.
9722 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9724 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9725 redundant WIDEN_SUM_EXPR handling.
9726 (vect_recog_sad_pattern): Likewise.
9728 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9730 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9731 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
9732 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
9733 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
9734 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
9736 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9738 * tree-vect-stmts.c (vectorizable_call): Make sure that we
9739 use the stmt_vec_info of the original bb statement for the
9740 new zero assignment, even if the call is part of a pattern.
9742 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9744 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
9745 that the sequence is attached to the original statement rather
9746 than the pattern statement.
9747 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
9748 PATTERN_DEF_SEQ from the original statement rather than
9749 the main pattern statement.
9750 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
9751 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
9752 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
9754 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9756 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
9757 definition statements before the early exit for statements that aren't
9759 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
9761 (vect_transform_loop): ...here. Process pattern definition
9762 statements without first checking whether the main pattern
9763 statement is live or relevant.
9765 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9767 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9768 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9770 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
9772 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9773 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9774 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9775 (expand_block_compare): Change select_block_compare_mode call.
9776 (expand_strncmp_align_check): Use new functions, fix comment.
9777 (emit_final_str_compare_gpr): New function.
9778 (expand_strn_compare): Refactor and clean up code.
9779 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9781 2018-06-19 Tony Reix <tony.reix@atos.com>
9782 Damien Bergamini <damien.bergamini@atos.com>
9783 David Edelsohn <dje.gcc@gmail.com>
9785 * collect2.c (static_obj): New variable.
9786 (static_libs): New variable.
9787 (is_in_list): Uncomment declaration.
9788 (main): Track AIX libraries linked statically.
9789 (is_in_list): Uncomment definition.
9790 (scan_prog_file): Don't add AIX shared libraries initializer
9791 to constructor list if linking statically.
9793 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
9795 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
9797 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
9799 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
9801 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
9804 2018-06-19 Martin Liska <mliska@suse.cz>
9806 * config/i386/i386.c (ix86_can_inline_p): Do not use
9807 ipa_fn_summaries::get_create.
9808 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
9810 (devirtualization_time_bonus): Likewise.
9811 (ipcp_propagate_stage): Likewise.
9812 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9813 (edge_set_predicate): Likewise.
9814 (evaluate_conditions_for_known_args): Likewise.
9815 (evaluate_properties_for_edge): Likewise.
9816 (ipa_call_summary::reset): Tranform to ...
9817 (ipa_call_summary::~ipa_call_summary): ... this.
9818 (ipa_fn_summary::reset): Transform to ...
9819 (ipa_fn_summary::~ipa_fn_summary): ... this.
9820 (ipa_fn_summary_t::remove): Rename to ...
9821 (ipa_fn_summary_t::remove_callees): ... this.
9822 (ipa_fn_summary_t::duplicate): Use placement new
9823 instead of memory copy.
9824 (ipa_call_summary_t::duplicate): Likewise.
9825 (ipa_call_summary_t::remove): Remove.
9826 (dump_ipa_call_summary): Change get_create to get.
9827 (ipa_dump_fn_summary): Dump only when summary exists.
9828 (analyze_function_body): Use symbol_summary::get instead
9830 (compute_fn_summary): Likewise.
9831 (estimate_edge_devirt_benefit): Likewise.
9832 (estimate_edge_size_and_time): Likewise.
9833 (inline_update_callee_summaries): Likewise.
9834 (remap_edge_change_prob): Likewise.
9835 (remap_edge_summaries): Likewise.
9836 (ipa_merge_fn_summary_after_inlining): Likewise.
9837 (write_ipa_call_summary): Likewise.
9838 (ipa_fn_summary_write): Likewise.
9839 (ipa_free_fn_summary): Likewise.
9840 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
9841 (struct ipa_call_summary): Likewise.
9842 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
9844 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
9845 (estimate_size_after_inlining): Likewise.
9846 (estimate_growth): Likewise.
9847 (growth_likely_positive): Likewise.
9848 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9849 (inline_call): Likewise.
9850 * ipa-inline.c (caller_growth_limits): Likewise.
9851 (can_inline_edge_p): Likewise.
9852 (can_inline_edge_by_limits_p): Likewise.
9853 (compute_uninlined_call_time): Likewise.
9854 (compute_inlined_call_time): Likewise.
9855 (want_inline_small_function_p): Likewise.
9856 (edge_badness): Likewise.
9857 (update_caller_keys): Likewise.
9858 (update_callee_keys): Likewise.
9859 (inline_small_functions): Likewise.
9860 (inline_to_all_callers_1): Likewise.
9861 (dump_overall_stats): Likewise.
9862 (early_inline_small_functions): Likewise.
9863 (early_inliner): Likewise.
9864 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9865 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9866 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9867 * ipa-split.c (execute_split_functions): Likewise.
9868 * symbol-summary.h: Likewise.
9869 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9871 2018-06-19 Richard Biener <rguenther@suse.de>
9873 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
9874 (vectorize_loops): ... here. Fix dbgcnt handling.
9875 (try_vectorize_loop): Wrap try_vectorize_loop_1.
9877 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
9880 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
9881 ieee128 argument takes up only one (vector) register, not two (floating
9884 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
9886 * gimplify.c (gimplify_init_constructor): Really never clear for an
9887 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
9889 2018-06-19 Richard Biener <rguenther@suse.de>
9891 PR tree-optimization/86179
9892 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
9893 after failed recognition.
9895 2018-06-18 Martin Sebor <msebor@redhat.com>
9898 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
9899 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
9900 Handle integer subtraction.
9901 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
9902 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
9904 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9906 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
9907 param from rtx to rtx_insn *.
9908 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
9910 (frv_ifcvt_modify_insn): Likwise.
9911 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
9912 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
9913 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
9914 as_a <rtx_insn *> cast to local "unprotected_region" once
9915 it's been established that it's not NULL or pc_rtx.
9916 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
9917 param "sethi" from rtx to rtx_insn *.
9918 (nds32_group_float_insns): Likewise for param "insn".
9919 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
9921 (vax_output_int_subtract): Likewise.
9922 * config/vax/vax.c (vax_output_int_add): Likewise for param
9924 (vax_output_int_subtract): Likewise.
9925 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
9926 (emit_pattern_after): Likewise for param "after".
9927 (emit_insn_after): Likewise.
9928 (emit_jump_insn_after): Likewise.
9929 (emit_call_insn_after): Likewise.
9930 (emit_debug_insn_after): Likewise.
9931 (emit_pattern_before): Likewise for param "before".
9932 (emit_insn_before): Likewise.
9933 (emit_jump_insn_before): Likewise.
9934 * final.c (get_insn_template): Likewise for param "insn", removing
9936 * output.h (get_insn_template): Likewise for 2nd param.
9937 * rtl.h (emit_insn_before): Likewise.
9938 (emit_jump_insn_before): Likewise.
9939 (emit_debug_insn_before_noloc): Likewise.
9940 (emit_insn_after): Likewise.
9941 (emit_jump_insn_after): Likewise.
9942 (emit_call_insn_after): Likewise.
9943 (emit_debug_insn_after): Likewise.
9944 (set_insn_deleted): Likewise for param.
9946 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
9949 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
9950 floating point modes, so that IFmode is numerically greater than
9951 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
9952 to declare the ordering. This prevents IFmode from being
9953 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
9954 machine. Include rs6000-modes.h to share the fractional values
9955 between genmodes* and the rest of the compiler.
9959 * config/rs6000/rs6000-modes.h: New file.
9960 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
9961 meaning of rs6000_long_double_size so that 126..128 selects an
9962 appropriate 128-bit floating point type.
9963 (rs6000_option_override_internal): Likewise.
9964 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
9965 (TARGET_LONG_DOUBLE_128): Change the meaning of
9966 rs6000_long_double_size so that 126..128 selects an appropriate
9967 128-bit floating point type.
9968 (LONG_DOUBLE_TYPE_SIZE): Update comment.
9969 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
9970 source and destination to match the standard usage.
9971 (truncifkf2): Likewise.
9972 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
9973 ISA 2.07 to use an explicit clobber, instead of passing in a
9975 (copysign<mode>3_soft): Likewise.
9977 2018-06-18 David Malcolm <dmalcolm@redhat.com>
9979 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9980 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
9981 (vect_slp_analyze_instance_dependence): Likewise.
9982 (vect_enhance_data_refs_alignment): Likewise.
9983 (vect_analyze_data_refs_alignment): Likewise.
9984 (vect_slp_analyze_and_verify_instance_alignment
9985 (vect_analyze_data_ref_accesses): Likewise.
9986 (vect_prune_runtime_alias_test_list): Likewise.
9987 (vect_analyze_data_refs): Likewise.
9988 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
9989 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9990 (vect_analyze_scalar_cycles_1): Likewise.
9991 (vect_get_loop_niters): Likewise.
9992 (vect_analyze_loop_form_1): Likewise.
9993 (vect_update_vf_for_slp): Likewise.
9994 (vect_analyze_loop_operations): Likewise.
9995 (vect_analyze_loop): Likewise.
9996 (vectorizable_induction): Likewise.
9997 (vect_transform_loop): Likewise.
9998 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
9999 * tree-vect-slp.c (vect_analyze_slp): Likewise.
10000 (vect_make_slp_decision): Likewise.
10001 (vect_detect_hybrid_slp): Likewise.
10002 (vect_slp_analyze_operations): Likewise.
10003 (vect_slp_bb): Likewise.
10004 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
10005 (vectorizable_bswap): Likewise.
10006 (vectorizable_call): Likewise.
10007 (vectorizable_simd_clone_call): Likewise.
10008 (vectorizable_conversion): Likewise.
10009 (vectorizable_assignment): Likewise.
10010 (vectorizable_shift): Likewise.
10011 (vectorizable_operation): Likewise.
10012 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
10014 2018-06-18 Martin Sebor <msebor@redhat.com>
10016 PR tree-optimization/81384
10017 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
10018 * builtins.c (expand_builtin_strnlen): New function.
10019 (expand_builtin): Call it.
10020 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
10021 * builtins.def (BUILT_IN_STRNLEN): New.
10022 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
10023 Warn for bounds in excess of maximum object size.
10024 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
10025 single-value ranges. Handle strnlen.
10026 (handle_builtin_strlen): Handle strnlen.
10027 (strlen_check_and_optimize_stmt): Same.
10028 * doc/extend.texi (Other Builtins): Document strnlen.
10030 2018-06-18 Maya Rashish <coypu@sdf.org>
10032 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
10033 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
10034 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
10036 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
10038 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
10040 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10042 * tree.c (escaped_string::escape): Replace cast to char * by
10043 const_cast<char *> (unescaped).
10045 2018-06-18 Nick Clifton <nickc@redhat.com>
10048 * tree.c (escaped_string): New class. Converts an unescaped
10049 string into its escaped equivalent.
10050 (warn_deprecated_use): Use the new class to convert the
10051 deprecation message, if present.
10052 (test_escaped_strings): New self test.
10053 (test_c_tests): Add test_escaped_strings.
10054 * doc/extend.texi (deprecated): Add a note that the
10055 deprecation message is affected by the -fmessage-length
10056 option, and that control characters will be escaped.
10057 (#pragma GCC error): Document this pragma.
10058 (#pragma GCC warning): Likewise.
10059 * doc/invoke.texi (-fmessage-length): Document this option's
10060 effect on the #warning and #error preprocessor directives and
10061 the deprecated attribute.
10063 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
10065 * tree.c (decl_value_expr_lookup): Revert latest change.
10066 (decl_value_expr_insert): Likewise.
10068 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
10070 * gimplify.c (nonlocal_vlas): Delete.
10071 (nonlocal_vla_vars): Likewise.
10072 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
10074 (gimplify_body): Do not create and destroy nonlocal_vlas.
10075 * tree-nested.c: Include diagnostic.h.
10076 (use_pointer_in_frame): Tweak.
10077 (lookup_field_for_decl): Add assertion and declare the transformation.
10078 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
10079 internal error when the reference is in a wrong context. Do not
10080 create a debug decl by default.
10081 (note_nonlocal_block_vlas): Delete.
10082 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
10083 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
10084 create a debug decl by default.
10085 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
10086 call is in a wrong context.
10087 (fixup_vla_decls): New function.
10088 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
10089 debug variables were created.
10090 * tree.c (decl_value_expr_lookup): Add checking assertion.
10091 (decl_value_expr_insert): Likewise.
10093 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
10095 PR middle-end/82479
10096 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
10097 * tree-scalar-evolution.c (interpret_expr): Likewise.
10098 (expression_expensive_p): Likewise.
10099 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
10100 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
10101 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
10102 (ssa_defined_by_minus_one_stmt_p): New.
10104 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
10106 PR middle-end/64946
10107 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
10108 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
10109 * dojump.c (do_jump): Likewise.
10110 * expr.c (expand_expr_real_2): Check operand type's sign.
10111 * fold-const.c (const_unop): Handle ABSU_EXPR.
10112 (fold_abs_const): Likewise.
10113 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
10114 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
10115 (strip_sign_op_1): Likesise.
10116 * match.pd: Add new pattern to generate ABSU_EXPR.
10117 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
10118 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10119 * tree-eh.c (operation_could_trap_helper_p): Likewise.
10120 * tree-inline.c (estimate_operator_cost): Likewise.
10121 * tree-pretty-print.c (dump_generic_node): Likewise.
10122 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
10123 * tree.def (ABSU_EXPR): New.
10125 2018-06-16 Jakub Jelinek <jakub@redhat.com>
10127 PR middle-end/86095
10128 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
10129 documented as preserved for backward compatibility only.
10130 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
10132 PR rtl-optimization/86108
10133 * bb-reorder.c (create_forwarder_block): Renamed to ...
10134 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
10135 jump from new landing pad to the second part.
10136 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
10139 2018-06-15 Jakub Jelinek <jakub@redhat.com>
10141 PR middle-end/85878
10142 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
10143 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
10144 Only call store_expr for halves if the mode is the same.
10146 PR middle-end/86123
10147 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
10148 Fix up comment formatting.
10150 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
10152 * typed-splay-tree.h (typed_splay_tree::remove): New function.
10153 (typed_splay_tree::closure,
10154 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
10155 (typed_splay_tree::typed_splay_tree,
10156 typed_splay_tree::operator =): Declared private.
10157 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
10158 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
10159 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
10160 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
10161 typed_splay_tree::splay_tree_splay,
10162 typed_splay_tree::splay_tree_foreach_helper,
10163 typed_splay_tree::splay_tree_insert,
10164 typed_splay_tree::splay_tree_remove,
10165 typed_splay_tree::splay_tree_lookup,
10166 typed_splay_tree::splay_tree_predecessor,
10167 typed_splay_tree::splay_tree_successor,
10168 typed_splay_tree::splay_tree_min,
10169 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
10170 (typed_splay_tree::root, typed_splay_tree::comp,
10171 typed_splay_tree::delete_key,
10172 typed_splay_tree::delete_value): New data members.
10173 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
10174 typed_splay_tree::remove.
10176 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
10178 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
10179 -mginv and -mno-ginv to the assembler.
10180 * config/mips/mips.opt (-mcrc): New option.
10181 (-mginv): Likewise.
10182 * doc/invoke.text (-mcrc): Document.
10183 (-mginv): Likewise.
10185 2018-06-15 Nick Clifton <nickc@redhat.com>
10188 * tree.c (escaped_string): New class. Converts an unescaped
10189 string into its escaped equivalent.
10190 (warn_deprecated_use): Use the new class to convert the
10191 deprecation message, if present.
10192 (test_escaped_strings): New self test.
10193 (test_c_tests): Add test_escaped_strings.
10194 * doc/extend.texi (deprecated): Add a note that the
10195 deprecation message is affected by the -fmessage-length
10196 option, and that control characters will be escaped.
10197 (#pragma GCC error): Document this pragma.
10198 (#pragma GCC warning): Likewise.
10199 * doc/invoke.texi (-fmessage-length): Document this option's
10200 effect on the #warning and #error preprocessor directives and
10201 the deprecated attribute.
10203 2018-06-15 Richard Biener <rguenther@suse.de>
10205 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
10206 here, also noting vector size used.
10207 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
10208 size used in MSG_OPTIMIZED_LOCATIONS dump.
10209 (pass_slp_vectorize::execute): Adjust.
10211 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
10214 * config/arc/arc.c (arc_return_address_register): Fix
10217 2018-06-15 Richard Biener <rguenther@suse.de>
10219 PR middle-end/86159
10220 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
10221 leave useless conversion stripping to force_gimple_operand_gsi.
10222 (gimplify_build2): Likewise.
10223 (gimplify_build1): Likewise.
10225 2018-06-15 Richard Biener <rguenther@suse.de>
10227 PR middle-end/86076
10228 * tree-cfg.c (move_stmt_op): unshare invariant addresses
10229 before adjusting their block.
10231 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
10233 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
10234 multilibs for *-*-rtems*.
10235 * config/riscv/t-rtems: New file.
10237 2018-06-14 Jakub Jelinek <jakub@redhat.com>
10239 PR middle-end/86122
10240 * match.pd ((A +- CST1) +- CST2): Punt if last resort
10241 unsigned_type_for returns NULL.
10244 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
10245 subregs of multi-word pseudos unless the float mode has word size.
10247 2018-06-14 Richard Biener <rguenther@suse.de>
10249 PR middle-end/86139
10250 * tree-vect-generic.c (build_word_mode_vector_type): Remove
10251 duplicate and harmful type_hash_canon.
10252 * tree.c (type_hash_canon): Assert we didn't find ourselves.
10254 2018-06-14 Richard Biener <rguenther@suse.de>
10257 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
10260 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
10262 * config/rtems.h (STDINT_LONG32): Define.
10264 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
10265 Prachi Godbole <prachi.godbole@imgtec.com>
10267 * config/mips/mips-cpus.def: Define P6600.
10268 * config/mips/mips-tables.opt: Regenerate.
10269 * config/mips/mips.c (mips_ucbranch_type): New enum.
10270 (mips_rtx_cost_data): Add support for P6600.
10271 (mips_issue_rate): Likewise.
10272 (mips_multipass_dfa_lookahead): Likewise.
10273 (mips_avoid_hazard): Likewise.
10274 (mips_reorg_process_insns): Likewise.
10275 (mips_classify_branch_p6600): New function.
10276 * config/mips/mips.h (TUNE_P6600): New define.
10277 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
10278 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
10279 * config/mips/mips.md: Include p6600.md.
10280 (processor): Add p6600.
10281 * config/mips/p6600.md: New file.
10282 * doc/invoke.texi: Add p6600 to supported architectures.
10284 2018-06-13 Martin Sebor <msebor@redhat.com>
10286 PR tree-optimization/86114
10287 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
10289 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
10291 2018-06-13 Richard Biener <rguenther@suse.de>
10293 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
10294 Properly set vector type of the intermediate stmt.
10295 * tree-vect-stmts.c (vectorizable_operation): The destination
10296 var always has vectype_out type.
10298 2018-06-13 Jeff Law <law@redhat.com>
10300 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
10301 integer 0 for argument to print_rtl_with_bb.
10302 (rl78_reorg): Likewise.
10304 2018-06-13 David Malcolm <dmalcolm@redhat.com>
10306 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
10307 from rtx to rtx_insn *.
10308 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
10310 (add_sched_insns_for_speculation): Likewise for local "target",
10311 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10312 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
10313 from rtx_insn ** to rtx_code_label **.
10314 (reorg_emit_nops): Likewise.
10315 (c6x_reorg): Likewise for local "call_labels".
10316 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
10318 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
10319 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
10320 the loops over LABEL_REFS.
10321 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
10323 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10324 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
10325 (split_branches): Strengthen local "olabel" from rtx to
10326 rtx_insn *, adding a safe_as_a cast.
10327 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
10329 (add_insn_after): Likewise for first two params.
10330 (add_insn_before): Likewise.
10331 (remove_insn): Likewise for param.
10332 (emit_pattern_before_noloc): Likewise for second and third params.
10333 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
10334 (emit_call_insn_before_noloc): Likewise.
10335 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
10337 (emit_barrier_before): Likewise.
10338 (emit_label_before): Strengthen "label" param from "rtx" to
10339 "rtx_code_label *". Strengthen "before" param from "rtx" to
10341 (emit_insn_after_1): Strengthen "after" param from "rtx" to
10343 (emit_pattern_after_noloc): Likewise.
10344 (emit_insn_after_noloc): Likewise.
10345 (emit_jump_insn_after_noloc): Likewise.
10346 (emit_call_insn_after_noloc): Likewise.
10347 (emit_debug_insn_after_noloc): Likewise.
10348 (emit_barrier_after): Likewise.
10349 (emit_label_after): Likewise for both params.
10350 (emit_pattern_after_setloc): Likewise for "after" param. Convert
10351 "loc" param from "int" to "location_t".
10352 (emit_insn_after_setloc): Likewise.
10353 (emit_jump_insn_after_setloc): Likewise.
10354 (emit_call_insn_after_setloc): Likewise.
10355 (emit_debug_insn_after_setloc): Likewise.
10356 (emit_pattern_before_setloc): Likewise for "before" param. Convert
10357 "loc" param from "int" to "location_t".
10358 (emit_pattern_before): Convert NULL_RTX to NULL.
10359 (emit_insn_before_setloc): Convert "loc" param from "int" to
10361 (emit_jump_insn_before_setloc): Likewise.
10362 (emit_call_insn_before_setloc): Likewise.
10363 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
10364 rtx_insn *. Convert "loc" param from "int" to "location_t".
10365 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
10366 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
10367 Convert 3rd param from "int" to "location_t".
10368 (emit_barrier_before, emit_barrier_after, next_real_insn):
10369 Strengthen param from rtx to rtx_insn *.
10370 (emit_label_before): Strengthen 1st param from "rtx" to
10371 "rtx_code_label *". Strengthen 2nd param from "rtx" to
10373 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
10374 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
10375 Strengthen 2nd param from "rtx" to "rtx_insn *".
10376 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
10377 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
10378 Likewise. Convert 3rd param from "int" to "location_t".
10379 (emit_label_after): Strengthen 1st param from "rtx" to
10380 "rtx_code_label *".
10381 (next_real_insn, remove_insn): Strengthen param from "rtx" to
10383 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
10384 from "rtx" to "rtx_insn *".
10386 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
10388 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
10389 bodies streamed in with -Q.
10390 * dumpfile.c (dump_files): Add lto-stream-out dump file.
10391 * dumpfile.h (tree_dump_index): Add lto_stream_out.
10392 * gimple-streamer-out.c: Include gimple-pretty-print.h
10393 (output_bb): Dump stmts streamed.
10394 * lto-section-out.c: Include print-tree.h
10395 (lto_begin_section): Dump sections created.
10396 (lto_output_decl_index): Dump decl encoded.
10397 * lto-streamer-out.c: Include print-tree.h
10398 (create_output_block): Dump output block created.
10399 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
10400 (output_function): Dump function output.
10401 (output_constructor): Dump constructor streamed.
10402 (write_global_stream): Output indexes encoded.
10403 (produce_asm_for_decls): Dump streams encoded.
10404 * lto-streamer.c (streamer_dump_file): New global var.
10405 * lto-streamer.h (streamer_dump_file): Declare.
10406 * passes.c (ipa_write_summaries): Initialize streamer dump.
10407 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
10410 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
10413 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
10414 offsets for register save directives. Emit a second batch of save
10415 directives, if need be, when the function accesses prior frames.
10417 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10419 * config/arc/fpu.md (fmasf4): Force operand to register.
10420 (fnmasf4): Likewise.
10422 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10424 * config/arc/arc-protos.h (arc_pad_return): Remove.
10425 * config/arc/arc.c (machine_function): Remove force_short_suffix
10427 (arc_print_operand): Adjust printing of '&'.
10428 (arc_verify_short): Remove conditional printing of short suffix.
10429 (arc_final_prescan_insn): Remove reference to size_reason.
10430 (pad_return): New function.
10431 (arc_reorg): Call pad_return.
10432 (arc_pad_return): Remove.
10433 (arc_init_machine_status): Remove reference to force_short_suffix.
10434 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
10435 (attr length): When attribute iscompact is true force to 2
10436 regardless; in the case of maybe check if we want to force the
10437 instruction to have 4 bytes length.
10438 (nopv): Change it to generate 4 byte long nop as well.
10439 (blockage): New pattern.
10440 (simple_return): Remove call to arc_pad_return.
10441 (p_return_i): Likewise.
10443 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10445 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
10447 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10449 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
10452 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10454 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
10455 for ARC700 and ARCv2.
10457 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
10460 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
10461 operands[2] instead of operands[1].
10464 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10466 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
10467 case, check whether the outer register overlaps an unallocatable
10468 register, not just whether it fits the required class.
10470 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10472 * poly-int.h (can_div_trunc_p): Add new overload in which all values
10474 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
10475 (memrefs_conflict_p): Likewise.
10476 (init_alias_analysis): Likewise.
10477 * cfgexpand.c (expand_debug_expr): Likewise.
10478 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
10479 * cse.c (fold_rtx): Likewise.
10480 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
10481 * expr.c (emit_block_move_hints): Likewise.
10482 (clear_storage_hints, push_block, emit_push_insn): Likewise.
10483 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
10484 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
10485 (emit_group_store): Likewise.
10486 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
10487 to read the PRE/POST_MODIFY increment.
10488 * calls.c (store_one_arg): Use strip_offset.
10489 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
10491 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
10493 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
10494 (simplify_binary_operation_1): Extend CONST_INT handling to
10496 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
10497 than a HOST_WIDE_INT.
10498 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
10500 (adjust_mems, add_stores): Update accodingly.
10501 (vt_canonicalize_addr): Track polynomial offsets.
10502 (emit_note_insn_var_location): Likewise.
10503 (vt_add_function_parameter): Likewise.
10504 (vt_initialize): Likewise.
10506 2018-06-12 Jeff Law <law@redhat.com>
10508 * config.gcc (alpha*-*-freebsd*): Remove.
10509 * config/alpha/freebsd.h: Remove.
10511 2018-06-12 David Malcolm <dmalcolm@redhat.com>
10514 * spellcheck-tree.c (levenshtein_distance): Rename to...
10515 (get_edit_distance): ...this, and update for underlying renaming.
10516 * spellcheck-tree.h (levenshtein_distance): Rename to...
10517 (get_edit_distance): ...this.
10518 * spellcheck.c (levenshtein_distance): Rename to...
10519 (get_edit_distance): ...this. Convert from Levenshtein distance
10520 to Damerau-Levenshtein distance by supporting transpositions of
10521 adjacent characters. Rename "v1" to "v_next" and "v0" to
10523 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
10524 (selftest::test_edit_distance_unit_test_oneway): ...this, and
10525 update for underlying renaming.
10526 (selftest::levenshtein_distance_unit_test): Rename to...
10527 (selftest::test_get_edit_distance_unit): ...this, and update for
10528 underlying renaming.
10529 (selftest::test_find_closest_string): Add example from PR 69968
10530 where transposition helps
10531 (selftest::test_metric_conditions): Update for renaming.
10532 (selftest::test_metric_conditions): Likewise.
10533 (selftest::spellcheck_c_tests): Likewise.
10534 * spellcheck.h (levenshtein_distance): Rename both overloads to...
10535 (get_edit_distance): ...this.
10536 (best_match::consider): Update for renaming.
10538 2018-06-12 Martin Sebor <msebor@redhat.com>
10540 PR tree-optimization/85259
10541 * builtins.c (compute_objsize): Handle constant offsets.
10542 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
10543 true iff a warning has been issued.
10544 * gimple.h (gimple_nonartificial_location): New function.
10545 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
10546 gimple_nonartificial_location and handle -Wno-system-headers.
10547 (handle_builtin_stxncpy): Same.
10549 2018-06-12 Martin Sebor <msebor@redhat.com>
10552 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
10554 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
10556 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
10557 BUILTIN_VEC_XST entries for pointer to double and long long.
10559 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
10562 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10564 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10567 2018-06-12 Martin Liska <mliska@suse.cz>
10569 * doc/options.texi: Document IntegerRange.
10571 2018-06-12 Martin Liska <mliska@suse.cz>
10573 * config/i386/i386.opt: Make MPX-related options as Deprecated.
10574 * opt-functions.awk: Handle Deprecated flag.
10575 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
10577 (read_cmdline_option): Report warning for a deprecated option.
10578 * opts.h (struct cl_option): Add new field cl_deprecated.
10579 (CL_ERR_DEPRECATED): New.
10581 2018-06-12 Martin Liska <mliska@suse.cz>
10583 * doc/options.texi: Document Deprecated option flag.
10585 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10587 * config/arc/arc-arch.h (arc_extras): New enum.
10588 (arc_cpu_t):Add field extra.
10589 (arc_cpu_types): Consider the extras.
10590 * config/arc/arc-cpus.def: Add extras info.
10591 * config/arc/arc-opts.h (processor_type): Consider extra field.
10592 * config/arc/arc.c (arc_override_options): Handle extra field.
10594 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10596 * config/arc/arc-arch.h: Update ARC_OPTX macro.
10597 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
10599 * config/arc/arc.c (arc_init): Update pic warning.
10600 (irq_range): Update irq range parsing warnings.
10601 (arc_override_options): Update various warning messages.
10602 (arc_handle_aux_attribute): Likewise.
10604 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
10606 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
10608 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10610 * doc/sourcebuild.texi: Document usage of line number 0 in verify
10611 compiler messages directives.
10613 2018-06-12 Matthew Fortune <mfortune@gmail.com>
10615 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
10616 * config/mips/mips-tables.opt: Regenerate.
10617 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
10619 * doc/invoke.texi: Document -march=i6500.
10621 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
10623 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
10624 (i6400_gpmul): Add cpu_unit.
10625 (i6400_gpdiv): Likewise.
10626 (i6400_msa_add_d): Update reservations.
10627 (i6400_msa_int_add) Likewise.
10628 (i6400_msa_short_logic3) Likewise.
10629 (i6400_msa_short_logic2) Likewise.
10630 (i6400_msa_short_logic) Likewise.
10631 (i6400_msa_move) Likewise.
10632 (i6400_msa_cmp) Likewise.
10633 (i6400_msa_short_float2) Likewise.
10634 (i6400_msa_div_d) Likewise.
10635 (i6400_msa_long_logic1) Likewise.
10636 (i6400_msa_long_logic2) Likewise.
10637 (i6400_msa_mult) Likewise.
10638 (i6400_msa_long_float2) Likewise.
10639 (i6400_msa_long_float4) Likewise.
10640 (i6400_msa_long_float5) Likewise.
10641 (i6400_msa_long_float8) Likewise.
10642 (i6400_fpu_fadd): Include frint type.
10643 (i6400_fpu_store): New define_insn_reservation.
10644 (i6400_fpu_load): Likewise.
10645 (i6400_fpu_move): Likewise.
10646 (i6400_fpu_fcmp): Likewise.
10647 (i6400_fpu_fmadd): Likewise.
10648 (i6400_int_mult): Include imul3nc type and update reservation.
10649 (i6400_int_div): Include idiv3 type and update reservation.
10650 (i6400_int_load): Update to check type not move_type.
10651 (i6400_int_store): Likewise.
10652 (i6400_int_prefetch): Set zero latency.
10654 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
10656 * gcc.c: Document new %@{...} sequence.
10657 (LINK_COMMAND_SPEC): Use it for the -L switches.
10658 (cpp_unique_options): Use it for the -I switches.
10659 (at_file_argbuf): New global variable.
10660 (in_at_file): Likewise.
10661 (alloc_args): Create at_file_argbuf.
10662 (clear_args): Truncate at_file_argbuf.
10663 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
10664 (open_at_file): New function.
10665 (close_at_file): Likewise.
10666 (create_at_file): Delete.
10667 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
10670 (validate_switches_from_spec): Deal with %@{...} sequence.
10671 (validate_switches): Likewise.
10672 (driver::finalize): Call clear_args.
10674 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
10676 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
10678 2018-06-11 Martin Sebor <msebor@redhat.com>
10680 * doc/invoke.texi (-Wall): List -Wc++17-compat.
10681 (Wno-class-memaccess): Add @opindex.
10682 (Wno-templates, Wno-multiple-inheritance): Same.
10683 (Wno-virtual-inheritance, Wno-namespaces): Same.
10684 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
10685 (Wno-format-overflow, Wno-format-truncation): Same.
10686 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
10687 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
10688 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
10689 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
10690 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
10691 (Wno-misspelled-isr): Same.
10693 2018-06-11 Martin Sebor <msebor@redhat.com>
10695 * PR tree-optimization/86083
10696 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
10698 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
10700 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
10702 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
10705 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
10706 on the correct operand.
10707 (*movdi_internal64): Ditto.
10709 2018-06-11 Martin Liska <mliska@suse.cz>
10711 PR tree-optimization/86089
10712 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
10714 2018-06-11 Julia Koval <julia.koval@intel.com>
10716 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
10717 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
10718 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
10720 2018-06-11 Olivier Hainque <hainque@adacore.com>
10722 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
10723 for Ada with strict dwarf2.
10725 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
10728 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
10731 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
10733 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
10735 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
10737 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
10740 2018-06-08 Martin Liska <mliska@suse.cz>
10742 * tree-cfg.h (debug_function): Fix argument type to match
10745 2018-06-08 Martin Liska <mliska@suse.cz>
10747 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
10748 Remove usage of MPX-related (and removed) fields.
10749 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
10751 2018-06-08 David Malcolm <dmalcolm@redhat.com>
10753 * cfg.c (debug): Use TDF_NONE rather than 0.
10754 * cfghooks.c (debug): Likewise.
10755 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
10756 (struct dump_option_value_info): Convert to...
10757 (struct kv_pair): ...this template type.
10758 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
10760 (optinfo_verbosity_options): Likewise.
10761 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
10763 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10764 than int for "optgroup_flags" param.
10765 (dump_generic_expr_loc): Use dump_flags_t rather than int for
10767 (dump_dec): Likewise.
10768 (dump_finish): Use TDF_NONE rather than 0.
10769 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10770 rather than int for "optgroup_flags" param. Use TDF_NONE rather
10771 than 0. Update for change to option_ptr.
10772 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10773 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
10774 0. Update for changes to optinfo_verbosity_options and
10776 (opt_info_switch_p): Convert optgroup_flags from int to
10778 (dump_basic_block): Use dump_flags_t rather than int
10779 for "dump_kind" param.
10780 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10781 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10782 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10783 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10784 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10785 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10786 TDF_NONE): Convert from macros to...
10787 (enum dump_flag): ...this new enum.
10788 (dump_flags_t): Update to use enum.
10789 (operator|, operator&, operator~, operator|=, operator&=):
10790 Implement for dump_flags_t.
10791 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10792 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10793 Convert from macros to...
10794 (enum optgroup_flag): ...this new enum.
10795 (optgroup_flags_t): New typedef.
10796 (operator|, operator|=): Implement for optgroup_flags_t.
10797 (struct dump_file_info): Convert field "alt_flags" to
10798 dump_flags_t. Convert field "optgroup_flags" to
10800 (dump_basic_block): Use dump_flags_t rather than int for param.
10801 (dump_generic_expr_loc): Likewise.
10802 (dump_dec): Likewise.
10803 (dump_register): Convert param "optgroup_flags" to
10805 (opt_info_enable_passes): Likewise.
10806 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
10808 * gimple-pretty-print.c (debug): Likewise.
10809 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
10810 (merged_store_group::apply_stores): Likewise.
10811 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
10812 * gimple.c (verify_gimple_pp): Likewise.
10813 * graphite-poly.c (print_pbb_body): Likewise.
10814 * passes.c (pass_manager::register_one_dump_file): Convert
10815 local "optgroup_flags" to optgroup_flags_t.
10816 * print-tree.c (print_node): Use TDF_NONE rather than 0.
10818 (debug_body): Likewise.
10819 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
10820 to optgroup_flags_t.
10821 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
10823 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
10824 (convert_mult_to_fma): Likewise.
10825 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
10826 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
10827 * tree-vect-data-refs.c (dump_lower_bound): Convert param
10828 "dump_kind" to dump_flags_t.
10830 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10832 * config/rs6000/rs6000.c (min, max): Delete.
10834 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
10836 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
10837 -mabi=spe and -mabi=no-spe.
10839 2018-06-08 Martin Liska <mliska@suse.cz>
10841 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
10842 where we expect an existing summary.
10844 2018-06-08 Martin Liska <mliska@suse.cz>
10846 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
10847 * ipa-inline.h (estimate_edge_growth): Likewise.
10849 2018-06-08 Martin Liska <mliska@suse.cz>
10851 * cgraph.c (function_version_hasher::hash): Use
10852 cgraph_node::get_uid ().
10853 (function_version_hasher::equal):
10854 * cgraph.h (cgraph_node::get_uid): New method.
10855 * ipa-inline.c (update_caller_keys): Use
10856 cgraph_node::get_uid ().
10857 (update_callee_keys): Likewise.
10858 * ipa-utils.c (searchc): Likewise.
10859 (ipa_reduced_postorder): Likewise.
10860 * lto-cgraph.c (input_node): Likewise.
10861 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
10862 * symbol-summary.h (symtab_insertion): Likewise.
10863 (symtab_removal): Likewise.
10864 (symtab_duplication): Likewise.
10865 * tree-pretty-print.c (dump_function_header): Likewise.
10866 * tree-sra.c (convert_callers_for_node): Likewise.
10868 2018-06-08 Martin Liska <mliska@suse.cz>
10870 * cgraph.c (symbol_table::create_edge): Always assign a new
10872 (symbol_table::free_edge): Do not recycle numbers.
10873 * cgraph.h (cgraph_edge::get): New method.
10874 * symbol-summary.h (symtab_removal): Use it.
10875 (symtab_duplication): Likewise.
10876 (call_summary::hashable_uid): Remove.
10878 2018-06-08 Martin Liska <mliska@suse.cz>
10880 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
10881 (initialize_growth_caches): Remove.
10882 (free_growth_caches): Likewise.
10883 (do_estimate_edge_time): Use edge_growth_cache.
10884 (do_estimate_edge_size): Likewise.
10885 (do_estimate_edge_hints): Likewise.
10886 * ipa-inline.c (reset_edge_caches): Likewise.
10887 (recursive_inlining): Likewise.
10888 (inline_small_functions): Likewise.
10889 * ipa-inline.h (initialize_growth_caches): Remove.
10890 (estimate_edge_size): Likewise.
10891 (estimate_edge_time): Likewise.
10892 (estimate_edge_hints): Likewise.
10893 (reset_edge_growth_cache): Likewise.
10894 * symbol-summary.h (call_summary::remove): New method.
10896 2018-06-08 Martin Liska <mliska@suse.cz>
10898 * ipa-cp.c (class edge_clone_summary): New summary.
10899 (grow_edge_clone_vectors): Remove.
10900 (ipcp_edge_duplication_hook): Remove.
10901 (class edge_clone_summary_t): New call_summary class.
10902 (ipcp_edge_removal_hook): Remove.
10903 (edge_clone_summary_t::duplicate): New function.
10904 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
10905 (create_specialized_node): Likewise.
10906 (ipcp_driver): Initialize edge_clone_summaries and do not
10909 2018-06-08 Martin Liska <mliska@suse.cz>
10911 * symbol-summary.h (get): New function.
10912 (call_summary::m_initialize_when_cloning): New class member.
10914 2018-06-08 Martin Liska <mliska@suse.cz>
10916 * cgraph.c (cgraph_node::remove): Do not recycle uid.
10917 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
10918 (symbol_table::allocate_cgraph_symbol): Do not set uid.
10919 * passes.c (uid_hash_t): Record removed_nodes by their uids.
10920 (remove_cgraph_node_from_order): Use the removed_nodes set.
10921 (do_per_function_toporder): Likwise.
10922 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
10923 instead of summary_uid.
10924 (symtab_removal): Likewise.
10925 (symtab_duplication): Likewise.
10927 2018-06-08 Martin Liska <mliska@suse.cz>
10929 * ipa-cp.c (ipcp_store_bits_results): Use
10930 ipcp_transformation_sum.
10931 (ipcp_store_vr_results): Likewise.
10932 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
10934 (ipcp_transformation_initialize): ... this.
10935 (ipa_set_node_agg_value_chain):
10936 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
10937 (write_ipcp_transformation_info): Likewise.
10938 (read_ipcp_transformation_info): Likewise.
10939 (ipcp_update_bits): Likewise.
10940 (ipcp_update_vr): Likewise.
10941 (ipcp_transform_function): Likewise.
10942 * ipa-prop.h: Rename ipcp_transformation_summary to
10943 ipcp_transformation.
10944 (class ipcp_transformation_t): New function summary.
10945 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
10946 (ipa_get_agg_replacements_for_node): Likewise.
10948 2018-06-08 Martin Liska <mliska@suse.cz>
10950 * ipa-pure-const.c (struct funct_state_d): Do it class instead
10952 (class funct_state_summary_t): New function_summary class.
10953 (has_function_state): Remove.
10954 (get_function_state): Likewise.
10955 (set_function_state): Likewise.
10956 (add_new_function): Likewise.
10957 (funct_state_summary_t::insert): New function.
10958 (duplicate_node_data): Remove.
10959 (remove_node_data): Remove.
10960 (funct_state_summary_t::duplicate): New function.
10961 (register_hooks): Create new funct_state_summaries.
10962 (pure_const_generate_summary): Use it.
10963 (pure_const_write_summary): Likewise.
10964 (pure_const_read_summary): Likewise.
10965 (propagate_pure_const): Likewise.
10966 (propagate_nothrow): Likewise.
10967 (dump_malloc_lattice): Likewise.
10968 (propagate_malloc): Likewise.
10969 (execute): Do not register hooks, just remove summary
10971 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
10974 2018-06-08 Martin Liska <mliska@suse.cz>
10976 * ipa-reference.c (remove_node_data): Remove.
10977 (duplicate_node_data): Likewise.
10978 (class ipa_ref_var_info_summary_t): New class.
10979 (class ipa_ref_opt_summary_t): Likewise.
10980 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
10981 (get_reference_optimization_summary): Use
10982 ipa_ref_opt_sum_summaries.
10983 (set_reference_vars_info): Remove.
10984 (set_reference_optimization_summary): Likewise.
10985 (ipa_init): Create summaries.
10986 (init_function_info): Use function summary.
10987 (ipa_ref_opt_summary_t::duplicate): New function.
10988 (ipa_ref_opt_summary_t::remove): New function.
10989 (get_read_write_all_from_node): Fix GNU coding style.
10990 (propagate): Use function summary.
10991 (write_node_summary_p): Fix GNU coding style.
10992 (stream_out_bitmap): Likewise.
10993 (ipa_reference_read_optimization_summary): Use function summary.
10994 (ipa_reference_c_finalize): Do not release hooks.
10996 2018-06-08 Martin Liska <mliska@suse.cz>
10998 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
10999 (analyze_function_body): Extract multiple calls of get_create.
11000 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11001 * ipa-inline.c (recursive_inlining): Use ::get method.
11002 * ipa-inline.h (estimate_edge_growth): Likewise.
11004 2018-06-08 Martin Liska <mliska@suse.cz>
11006 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
11008 (hsa_function_summary::hsa_function_summary): Use the new enum
11010 (hsa_gpu_implementation_p): Use hsa_summaries::get.
11011 * hsa-gen.c (hsa_get_host_function): Likewise.
11012 (get_brig_function_name): Likewise.
11013 * ipa-hsa.c (process_hsa_functions): Likewise.
11014 (ipa_hsa_write_summary): Likewise.
11015 * symbol-summary.h (symtab_duplication): Use ::get function/
11016 (get): New function.
11018 2018-06-08 Martin Liska <mliska@suse.cz>
11020 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
11022 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
11023 (hsa_register_kernel): Likewise.
11024 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
11025 * hsa-gen.c (hsa_get_host_function): Likewise.
11026 (get_brig_function_name): Likewise.
11027 (generate_hsa): Likewise.
11028 (pass_gen_hsail::execute): Likewise.
11029 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
11030 (devirtualization_time_bonus): Likewise.
11031 (ipcp_propagate_stage): Likewise.
11032 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11033 (edge_set_predicate): Likewise.
11034 (evaluate_conditions_for_known_args): Likewise.
11035 (evaluate_properties_for_edge): Likewise.
11036 (ipa_fn_summary::reset): Likewise.
11037 (ipa_fn_summary_t::duplicate): Likewise.
11038 (dump_ipa_call_summary): Likewise.
11039 (ipa_dump_fn_summary): Likewise.
11040 (analyze_function_body): Likewise.
11041 (compute_fn_summary): Likewise.
11042 (estimate_edge_devirt_benefit): Likewise.
11043 (estimate_edge_size_and_time): Likewise.
11044 (estimate_calls_size_and_time): Likewise.
11045 (estimate_node_size_and_time): Likewise.
11046 (inline_update_callee_summaries): Likewise.
11047 (remap_edge_change_prob): Likewise.
11048 (remap_edge_summaries): Likewise.
11049 (ipa_merge_fn_summary_after_inlining): Likewise.
11050 (ipa_update_overall_fn_summary): Likewise.
11051 (read_ipa_call_summary): Likewise.
11052 (inline_read_section): Likewise.
11053 (write_ipa_call_summary): Likewise.
11054 (ipa_fn_summary_write): Likewise.
11055 (ipa_free_fn_summary): Likewise.
11056 * ipa-hsa.c (process_hsa_functions): Likewise.
11057 (ipa_hsa_write_summary): Likewise.
11058 (ipa_hsa_read_section): Likewise.
11059 * ipa-icf.c (sem_function::merge): Likewise.
11060 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11061 (do_estimate_edge_time): Likewise.
11062 (estimate_size_after_inlining): Likewise.
11063 (estimate_growth): Likewise.
11064 (growth_likely_positive): Likewise.
11065 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11066 (inline_call): Likewise.
11067 * ipa-inline.c (caller_growth_limits): Likewise.
11068 (can_inline_edge_p): Likewise.
11069 (can_inline_edge_by_limits_p): Likewise.
11070 (compute_uninlined_call_time): Likewise.
11071 (compute_inlined_call_time): Likewise.
11072 (want_inline_small_function_p): Likewise.
11073 (edge_badness): Likewise.
11074 (update_caller_keys): Likewise.
11075 (update_callee_keys): Likewise.
11076 (recursive_inlining): Likewise.
11077 (inline_small_functions): Likewise.
11078 (inline_to_all_callers_1): Likewise.
11079 (dump_overall_stats): Likewise.
11080 (early_inline_small_functions): Likewise.
11081 (early_inliner): Likewise.
11082 * ipa-inline.h (estimate_edge_growth): Likewise.
11083 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11084 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11085 * ipa-prop.h (IPA_NODE_REF): Likewise.
11086 (IPA_EDGE_REF): Likewise.
11087 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11088 (propagate_malloc): Likewise.
11089 * ipa-split.c (execute_split_functions): Likewise.
11090 * symbol-summary.h: Rename get to get_create.
11092 (get_create): Likewise.
11093 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11095 2018-06-08 Martin Liska <mliska@suse.cz>
11097 * symbol-summary.h (release): Move definition out of class
11099 (symtab_removal): Likewise.
11100 (symtab_duplication): Likewise.
11102 2018-06-08 Martin Liska <mliska@suse.cz>
11104 * symbol-summary.h (function_summary): Move constructor
11105 implementation out of class declaration.
11106 (release): Likewise.
11107 (symtab_insertion): Likewise.
11108 (symtab_removal): Likewise.
11109 (symtab_duplication): Likewise.
11112 2018-06-08 Martin Liska <mliska@suse.cz>
11114 * Makefile.in: Remove support for MPX (macros, related functions,
11115 fields in cgraph_node, ...).
11116 * builtin-types.def (BT_BND): Likewise.
11117 (BT_FN_BND_CONST_PTR): Likewise.
11118 (BT_FN_CONST_PTR_BND): Likewise.
11119 (BT_FN_VOID_PTR_BND): Likewise.
11120 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
11121 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
11122 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
11123 (expand_builtin_mempcpy_with_bounds): Likewise.
11124 (expand_builtin_memset_with_bounds): Likewise.
11125 (expand_builtin_memset_args): Likewise.
11126 (std_expand_builtin_va_start): Likewise.
11127 (expand_builtin): Likewise.
11128 (expand_builtin_with_bounds): Likewise.
11129 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
11130 (DEF_LIB_BUILTIN_CHKP): Likewise.
11131 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
11132 (DEF_CHKP_BUILTIN): Likewise.
11133 (BUILT_IN_MEMCPY): Likewise.
11134 (BUILT_IN_MEMMOVE): Likewise.
11135 (BUILT_IN_MEMPCPY): Likewise.
11136 (BUILT_IN_MEMSET): Likewise.
11137 (BUILT_IN_STPCPY): Likewise.
11138 (BUILT_IN_STRCAT): Likewise.
11139 (BUILT_IN_STRCHR): Likewise.
11140 (BUILT_IN_STRCPY): Likewise.
11141 (BUILT_IN_STRLEN): Likewise.
11142 (BUILT_IN_MEMCPY_CHK): Likewise.
11143 (BUILT_IN_MEMMOVE_CHK): Likewise.
11144 (BUILT_IN_MEMPCPY_CHK): Likewise.
11145 (BUILT_IN_MEMSET_CHK): Likewise.
11146 (BUILT_IN_STPCPY_CHK): Likewise.
11147 (BUILT_IN_STRCAT_CHK): Likewise.
11148 (BUILT_IN_STRCPY_CHK): Likewise.
11149 * calls.c (store_bounds): Likewise.
11150 (emit_call_1): Likewise.
11151 (special_function_p): Likewise.
11152 (maybe_warn_nonstring_arg): Likewise.
11153 (initialize_argument_information): Likewise.
11154 (finalize_must_preallocate): Likewise.
11155 (compute_argument_addresses): Likewise.
11156 (expand_call): Likewise.
11157 * cfgexpand.c (expand_call_stmt): Likewise.
11158 (expand_return): Likewise.
11159 (expand_gimple_stmt_1): Likewise.
11160 (pass_expand::execute): Likewise.
11161 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
11162 (cgraph_node::remove): Likewise.
11163 (cgraph_node::dump): Likewise.
11164 (cgraph_node::verify_node): Likewise.
11165 * cgraph.h (chkp_function_instrumented_p): Likewise.
11166 (symtab_node::get_alias_target): Likewise.
11167 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
11168 (cgraph_local_p): Likewise.
11169 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
11170 (cgraph_edge::rebuild_references): Likewise.
11171 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
11172 (walk_polymorphic_call_targets): Likewise.
11173 (cgraph_node::expand_thunk): Likewise.
11174 (symbol_table::output_weakrefs): Likewise.
11175 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
11176 (ix86_handle_option): Likewise.
11177 * config/i386/constraints.md: Likewise.
11178 * config/i386/i386-builtin-types.def (BND): Likewise.
11182 * config/i386/i386-builtin.def (BDESC_END): Likewise.
11183 (BDESC_FIRST): Likewise.
11185 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
11186 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
11187 * config/i386/i386.c (enum reg_class): Likewise.
11188 (ix86_target_string): Likewise.
11189 (ix86_option_override_internal): Likewise.
11190 (ix86_conditional_register_usage): Likewise.
11191 (ix86_valid_target_attribute_inner_p): Likewise.
11192 (ix86_set_indirect_branch_type): Likewise.
11193 (ix86_set_current_function): Likewise.
11194 (ix86_function_arg_regno_p): Likewise.
11195 (init_cumulative_args): Likewise.
11196 (ix86_function_arg_advance): Likewise.
11197 (ix86_function_arg): Likewise.
11198 (ix86_pass_by_reference): Likewise.
11199 (ix86_function_value_regno_p): Likewise.
11200 (ix86_function_value_1): Likewise.
11201 (ix86_function_value_bounds): Likewise.
11202 (ix86_return_in_memory): Likewise.
11203 (ix86_setup_incoming_vararg_bounds): Likewise.
11204 (ix86_va_start): Likewise.
11205 (indirect_thunk_need_prefix): Likewise.
11206 (print_reg): Likewise.
11207 (ix86_print_operand): Likewise.
11208 (ix86_expand_call): Likewise.
11209 (ix86_output_function_return): Likewise.
11210 (reg_encoded_number): Likewise.
11211 (BDESC_VERIFYS): Likewise.
11212 (ix86_init_mpx_builtins): Likewise.
11213 (ix86_init_builtins): Likewise.
11214 (ix86_emit_cmove): Likewise.
11215 (ix86_emit_move_max): Likewise.
11216 (ix86_expand_builtin): Likewise.
11217 (ix86_builtin_mpx_function): Likewise.
11218 (ix86_get_arg_address_for_bt): Likewise.
11219 (ix86_load_bounds): Likewise.
11220 (ix86_store_bounds): Likewise.
11221 (ix86_load_returned_bounds): Likewise.
11222 (ix86_store_returned_bounds): Likewise.
11223 (ix86_class_likely_spilled_p): Likewise.
11224 (ix86_hard_regno_mode_ok): Likewise.
11225 (x86_order_regs_for_local_alloc): Likewise.
11226 (ix86_mitigate_rop): Likewise.
11227 (ix86_bnd_prefixed_insn_p): Likewise.
11228 (ix86_mpx_bound_mode): Likewise.
11229 (ix86_make_bounds_constant): Likewise.
11230 (ix86_initialize_bounds): Likewise.
11231 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
11232 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
11233 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
11234 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
11235 (TARGET_CHKP_BOUND_MODE): Likewise.
11236 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
11237 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
11238 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
11239 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
11240 * config/i386/i386.h (TARGET_MPX): Likewise.
11241 (TARGET_MPX_P): Likewise.
11242 (VALID_BND_REG_MODE): Likewise.
11243 (FIRST_BND_REG): Likewise.
11244 (LAST_BND_REG): Likewise.
11245 (enum reg_class): Likewise.
11246 (BND_REG_P): Likewise.
11247 (BND_REGNO_P): Likewise.
11248 (BNDmode): Likewise.
11249 (ADJUST_INSN_LENGTH): Likewise.
11250 * config/i386/i386.md: Likewise.
11251 * config/i386/i386.opt: Likewise.
11252 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
11253 (defined): Likewise.
11254 (LINK_MPX): Likewise.
11255 (MPX_SPEC): Likewise.
11256 (LIBMPX_SPEC): Likewise.
11257 (LIBMPXWRAPPERS_SPEC): Likewise.
11258 (CHKP_SPEC): Likewise.
11259 * config/i386/predicates.md: Likewise.
11260 * dbxout.c (dbxout_type): Likewise.
11261 * doc/extend.texi: Likewise.
11262 * doc/invoke.texi: Likewise.
11263 * doc/md.texi: Likewise.
11264 * doc/tm.texi: Likewise.
11265 * doc/tm.texi.in: Likewise.
11266 * dwarf2out.c (is_base_type): Likewise.
11267 (gen_formal_types_die): Likewise.
11268 (gen_subprogram_die): Likewise.
11269 (gen_type_die_with_usage): Likewise.
11270 (gen_decl_die): Likewise.
11271 (dwarf2out_late_global_decl): Likewise.
11272 * expr.c (expand_assignment): Likewise.
11273 (emit_storent_insn): Likewise.
11274 (store_expr_with_bounds): Likewise.
11275 (store_expr): Likewise.
11276 (expand_expr_real_1): Likewise.
11277 * expr.h (store_expr_with_bounds): Likewise.
11278 * function.c (use_register_for_decl): Likewise.
11279 (struct bounds_parm_data): Likewise.
11280 (assign_parms_augmented_arg_list): Likewise.
11281 (assign_parm_find_entry_rtl): Likewise.
11282 (assign_parm_is_stack_parm): Likewise.
11283 (assign_parm_load_bounds): Likewise.
11284 (assign_bounds): Likewise.
11285 (assign_parms): Likewise.
11286 (expand_function_start): Likewise.
11287 * gcc.c (CHKP_SPEC): Likewise.
11288 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
11289 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
11290 (wrestrict_dom_walker::check_call): Likewise.
11291 * gimple.c (gimple_build_call_from_tree): Likewise.
11292 * gimple.h (enum gf_mask): Likewise.
11293 (gimple_call_with_bounds_p): Likewise.
11294 (gimple_call_set_with_bounds): Likewise.
11295 * gimplify.c (gimplify_init_constructor): Likewise.
11296 * ipa-cp.c (initialize_node_lattices): Likewise.
11297 (propagate_constants_across_call): Likewise.
11298 (find_more_scalar_values_for_callers_subset): Likewise.
11299 * ipa-hsa.c (process_hsa_functions): Likewise.
11300 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
11301 * ipa-icf.c (sem_function::merge): Likewise.
11302 * ipa-inline.c (early_inliner): Likewise.
11303 * ipa-pure-const.c (warn_function_noreturn): Likewise.
11304 (warn_function_cold): Likewise.
11305 (propagate_pure_const): Likewise.
11306 * ipa-ref.h (enum GTY): Likewise.
11307 * ipa-split.c (find_retbnd): Likewise.
11308 (consider_split): Likewise.
11309 (split_function): Likewise.
11310 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11311 * ipa.c (walk_polymorphic_call_targets): Likewise.
11312 (symbol_table::remove_unreachable_nodes): Likewise.
11313 (process_references): Likewise.
11314 (cgraph_build_static_cdtor_1): Likewise.
11315 * lto-cgraph.c (lto_output_node): Likewise.
11316 (output_refs): Likewise.
11317 (compute_ltrans_boundary): Likewise.
11318 (input_overwrite_node): Likewise.
11319 (input_node): Likewise.
11320 (input_cgraph_1): Likewise.
11321 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
11322 * passes.c (pass_manager::execute_early_local_passes): Likewise.
11323 (class pass_chkp_instrumentation_passes): Likewise.
11324 (make_pass_chkp_instrumentation_passes): Likewise.
11325 * passes.def: Likewise.
11326 * rtl.h (struct GTY): Likewise.
11327 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
11328 * stor-layout.c (layout_type): Likewise.
11329 * symtab.c: Likewise.
11330 * target.def: Likewise.
11331 * targhooks.c (default_chkp_bound_type): Likewise.
11332 (default_chkp_bound_mode): Likewise.
11333 (default_builtin_chkp_function): Likewise.
11334 (default_chkp_function_value_bounds): Likewise.
11335 (default_chkp_make_bounds_constant): Likewise.
11336 (default_chkp_initialize_bounds): Likewise.
11337 * targhooks.h (default_chkp_bound_type): Likewise.
11338 (default_chkp_bound_mode): Likewise.
11339 (default_builtin_chkp_function): Likewise.
11340 (default_chkp_function_value_bounds): Likewise.
11341 (default_chkp_make_bounds_constant): Likewise.
11342 (default_chkp_initialize_bounds): Likewise.
11343 * toplev.c (compile_file): Likewise.
11344 (process_options): Likewise.
11345 * tree-core.h (DEF_BUILTIN): Likewise.
11346 (DEF_BUILTIN_CHKP): Likewise.
11347 * tree-inline.c (declare_return_variable): Likewise.
11348 (remap_gimple_stmt): Likewise.
11349 (copy_bb): Likewise.
11350 (initialize_inlined_parameters): Likewise.
11351 (expand_call_inline): Likewise.
11352 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
11353 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
11354 (make_pass_ipa_chkp_produce_thunks): Likewise.
11355 (make_pass_chkp): Likewise.
11356 (make_pass_chkp_opt): Likewise.
11357 (make_pass_chkp_instrumentation_passes): Likewise.
11358 * tree-pretty-print.c (dump_generic_node): Likewise.
11359 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
11360 * tree-ssa-dce.c (propagate_necessity): Likewise.
11361 (eliminate_unnecessary_stmts): Likewise.
11362 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
11363 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
11364 * tree-ssa-sccvn.h: Likewise.
11365 * tree-ssa-strlen.c (get_string_length): Likewise.
11366 (valid_builtin_call): Likewise.
11367 (adjust_last_stmt): Likewise.
11368 (handle_builtin_strchr): Likewise.
11369 (handle_builtin_strcpy): Likewise.
11370 (handle_builtin_stxncpy): Likewise.
11371 (handle_builtin_memcpy): Likewise.
11372 (handle_builtin_strcat): Likewise.
11373 (strlen_check_and_optimize_stmt): Likewise.
11374 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
11375 * tree-streamer-in.c: Likewise.
11376 * tree-streamer.c (record_common_node): Likewise.
11377 * tree.c (tree_code_size): Likewise.
11378 (wide_int_to_tree_1): Likewise.
11379 (type_contains_placeholder_1): Likewise.
11380 (build_common_tree_nodes): Likewise.
11381 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
11382 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
11383 (POINTER_BOUNDS_P): Likewise.
11384 (BOUNDED_TYPE_P): Likewise.
11385 (BOUNDED_P): Likewise.
11386 (CALL_WITH_BOUNDS_P): Likewise.
11387 (pointer_bounds_type_node): Likewise.
11388 * value-prof.c (gimple_ic): Likewise.
11389 * var-tracking.c (vt_add_function_parameters): Likewise.
11390 * varasm.c (make_decl_rtl): Likewise.
11391 (assemble_start_function): Likewise.
11392 (output_constant): Likewise.
11393 (maybe_assemble_visibility): Likewise.
11394 * varpool.c (ctor_for_folding): Likewise.
11395 * chkp-builtins.def: Remove.
11396 * ipa-chkp.c: Remove.
11397 * ipa-chkp.h: Remove.
11398 * rtl-chkp.c: Remove.
11399 * rtl-chkp.h: Remove.
11400 * tree-chkp-opt.c: Remove.
11401 * tree-chkp.c: Remove.
11402 * tree-chkp.h: Remove.
11404 2018-06-07 Carl Love <cel@us.ibm.com>
11406 * config/rs6000/vsx.md (vextract_fp_from_shorth,
11407 vextract_fp_from_shortl): Add BE support.
11409 2018-06-07 Paul Koning <ni1d@arrl.net>
11411 * compare-elim.c (try_merge_compare): Don't merge compare if
11412 address contains a side effect.
11413 (try_eliminate_compare): Likewise.
11415 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
11417 * config.gcc: Support "tremont".
11418 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
11419 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11421 * config/i386/i386.c (m_TREMONT): Define.
11422 (processor_target_table): Add "tremont".
11423 (PTA_TREMONT): Define.
11424 (ix86_lea_outperforms): Add TARGET_TREMONT.
11425 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
11426 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
11427 and M_INTEL_GOLDMONT_PLUS.
11428 (fold_builtin_cpu): Add "tremont".
11429 (ix86_add_stmt_cost): Add TARGET_TREMONT.
11430 (ix86_option_override_internal): Add "tremont".
11431 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
11432 (processor_type): Add PROCESSOR_TREMONT.
11433 * config/i386/x86-tune.def: Add m_TREMONT.
11434 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
11436 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11438 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
11439 symbol defined for msp430i* devices to be lower case.
11441 2018-06-07 Richard Biener <rguenther@suse.de>
11443 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
11444 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
11445 Properly wrap signed arithmetic if overflow wraps.
11447 2018-06-07 Jakub Jelinek <jakub@redhat.com>
11449 PR tree-optimization/69615
11450 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
11451 of a cast from a same precision integral SSA_NAME in a bb dominated
11452 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
11453 cast to utype if rhs2 has already a compatible type.
11455 2018-06-07 Richard Biener <rguenther@suse.de>
11457 PR tree-optimization/85935
11458 * graphite-scop-detection.c (find_params_in_bb): Analyze
11459 condition operands with respect to the correct loop. Assert
11460 the analysis doesn't fail.
11462 2018-06-04 Carl Love <cel@us.ibm.com>
11464 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
11465 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
11466 as it is slightly cheaper.
11467 (first_match_or_eos_index_<mode>):
11468 Calculate index using natural element order.
11469 (first_match_index_<mode>):
11470 Calculate index using natural element order.
11471 (first_match_or_eos_index_<mode>):
11472 Calculate index using natural order.
11473 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
11474 for BE and LE modes.
11475 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
11476 P9V_BUILTIN_VCLZLSBB_V16QI.
11477 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
11480 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11482 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
11483 indentation and line wrap for many prototypes. Add missing
11484 @smallexample directives around block of prototypes for vec_xl and
11487 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
11489 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
11490 track if we pass or return IEEE 128-bit floating point.
11491 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
11492 C++ mangling that is compatible with GCC 8.1.
11493 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
11494 (init_cumulative_args): Note if we pass or return IEEE 128-bit
11495 floating point types.
11496 (rs6000_function_arg_advance_1): Likewise.
11497 (rs6000_mangle_type): Optionally generate mangled names that match
11498 what GCC 8.1 generated for IEEE 128-bit floating point types.
11499 (rs6000_globalize_decl_name): If we have an external function that
11500 passes or returns IEEE 128-bit types, generate a weak reference
11501 from the mangled name used in GCC 8.1 to the current mangled
11503 (rs6000_init_builtins): Make __ibm128 use the long double type if
11504 long double is IBM extended double. Make __float128 use the long
11505 double type if long double is IEEE 128-bit.
11508 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
11509 macro for __ibm128 built-in functions.
11510 (PACK_IF): Add __ibm128 pack/unpack functions.
11511 (UNPACK_IF): Likewise.
11512 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
11513 enable long double built-in functions if long double is IEEE
11514 128-bit floating point.
11515 (rs6000_invalid_builtin): Update long double built-in function
11517 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
11518 functions, adjust the built-in function to use the long double
11519 built-in function if __ibm128 and long double are the same type.
11520 * doc/extend.texi (PowerPC builtins): Update documention for
11521 __builtin_{,un}pack_longdouble. Add documentation for
11522 __builtin_{,un}pack_ibm128.
11524 2018-06-06 Jim Wilson <jimw@sifive.com>
11526 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
11527 (struct machine_function): New field interrupt_mode.
11528 (riscv_handle_type_attribute): New function. Add forward declaration.
11529 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
11530 (riscv_expand_epilogue): Check interrupt_mode field.
11531 (riscv_set_current_function): Check interrupt attribute args and
11532 set interrupt_mode field.
11533 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
11534 (riscv_sret, riscv_uret): New.
11535 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
11536 new arguments to interrupt attribute.
11538 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
11541 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
11542 Don't handle -mcpu=power8 if -mpower9-vector is also used.
11544 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11546 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
11547 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
11548 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11549 several redundant entries.
11551 2018-06-06 David Malcolm <dmalcolm@redhat.com>
11553 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
11554 type from "rtx" to "rtx_insn *".
11555 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
11556 for local "call_insn", removing cast.
11557 (ix86_expand_call): Likewise, introducing a "call_insn" local.
11559 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
11561 PR tree-optimization/86066
11562 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
11563 for BIT_INSERT_EXPR stores.
11565 2018-06-06 Richard Biener <rguenther@suse.de>
11567 PR tree-optimization/86062
11568 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11569 component refs ontop
11570 of to be offsetted base.
11572 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11574 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
11575 to be static and remove check on interrupt attribute name.
11577 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
11579 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11580 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
11582 2018-06-05 Steve Ellcey <sellcey@cavium.com>
11585 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
11587 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
11588 Remove second argument, change how error is called.
11589 (aarch64_layout_arg): Remove second argument from
11590 aarch64_err_no_fpadvsimd call.
11591 (aarch64_init_cumulative_args): Ditto.
11592 (aarch64_gimplify_va_arg_expr): Ditto.
11593 * config/aarch64/aarch64.md (mov<mode>): Ditto.
11595 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
11597 * config/i386/i386.md (simple_return_indirect_internal): New expander.
11598 (*simple_return_indirect_internal<mode>): Rename from
11599 simple_return_indirect_internal. Use W mode iterator.
11600 (rstorssp): New expander.
11601 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
11602 (clrssbsy): New expander.
11603 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
11605 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
11607 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
11609 (cmse_check_pointed_object): Likewise.
11611 2018-06-05 Martin Liska <mliska@suse.cz>
11613 PR gcov-profile/47618
11614 * doc/invoke.texi: Document how -fprofile-dir format
11617 2018-06-05 Richard Biener <rguenther@suse.de>
11619 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
11620 removal pretend DOM info isn't available so we do not update
11621 it and only remove edges, not dominated blocks. Actually free
11622 DOM info in case we removed something. Remove unreachable blocks.
11623 (mfb_keep_latches): Work with either DOM info or marked backedges.
11624 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
11625 first. Mark backedges if DOM info isn't available.
11626 (Re-)compute DOM info after cleanup_control_flow_pre.
11628 2018-06-05 Richard Biener <rguenther@suse.de>
11630 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
11631 (locus_discrim_hasher::hash): Adjust.
11632 (locus_discrim_hasher::equal): Likewise.
11633 (next_discriminator_for_locus): Work on line directly.
11634 (same_line_p): Pass in expanded locus1 as well.
11635 (assign_discriminators): Avoid redundant location expansions.
11637 2018-06-05 Richard Biener <rguenther@suse.de>
11639 PR tree-optimization/86046
11640 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
11641 if required after clearing TREE_ADDRESSABLE.
11643 2018-06-05 Richard Biener <rguenther@suse.de>
11645 PR tree-optimization/86047
11646 * tree-ssa-loop.c (for_each_index): Glob handling of all
11647 decls and constants and really handle all of them.
11649 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11652 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
11653 qualifier_void_pointer and qualifier_const_void_pointer.
11654 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
11655 (arm_init_builtins): Handle the above.
11656 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
11657 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
11658 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
11661 2018-06-05 Martin Liska <mliska@suse.cz>
11663 * auto-profile.c (read_autofdo_file): Do not use
11664 gcov_ctr_summary struct.
11665 (afdo_callsite_hot_enough_for_early_inline): Likewise.
11666 * coverage.c (struct counts_entry): Likewise.
11667 (read_counts_file): Read just single summary entry.
11668 (get_coverage_counts): Use gcov_summary struct.
11669 * coverage.h (get_coverage_counts): Likewise.
11670 * gcov-dump.c (dump_working_sets): Likewise.
11671 (tag_summary): Dump just single summary.
11672 * gcov-io.c (gcov_write_summary): Write just histogram
11674 (gcov_read_summary): Read just single summary.
11675 (compute_working_sets): Use gcov_summary struct.
11676 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
11677 of GCOV_COUNTERS_SUMMABLE.
11678 (GCOV_COUNTERS_SUMMABLE): Remove.
11679 (GCOV_FIRST_VALUE_COUNTER): Replace with
11680 GCOV_COUNTER_V_INTERVAL.
11681 (struct gcov_ctr_summary): Remove.
11682 (struct gcov_summary): Directly use fields of former
11684 (compute_working_sets): Use gcov_summary struct.
11685 * gcov.c (read_count_file): Do not use ctrs fields.
11686 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
11688 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
11690 * profile.c: Likewise.
11691 * profile.h: Likewise.
11693 2018-06-05 Martin Liska <mliska@suse.cz>
11695 PR gcov-profile/84846
11696 * gcov.c (output_lines): Print working directory only
11697 in intermediate format.
11699 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
11701 * config/s390/s390-builtin-types.def: Add void function type.
11702 * config/s390/s390-builtins.def: Use the function type for the
11705 2018-06-04 Jim Wilson <jimw@sifive.com>
11707 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
11709 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
11710 and maybe_eh_return. Change regno to unsigned int. Use new args to
11711 handle EH_RETURN_DATA_REGNO registers properly.
11712 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
11713 (riscv_expand_epilogue): Update comment. Change argument name and
11714 type. Update code to use new name and type. Pass new args to
11715 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
11717 * config/riscv/riscv.md (NORMAL_RETURN): New.
11718 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
11719 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
11720 (eh_return): Call gen_eh_return_internal and emit barrier.
11721 (eh_return_internal): Call riscv_expand_epilogue.
11723 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
11725 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
11726 bit_insertion field and declare can_be_merged_into method.
11727 (merged_store_group::can_be_merged_into): New method.
11728 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
11729 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
11730 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
11732 2018-06-04 Richard Biener <rguenther@suse.de>
11734 PR tree-optimization/85955
11735 * builtins.c (fold_builtin_sincos): Convert pointers to
11736 destination to appropriate type before dereferencing.
11738 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
11740 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
11742 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
11744 * expr.c (expand_expr_real_1): Force the operand into memory if
11745 its TYPE_MODE is BLKmode and if there is no integer mode for
11746 the number of bits being extracted.
11748 2018-06-04 Jakub Jelinek <jakub@redhat.com>
11752 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11753 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
11755 2018-06-04 Richard Biener <rguenther@suse.de>
11757 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
11758 (cleanup_tree_cfg_noloop): ... single caller. Do
11759 start_recording_case_labels later.
11761 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
11763 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11764 to _IMMINTRIN_H_INCLUDED.
11765 * config/i386/pconfigintrin.h: Ditto.
11766 * config/i386/waitpkgintrin.h: Ditto.
11767 * config/i386/immintrin.h: Add includes for sgxintrin.h,
11768 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11769 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11770 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11771 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11772 waitpkgintrin.h and cldemoteintrin.h.
11774 2018-06-04 Richard Biener <rguenther@suse.de>
11776 PR tree-optimization/86038
11777 * tracer.c (find_best_successor): Check probability for
11778 being initialized, bail out if not.
11780 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
11783 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11784 of bits to ignore when comparing architectures.
11786 2018-06-04 Jakub Jelinek <jakub@redhat.com>
11788 PR tree-optimization/69615
11789 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11790 maximum or minimum of the type, try to merge it also as if
11791 range1 is + [-, x - 1] or + [x + 1, -].
11794 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
11796 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
11798 PR tree-optimization/86034
11799 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
11800 the unsigned bitfield type in a bit insertion sequence if it does not
11801 have a larger precision than the bitfield size.
11802 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
11804 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
11806 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
11808 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11810 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
11811 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
11812 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
11813 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
11815 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
11817 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
11818 Disable -fdelete-null-pointer-checks for ELF toolchain.
11820 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11821 Kito Cheng <kito.cheng@gmail.com>
11823 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
11824 (nds32le-*-*, nds32be-*-*): Integrate checking process.
11825 (nds32*-*-*): Add glibc and uclibc conditions.
11826 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
11827 (TARGET_EXCEPT_UNWIND_INFO): Define.
11828 * config/nds32/elf.h: New file.
11829 * config/nds32/linux.h: New file.
11830 * config/nds32/nds32-elf.opt: New file.
11831 * config/nds32/nds32-linux.opt: New file.
11832 * config/nds32/nds32-fp-as-gp.c
11833 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
11834 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
11836 (nds32_asm_file_end): Ditto.
11837 (nds32_print_operand): Ditto.
11838 (nds32_insert_attributes): Ditto.
11839 (nds32_init_libfuncs): New function.
11840 (TARGET_HAVE_TLS): Define.
11841 (TARGET_INIT_LIBFUNCS): Define.
11842 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
11844 (TARGET_ELF): Apply different mcmodel setting.
11845 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
11846 been migrated into elf.h and linux.h files.
11847 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
11848 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
11849 (mcmodel): The content has been migrated into nds32-elf.opt and
11850 nds32-linux.opt files.
11851 * config/nds32/t-elf: New file.
11852 * config/nds32/t-linux: New file.
11854 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
11855 Shiva Chen <shiva0217@gmail.com>
11857 * config/nds32/constants.md (unspec_volatile_element): Add
11858 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
11859 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
11861 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
11862 (make_pass_nds32_fp_as_gp): Declare.
11863 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
11865 (nds32_asm_function_end_prologue): Remove unused asm output.
11866 (nds32_asm_function_begin_epilogue): Remove unused asm output.
11867 (nds32_asm_file_start): Output necessary fp_as_gp information.
11868 (nds32_option_override): Adjust register usage.
11869 (nds32_expand_prologue): Consider fp_as_gp situation.
11870 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
11871 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
11874 (simple_return): Ditto.
11875 (omit_fp_begin): Output special directive for fp_as_gp.
11876 (omit_fp_end): Output special directive for fp_as_gp.
11877 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
11878 mforbid-fp-as-gp): New options.
11880 2018-06-01 Mark Wielaard <mark@klomp.org>
11882 * dwarf2out.c (dwarf2out_finish): Remove generation of
11883 DW_AT_loclists_base.
11885 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
11887 * gimple-ssa-store-merging.c: Include gimple-fold.h.
11888 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
11889 (struct merged_store_group): Add bit_insertion field.
11890 (dump_char_array): Use standard hexadecimal format.
11891 (merged_store_group::merged_store_group): Set bit_insertion to false.
11892 (merged_store_group::apply_stores): Use optimal buffer size. Deal
11893 with BIT_INSERT_EXPR stores. Move up code updating the mask and
11894 also print the mask in the dump file.
11895 (pass_store_merging::gate): Minor tweak.
11896 (imm_store_chain_info::coalesce_immediate): Fix wrong association
11897 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
11898 stores with INTEGER_CST stores.
11899 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
11900 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
11901 and use it throughout. Generate bit insertion sequences if need be.
11902 (pass_store_merging::process_store): Remove redundant condition.
11903 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
11905 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
11907 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
11908 the 128-bit floating point types. Fix function comment.
11910 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11912 * config/aarch64/aarch64-simd.md
11913 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
11915 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
11918 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11920 PR tree-optimization/85989
11921 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
11923 (backprop::intersect_uses): Check it when deciding whether this
11924 is a backedge reference.
11925 (backprop::process_block): Add each phi to m_visited_phis
11926 after visiting it, then clear it at the end.
11928 2018-06-01 Richard Biener <rguenther@suse.de>
11930 * tree-vectorizer.h (vect_dr_stmt): New function.
11931 (vect_get_load_cost): Adjust.
11932 (vect_get_store_cost): Likewise.
11933 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
11934 Use vect_dr_stmt instead of DR_SMTT.
11935 (vect_record_base_alignments): Likewise.
11936 (vect_calculate_target_alignment): Likewise.
11937 (vect_compute_data_ref_alignment): Likewise and make static.
11938 (vect_update_misalignment_for_peel): Likewise.
11939 (vect_verify_datarefs_alignment): Likewise.
11940 (vector_alignment_reachable_p): Likewise.
11941 (vect_get_data_access_cost): Likewise. Pass down
11942 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
11943 (vect_get_peeling_costs_all_drs): Likewise.
11944 (vect_peeling_hash_get_lowest_cost): Likewise.
11945 (vect_enhance_data_refs_alignment): Likewise.
11946 (vect_find_same_alignment_drs): Likewise.
11947 (vect_analyze_data_refs_alignment): Likewise.
11948 (vect_analyze_group_access_1): Likewise.
11949 (vect_analyze_group_access): Likewise.
11950 (vect_analyze_data_ref_access): Likewise.
11951 (vect_analyze_data_ref_accesses): Likewise.
11952 (vect_vfa_segment_size): Likewise.
11953 (vect_small_gap_p): Likewise.
11954 (vectorizable_with_step_bound_p): Likewise.
11955 (vect_prune_runtime_alias_test_list): Likewise.
11956 (vect_analyze_data_refs): Likewise.
11957 (vect_supportable_dr_alignment): Likewise.
11958 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11959 (vect_gen_prolog_loop_niters): Likewise.
11960 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11961 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
11963 (vect_recog_mask_conversion_pattern): Likewise.
11964 (vect_try_gather_scatter_pattern): Likewise.
11965 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
11966 to vect_get_store_cost.
11967 (vect_get_store_cost): Get stmt_info instead of DR.
11968 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
11969 (vect_get_load_cost): Get stmt_info instead of DR.
11971 2018-06-01 Richard Biener <rguenther@suse.de>
11973 PR middle-end/86017
11974 * gimple-fold.c (var_decl_component_p): Also allow offsetted
11975 vars wrapped in MEM_REFs.
11977 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
11979 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11980 Fix subreg tests so that we only return a choice between
11981 GENERAL_REGS and FP_REGS if the original classes included both.
11983 2018-06-01 Richard Biener <rguenther@suse.de>
11986 * tree-ssa-structalias.c (get_function_part_constraint):
11987 Handle NULL fi->decl.
11988 (find_func_aliases_for_call): Properly handle indirect
11989 fi from direct call.
11990 (find_func_clobbers): Likewise.
11991 (ipa_pta_execute): Likewise.
11992 (create_variable_info_for): For functions that are ifunc_resolver
11993 resolve to a varinfo that contains the result of the resolver call.
11994 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
11997 2018-05-31 Michael Collison <michael.collison@arm.com>
11999 * config/aarch64/aarch64.md:
12000 (*fix_to_zero_extenddfdi2): New pattern.
12001 * gcc.target/aarch64/fix_extend1.c: New testcase.
12003 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
12005 PR middle-end/78809
12006 PR middle-end/83026
12007 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
12008 and BUILT_IN_STRNCMP_EQ.
12009 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
12010 BUILT_IN_STRNCMP_EQ.
12011 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
12012 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12013 (gimple_fold_builtin): Likewise.
12014 * tree-ssa-strlen.c (compute_string_length): New function.
12015 (determine_min_obsize): New function.
12016 (handle_builtin_string_cmp): New function to handle calls to
12017 string compare functions.
12018 (strlen_optimize_stmt): Add handling to builtin string compare
12020 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
12021 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12022 * tree.c (build_common_builtin_nodes): Add new defines of
12023 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
12025 2018-05-31 Jakub Jelinek <jakub@redhat.com>
12028 * bb-reorder.c (pass_partition_blocks::gate): Return false for
12029 functions with naked attribute.
12031 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
12033 * config/i386/sse.md (avx_vec_concat<mode>):
12034 Substitute concat_tg_mode mode attribute with xtg_mode.
12035 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
12036 (concat_tg_mode): Remove mode attribute.
12038 2018-05-31 Martin Sebor <msebor@redhat.com>
12041 * calls.c (alloc_max_size): Correct a logic error/typo.
12042 Treat excessive arguments as infinite. Warn for invalid arguments.
12043 * doc/invoke.texi (-Walloc-size-larger-than): Update.
12045 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
12048 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
12049 and movx for Haswell.
12051 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
12052 Cesar Philippidis <cesar@codesourcery.com>
12054 PR middle-end/85879
12055 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
12056 when emitting error on private/firstprivate reductions.
12057 * omp-low.c (lower_omp_target): Avoid reference-type processing
12058 on pointers for firstprivate clause.
12060 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12062 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
12065 * config/aarch64/aarch64-simd.md
12066 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
12067 (aarch64_ld1_x3_<mode>): Likewise
12068 (aarch64_st1x2<VALLDIF:mode>): Likewise
12069 (aarch64_st1_x2_<mode>): Likewise
12070 (aarch64_st1x3<VALLDIF:mode>): Likewise
12071 (aarch64_st1_x3_<mode>): Likewise
12072 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
12073 (vld1_s8_x3): Likewise.
12074 (vld1_u16_x3): Likewise.
12075 (vld1_s16_x3): Likewise.
12076 (vld1_u32_x3): Likewise.
12077 (vld1_s32_x3): Likewise.
12078 (vld1_u64_x3): Likewise.
12079 (vld1_s64_x3): Likewise.
12080 (vld1_f16_x3): Likewise.
12081 (vld1_f32_x3): Likewise.
12082 (vld1_f64_x3): Likewise.
12083 (vld1_p8_x3): Likewise.
12084 (vld1_p16_x3): Likewise.
12085 (vld1_p64_x3): Likewise.
12086 (vld1q_u8_x3): Likewise.
12087 (vld1q_s8_x3): Likewise.
12088 (vld1q_u16_x3): Likewise.
12089 (vld1q_s16_x3): Likewise.
12090 (vld1q_u32_x3): Likewise.
12091 (vld1q_s32_x3): Likewise.
12092 (vld1q_u64_x3): Likewise.
12093 (vld1q_s64_x3): Likewise.
12094 (vld1q_f16_x3): Likewise.
12095 (vld1q_f32_x3): Likewise.
12096 (vld1q_f64_x3): Likewise.
12097 (vld1q_p8_x3): Likewise.
12098 (vld1q_p16_x3): Likewise.
12099 (vld1q_p64_x3): Likewise.
12100 (vst1_s64_x2): Likewise.
12101 (vst1_u64_x2): Likewise.
12102 (vst1_f64_x2): Likewise.
12103 (vst1_s8_x2): Likewise.
12104 (vst1_p8_x2): Likewise.
12105 (vst1_s16_x2): Likewise.
12106 (vst1_p16_x2): Likewise.
12107 (vst1_s32_x2): Likewise.
12108 (vst1_u8_x2): Likewise.
12109 (vst1_u16_x2): Likewise.
12110 (vst1_u32_x2): Likewise.
12111 (vst1_f16_x2): Likewise.
12112 (vst1_f32_x2): Likewise.
12113 (vst1_p64_x2): Likewise.
12114 (vst1q_s8_x2): Likewise.
12115 (vst1q_p8_x2): Likewise.
12116 (vst1q_s16_x2): Likewise.
12117 (vst1q_p16_x2): Likewise.
12118 (vst1q_s32_x2): Likewise.
12119 (vst1q_s64_x2): Likewise.
12120 (vst1q_u8_x2): Likewise.
12121 (vst1q_u16_x2): Likewise.
12122 (vst1q_u32_x2): Likewise.
12123 (vst1q_u64_x2): Likewise.
12124 (vst1q_f16_x2): Likewise.
12125 (vst1q_f32_x2): Likewise.
12126 (vst1q_f64_x2): Likewise.
12127 (vst1q_p64_x2): Likewise.
12128 (vst1_s64_x3): Likewise.
12129 (vst1_u64_x3): Likewise.
12130 (vst1_f64_x3): Likewise.
12131 (vst1_s8_x3): Likewise.
12132 (vst1_p8_x3): Likewise.
12133 (vst1_s16_x3): Likewise.
12134 (vst1_p16_x3): Likewise.
12135 (vst1_s32_x3): Likewise.
12136 (vst1_u8_x3): Likewise.
12137 (vst1_u16_x3): Likewise.
12138 (vst1_u32_x3): Likewise.
12139 (vst1_f16_x3): Likewise.
12140 (vst1_f32_x3): Likewise.
12141 (vst1_p64_x3): Likewise.
12142 (vst1q_s8_x3): Likewise.
12143 (vst1q_p8_x3): Likewise.
12144 (vst1q_s16_x3): Likewise.
12145 (vst1q_p16_x3): Likewise.
12146 (vst1q_s32_x3): Likewise.
12147 (vst1q_s64_x3): Likewise.
12148 (vst1q_u8_x3): Likewise.
12149 (vst1q_u16_x3): Likewise.
12150 (vst1q_u32_x3): Likewise.
12151 (vst1q_u64_x3): Likewise.
12152 (vst1q_f16_x3): Likewise.
12153 (vst1q_f32_x3): Likewise.
12154 (vst1q_f64_x3): Likewise.
12155 (vst1q_p64_x3): Likewise.
12157 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12159 * config/msp430/msp430.c (msp430_output_labelref): Prepend
12160 user_label_prefix to name.
12162 * tree-core.h: Update comment about the format of NAME string
12163 passed to handler in attribute_spec.
12165 * config/msp430/msp430.md: Remove erroneous subreg expression from
12166 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
12167 zero_extend{q,h}isi2.
12169 2018-05-30 Borislav Petkov <bp@suse.de>
12171 * doc/extend.texi: Document some architecture specific
12172 constraints and sort entries.
12174 2018-05-30 Martin Sebor <msebor@redhat.com>
12176 PR middle-end/85369
12177 * builtins.c (expand_builtin_stpcpy_1): New function.
12178 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
12179 only if the former succeeds.
12181 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12183 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
12186 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12188 * doc/invoke.texi (-flinker-output): Document
12190 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12192 * passes.c (ipa_write_summaries): Only modify statements if body
12194 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
12195 incrementally linking.
12196 (ipa_passes): Likewise.
12197 * lto-cgraph.c (lto_output_node): When incrementally linking do not
12198 pass down resolution info.
12199 * common.opt (flag_incremental_link): Update info.
12200 * gcc.c (plugin specs): Turn flinker-output=* to
12201 -plugin-opt=-linker-output-known
12202 * toplev.c (compile_file): Also cut compilation when doing incremental
12204 * flag-types. (enum lto_partition_model): Add
12205 LTO_LINKER_OUTPUT_NOLTOREL.
12206 (invoke.texi): Add -flinker-output docs.
12207 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
12208 link same way as WPA; do not stream in dead initializers.
12210 * dwarf2out.c (dwarf2out_die_ref_for_decl,
12211 darf2out_register_external_decl): Support incremental link.
12213 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12215 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
12217 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12219 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
12220 it down to simple_object_copy_lto_debug_sections.
12221 (run_gcc): Determine incremental LTO link time and configure
12222 lto1 into non-wpa mode, disable renaming of debug sections.
12224 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
12226 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12227 descriptions of various incorrectly documented functions.
12229 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
12232 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12233 address check not strict.
12235 2018-05-30 Richard Biener <rguenther@suse.de>
12237 PR tree-optimization/85964
12238 * tracer.c (better_p): Drop initialized count check, we only
12239 call the function with initialized counts now.
12240 (find_best_successor): Do find a best edge if one
12241 has uninitialized count.
12242 (find_best_predecessor): Likewise. Do BB frequency check only
12243 if count is initialized.
12245 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
12247 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
12248 (aarch64_ldrstr_offset_compare): New.
12249 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
12250 load/store orderings.
12251 (aarch64_gen_adjusted_ldpstp): Likewise.
12253 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
12255 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12256 Check for subset of GENERAL_REGS and FP_REGS.
12257 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
12260 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
12262 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
12263 and wi::to_poly_offset. Add the current offset and then check
12264 whether the sum fits, rather than using an unchecked addition of
12265 a checked term. Check for a shwi rather than a uhwi.
12266 * expr.c (get_bit_range): Use tree_to_poly_uint64.
12267 (store_constructor): Use poly_int_tree_p.
12268 (expand_expr_real_1): Likewise.
12269 * function.c (assign_temp): Likewise.
12270 * fold-const.c (const_binop): Use poly_int_tree_p and
12271 wi::to_poly_offset.
12272 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
12274 * ipa-icf-gimple.c (func_checker::compare_operand): Use
12275 to_poly_offset for MEM offsets.
12276 * ipa-icf.c (sem_variable::equals): Likewise.
12277 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
12278 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
12279 wi::to_poly_offset for BIT_FIELD_REF offsets.
12280 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
12281 wi::to_poly_offset.
12282 * var-tracking.c (emit_note_insn_var_location): Use
12283 tree_to_poly_uint64.
12285 2018-05-29 Jim Wilson <jimw@sifive.com>
12287 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
12289 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
12292 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
12293 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
12295 (sse4_1_round<mode>2): Use nonimmediate_operand
12296 for operand 1 predicate.
12298 2018-05-29 Martin Sebor <msebor@redhat.com>
12299 Richard Biener <rguenther@suse.de>
12302 * calls.c (get_size_range): Call determine_value_range instead
12303 of get_value_range..
12304 * tree-vrp.h (determine_value_range): Declared new function.
12305 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
12307 2018-05-29 Richard Biener <rguenther@suse.de>
12309 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
12310 sure to use non-pattern stmts for get_earlier_stmt arguments.
12311 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
12312 called on pattern stmts.
12313 (get_later_stmt): Likewise.
12315 2018-05-29 Martin Liska <mliska@suse.cz>
12317 PR gcov-profile/85759
12318 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
12321 2018-05-29 Jakub Jelinek <jakub@redhat.com>
12323 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
12325 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
12329 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12330 VEC_PACK_FLOAT_EXPR): New tree codes.
12331 * tree-pretty-print.c (op_code_prio): Handle
12332 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12333 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12334 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12335 * tree-inline.c (estimate_operator_cost): Likewise.
12336 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
12337 * fold-const.c (const_binop): Likewise.
12338 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
12339 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12340 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12341 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
12342 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12343 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12344 * expr.c (expand_expr_real_2): Likewise.
12345 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
12346 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
12347 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
12349 * optabs.c (expand_widen_pattern_expr): For
12350 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
12351 sign from result type rather than operand's type.
12352 (expand_binop_directly): For vec_packu_float_optab and
12353 vec_packs_float_optab allow result type to be different from operand's
12355 * optabs-tree.c (optab_for_tree_code): Handle
12356 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12357 VEC_PACK_FLOAT_EXPR. Formatting fixes.
12358 * tree-vect-generic.c (expand_vector_operations_1): Handle
12359 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12360 VEC_PACK_FLOAT_EXPR.
12361 * tree-vect-stmts.c (supportable_widening_operation): Handle
12363 (supportable_narrowing_operation): Handle FLOAT_EXPR.
12364 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
12365 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
12366 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
12367 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
12369 (vec_pack<floatprefix>_float_<mode>): New expander.
12370 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
12372 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
12373 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
12374 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
12375 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
12376 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
12378 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
12379 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
12380 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12381 VEC_PACK_FLOAT_EXPR): Document.
12383 2018-05-29 Richard Biener <rguenther@suse.de>
12385 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
12387 (stmt_vec_info_vec): Make pointer.
12388 (init_stmt_vec_info_vec): Remove.
12389 (free_stmt_vec_info_vec): Likewise.
12390 (set_stmt_vec_info_vec): New function.
12391 (free_stmt_vec_infos): Likewise.
12392 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
12393 (set_vinfo_for_stmt): Likewise.
12394 (get_earlier_stmt): Likewise.
12395 (get_later_stmt): Likewise.
12396 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
12397 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
12398 (vec_info::~vec_info): Free stmt_vec_infos.
12399 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
12400 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
12401 (pass_slp_vectorize::execute): Likewise.
12402 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
12403 (free_stmt_vec_info_vec): Likewise.
12404 (set_stmt_vec_info_vec): New function.
12405 (free_stmt_vec_infos): Likewise.
12406 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
12407 the global stmt_vec_info_vec.
12408 * tree-parloops.c (gather_scalar_reductions): Use
12409 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
12412 2018-05-29 Richard Biener <rguenther@suse.de>
12414 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
12416 2018-05-29 Martin Liska <mliska@suse.cz>
12417 David Malcolm <dmalcolm@redhat.com>
12419 * vec.c (test_reverse): New.
12420 (vec_c_tests): Add new test.
12421 * vec.h (vl_ptr>::reverse): New function.
12423 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
12425 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
12427 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
12430 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
12432 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
12434 2018-05-28 Richard Biener <rguenther@suse.de>
12436 PR tree-optimization/85933
12437 * tree-vect-data-refs.c (vect_record_base_alignments): Only
12438 look at stmts marked as vectorizable.
12440 2018-05-28 Richard Biener <rguenther@suse.de>
12442 PR tree-optimization/85934
12443 * tree-vect-generic.c (expand_vector_operations_1): Hoist
12444 vector boolean check before scalar optimization.
12446 2018-05-28 Jakub Jelinek <jakub@redhat.com>
12448 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
12451 2018-05-28 Mark Wielaard <mark@klomp.org>
12453 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
12454 if it is an expression containing a minus sign.
12456 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
12458 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
12460 2018-05-27 Paul Koning <ni1d@arrl.net>
12462 * config/pdp11/pdp11.md (truncsihi2): Remove.
12464 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
12465 Chung-Ju Wu <jasonwucj@gmail.com>
12467 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
12469 (unaligned_store_dw): Ditto.
12470 * config/nds32/nds32-memory-manipulation.c
12471 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
12472 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
12473 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
12474 (emit_setmem_word_loop): Rename to ...
12475 (emit_setmem_doubleword_loop): ... this.
12476 (nds32_gen_dup_4_byte_to_word_value): New function.
12477 (nds32_gen_dup_8_byte_to_double_word_value): New function.
12478 (nds32_expand_setmem_loop): Refine implementation.
12479 (nds32_expand_setmem_loop_v3m): Ditto.
12480 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
12483 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12485 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
12487 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12489 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
12490 (nds32_init_machine_status): Initialize machine->attr_naked_p and
12491 machine->attr_no_prologue_p.
12492 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
12493 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
12494 (nds32_expand_epilogue): Consider attr_naked_p.
12495 (nds32_expand_epilogue_v3pop): Likewise.
12496 (nds32_can_use_return_insn): Likewise.
12497 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
12498 attr_no_prologue_p fields.
12499 * config/nds32/nds32.opt (mret-in-naked-func): New option.
12501 2018-05-27 Jakub Jelinek <jakub@redhat.com>
12504 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
12506 * config/i386/sse.md
12507 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
12509 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
12511 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
12513 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
12515 (*<floatsuffix>floatv2div2sf2): Rename to ...
12516 (*float<floatunssuffix>v2div2sf2): ... this.
12517 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
12518 (float<floatunssuffix>v2div2sf2_mask): ... this.
12519 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
12520 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
12521 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
12523 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
12525 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12527 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12529 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12531 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12533 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
12534 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
12535 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
12536 gen_ufix_truncv8dfv8si2.
12537 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
12538 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
12539 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
12540 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
12541 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
12542 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
12543 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
12544 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
12546 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12550 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
12552 2018-05-25 Jim Wilson <jimw@sifive.com>
12554 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
12555 * config/riscv/riscv.c (struct machine_function): Add
12556 interrupt_handler_p and attribute_checked_p fields.
12557 (riscv_attribute_table): Add interrupt.
12558 (riscv_interrupt_type_p): New.
12559 (riscv_save_reg_p): Save extra regs for interrupt handler.
12560 (riscv_use_save_libcall): Return false for interrupt handler.
12561 (riscv_first_stack_step): Add forward declaration.
12562 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
12563 for interrupt handler with large frame. Use it for saved reg list.
12564 (riscv_expand_prologue): Move flag_stack_usage_info support to
12565 eliminate duplication.
12566 (riscv_expand_epilogue): Generate mret for interrupt handler.
12567 (riscv_epilogue_uses): New.
12568 (riscv_can_use_return_insn): Return false for interrupt handler.
12569 (riscv_function_ok_for_sibcall): Likewise.
12570 (riscv_set_current_function): Add interrupt handler support.
12571 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
12572 * config/riscv/riscv.md (UNSPECV_MRET): New.
12574 (riscv_frflags, riscv_fsflags): Use tab after opcode.
12576 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
12578 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
12580 PR tree-optimization/85712
12581 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
12582 this candidate has already been replaced in-situ by a copy.
12584 2018-05-25 Jason Merrill <jason@redhat.com>
12586 PR c++/80485 - inline function non-zero address.
12587 * symtab.c (nonzero_address): Check DECL_COMDAT.
12589 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
12592 * config/alpha/alpha.md (ashlsi3): New insn pattern.
12593 (*ashlsi_se): Rename from *ashldi_se. Define as sign
12594 extension of SImode operation. Use const123_operand predicate.
12595 (*saddsi_1): Remove.
12596 (*saddl_se_1): Ditto.
12597 (*ssubsi_1): Ditto.
12598 (*ssubl_se_1): Ditto.
12599 * config/alpha/predicates.md (const123_operand): New predicate.
12600 * config/alpha/constraints.md (P): Use IN_RANGE.
12602 2018-05-25 Richard Biener <rguenther@suse.de>
12604 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
12606 (ref_maybe_used_by_stmt_p): Likewise.
12607 (stmt_may_clobber_ref_p): Likewise.
12608 (stmt_may_clobber_ref_p_1): Likewise.
12609 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
12611 (ref_maybe_used_by_stmt_p): Likewise.
12612 (stmt_may_clobber_ref_p): Likewise.
12613 (stmt_may_clobber_ref_p_1): Likewise.
12614 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
12615 the alias oracle to disambiguate DRs with stmts DR analysis
12617 (vect_analyze_data_refs): Do not give up on not analyzable
12618 DRs for BB vectorization. Remove code truncating the dataref
12621 2018-05-25 Jakub Jelinek <jakub@redhat.com>
12624 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12625 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
12626 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
12628 2018-05-25 Richard Biener <rguenther@suse.de>
12630 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
12631 function, combining stmt data ref gathering and fatal analysis
12633 (vect_analyze_data_refs): Remove now redudnant code and simplify.
12634 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
12635 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
12636 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
12637 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
12639 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12641 PR tree-optimization/85720
12642 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
12643 SCC if all partitions are builtins.
12644 (version_loop_by_alias_check): New parameter. Generate cancelable
12645 runtime alias check if all partitions are builtins.
12646 (distribute_loop): Update call to above function.
12648 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12650 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
12651 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
12652 (parm_default_def_partition_arg): Ditto.
12653 (set_parm_default_def_partition): Ditto.
12654 (get_parm_default_def_partitions): Ditto and make it static.
12655 (get_undefined_value_partitions): Ditto and make it static.
12656 (remove_ssa_form): Refactor call to init_var_map here.
12657 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
12658 computation for loop region.
12659 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
12660 (register_default_def): Delete.
12661 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
12662 (parm_default_def_partition_arg): Ditto.
12663 (set_parm_default_def_partition): Ditto.
12664 (get_parm_default_def_partitions): Ditto and make it static.
12665 (get_undefined_value_partitions): Ditto and make it static.
12666 (coalesce_with_default, coalesce_with_default): Update comment.
12667 (create_coalesce_list_for_region): New func factored out from
12668 create_outofssa_var_map.
12669 (populate_coalesce_list_for_outofssa): New func factored out from
12670 create_outofssa_var_map and coalesce_ssa_name.
12671 (create_outofssa_var_map): Delete.
12672 (coalesce_ssa_name): Refactor to support live range computation.
12673 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
12674 (get_parm_default_def_partitions): Delete.
12675 (get_undefined_value_partitions): Ditto.
12676 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
12677 computation for loop region.
12678 (new_tree_live_info, loe_visit_block): Ditto.
12679 (live_worklist, set_var_live_on_entry): Ditto.
12680 (calculate_live_on_exit, verify_live_on_entry): Ditto.
12681 * tree-ssa-live.h (struct _var_map): New fields.
12682 (init_var_map): Change decl.
12683 (region_contains_p): New.
12685 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12687 * tree-ssa-live.h (live_merge_and_clear): Delete.
12689 2018-05-25 Richard Biener <rguenther@suse.de>
12692 * tree-dump.c (dequeue_and_dump): Remove access to removed
12693 operand 2 of a SWITCH_EXPR.
12695 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12697 * doc/sourcebuild.texi (vect_double_cond_arith): Include
12698 multiplication and division.
12699 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
12700 (cond_udiv@var{m}, cond_umod@var{m}): Document.
12701 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
12702 (cond_udiv_optab, cond_umod_optab): New optabs.
12703 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
12704 (IFN_COND_RDIV): New internal functions.
12705 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
12706 TRUNC_MOD_EXPR and RDIV_EXPR.
12707 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
12708 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
12710 (SVE_INT_BINARY): Include mult.
12711 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
12712 (optab, sve_int_op): Handle mult.
12713 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
12715 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
12716 for SVE_INT_BINARY_SD.
12718 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12720 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
12721 (optab, sve_int_op): Handle div and udiv.
12722 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
12723 for SVE_INT_BINARY_SD.
12724 (*<optab><mode>3): New insn for the same.
12726 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12728 * tree-vect-patterns.c: Include predict.h.
12729 (vect_recog_divmod_pattern): Restrict check for division support
12730 to when optimizing for size.
12732 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12734 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
12735 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
12736 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
12737 (gimple_match_op::set_op): Likewise.
12738 (gimple_resimplify4): Declare.
12739 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
12740 (expr::gen_transform): Likewise.
12741 (decision_tree::gen): Generate a simplification routine for 4 operands.
12742 * gimple-match-head.c (gimple_simplify): Add an overload for
12743 4 operands. In the top-level function, handle up to 4 call
12744 arguments and call gimple_resimplify4.
12745 (gimple_resimplify4): New function.
12746 (build_call_internal): Pass a fourth operand.
12747 (maybe_push_to_seq): Likewise.
12748 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
12749 Fold VEC_COND_EXPRs of an operation and a default value into
12750 an IFN_COND_* function if possible.
12751 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
12753 (SVE_COND_FP_BINARY): Include them.
12754 (optab, sve_fp_op): Handle them.
12755 (SVE_INT_BINARY_REV): New code iterator.
12756 (SVE_COND_FP_BINARY_REV): New int iterator.
12757 (commutative): New int attribute.
12758 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
12760 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
12762 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
12763 (*cond_<optab><mode>): New patterns for reversed operands.
12765 2018-05-25 Richard Biener <rguenther@suse.de>
12767 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12768 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12769 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12770 (STMT_VINFO_GROUPED_ACCESS): Adjust.
12771 * tree-vect-data-refs.c (everywhere): Adjust users.
12772 * tree-vect-loop.c (everywhere): Likewise.
12773 * tree-vect-slp.c (everywhere): Likewise.
12774 * tree-vect-stmts.c (everywhere): Likewise.
12775 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12777 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12779 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12781 (gcc_cv_as_section_exclude): ... this.
12782 Try Solaris as #exclude syntax.
12783 * configure: Regenerate.
12784 * config.in: Regenerate.
12785 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12787 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12788 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12790 * varasm.c (default_elf_asm_named_section): Don't check if
12791 HAVE_GAS_SECTION_EXCLUDE is defined.
12793 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12795 * doc/md.texi: Update the documentation of the cond_* optabs
12796 to mention the new final operand. Fix GET_MODE_NUNITS call.
12797 Describe the scalar case too.
12798 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
12799 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
12801 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
12802 (get_conditional_internal_fn): Update comment.
12803 * tree-vect-loop.c (vectorizable_reduction): Pass the original
12804 accumulator value as a final argument to conditional functions.
12805 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
12806 a define_expand and add an "else" operand. Assert for now that
12807 the else operand is equal to operand 2. Use SVE_INT_BINARY and
12808 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
12809 (*cond_<optab><mode>): New patterns.
12810 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
12811 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12812 (UNSPEC_COND_EOR): Delete.
12813 (optab): Remove associated mappings.
12814 (SVE_INT_BINARY): New code iterator.
12815 (sve_int_op): Remove int attribute and add "minus" to the code
12817 (SVE_COND_INT_OP): Delete.
12818 (SVE_COND_FP_OP): Rename to...
12819 (SVE_COND_FP_BINARY): ...this.
12821 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12823 * optabs.c (can_reuse_operands_p): New function.
12824 (maybe_legitimize_operands): Try to reuse the results for
12827 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12829 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
12830 Add {q} suffix to insn mnemonic.
12832 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12834 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
12835 (msp430_warn_func_return): New.
12837 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
12839 * fold-const.c (tree_nonzero_bits): New function.
12840 * fold-const.h (tree_nonzero_bits): Likewise.
12841 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
12842 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
12844 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12848 * varasm.c (assemble_alias): Check ifunc_resolver only on
12851 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
12854 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
12855 when memory input operand is handled.
12857 2018-05-24 Luis Machado <luis.machado@linaro.org>
12859 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
12861 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
12863 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12865 * match.pd: Delay FMA folds until after vectorization.
12867 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
12870 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12871 address check not strict.
12873 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
12875 * gimple-match.h (gimple_match_op): New class.
12876 (mprts_hook): Replace parameters with a gimple_match_op *.
12877 (maybe_build_generic_op): Likewise.
12878 (gimple_simplified_result_is_gimple_val): Replace parameters with
12879 a const gimple_match_op *.
12880 (gimple_simplify): Replace code_helper * and tree * parameters with
12881 a gimple_match_op * parameter.
12882 (gimple_resimplify1): Replace code_helper *, tree and tree *
12883 parameters with a gimple_match_op * parameter.
12884 (gimple_resimplify2): Likewise.
12885 (gimple_resimplify3): Likewise.
12886 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
12887 parameters with a gimple_match_op * parameter.
12888 * gimple-match-head.c (gimple_simplify): Change prototypes of
12889 auto-generated functions to take a gimple_match_op * instead of
12890 separate code_helper * and tree * parameters. Make the same
12891 change in the top-level overload and update calls to the
12892 gimple_resimplify routines. Update calls to the auto-generated
12893 functions and to maybe_push_res_to_seq in the publicly-facing
12894 operation-specific gimple_simplify overloads.
12895 (gimple_match_op::MAX_NUM_OPS): Define.
12896 (gimple_resimplify1): Replace rcode and ops with a single res_op
12897 parameter. Update call to gimple_simplify.
12898 (gimple_resimplify2): Likewise.
12899 (gimple_resimplify3): Likewise.
12900 (mprts_hook): Replace parameters with a gimple_match_op *.
12901 (maybe_build_generic_op): Likewise.
12902 (build_call_internal): Replace type, nargs and ops with
12903 a gimple_match_op *.
12904 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
12905 with a single gimple_match_op *. Update calls to mprts_hook,
12906 build_call_internal and gimple_simplified_result_is_gimple_val.
12907 Factor out code that is common to the tree_code and combined_fn cases.
12908 * genmatch.c (expr::gen_transform): Replace tem_code and
12909 tem_ops with a gimple_match_op called tem_op. Update calls
12910 to the gimple_resimplify functions and maybe_push_res_to_seq.
12911 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
12912 res_ops. Update call to the gimple_resimplify functions.
12913 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
12914 (decision_tree::gen): Make the functions take a gimple_match_op *
12915 called res_op instead of separate res_code and res_ops parameters.
12916 Update call accordingly.
12917 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
12918 and ops with a single res_op parameter. Update calls to
12919 maybe_build_generic_op and maybe_push_res_to_seq.
12920 (fold_stmt_1): Update calls to gimple_simplify and
12921 replace_stmt_with_simplification.
12922 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
12923 and gimple_simplified_result_is_gimple_val.
12924 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
12926 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
12927 with a gimple_match_op *.
12928 (vn_nary_build_or_lookup): Likewise. Update call to
12929 vn_nary_build_or_lookup_1.
12930 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
12931 gimple_match_op *. Update calls to the gimple_resimplify routines
12932 and to gimple_simplified_result_is_gimple_val.
12933 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
12934 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
12935 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
12936 (visit_nary_op): Likewise.
12937 (visit_reference_op_load): Likewise.
12939 2018-05-23 Luis Machado <luis.machado@linaro.org>
12941 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
12942 modifier for printing the step amount.
12944 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
12947 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
12950 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
12952 * doc/sourcebuild.texi (Endianness): New subsubsection.
12954 2018-05-23 Luis Machado <luis.machado@linaro.org>
12956 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12957 <prefetch_dynamic_strides>: New const bool field.
12958 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12959 prefetch_dynamic_strides.
12960 (exynosm1_prefetch_tune): Likewise.
12961 (thunderxt88_prefetch_tune): Likewise.
12962 (thunderx_prefetch_tune): Likewise.
12963 (thunderx2t99_prefetch_tune): Likewise.
12964 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
12966 (aarch64_override_options_internal): Update to set
12967 PARAM_PREFETCH_DYNAMIC_STRIDES.
12968 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12969 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12970 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12971 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12972 prefetch-dynamic-strides setting.
12974 2018-05-23 Luis Machado <luis.machado@linaro.org>
12976 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12977 <minimum_stride>: New const int field.
12978 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12979 minimum_stride field defaulting to -1.
12980 (exynosm1_prefetch_tune): Likewise.
12981 (thunderxt88_prefetch_tune): Likewise.
12982 (thunderx_prefetch_tune): Likewise.
12983 (thunderx2t99_prefetch_tune): Likewise.
12984 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
12985 <default_opt_level>: Set to 3.
12986 (aarch64_override_options_internal): Update to set
12987 PARAM_PREFETCH_MINIMUM_STRIDE.
12988 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12989 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12990 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12991 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12992 stride is constant and is below the minimum stride threshold.
12994 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12996 * config/arm/arm-cpus.in (mode26): Delete.
12997 (armv4): Delete mode26 reference.
12998 * config/arm/arm.c (arm_configure_build_target): Delete use of
13001 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
13003 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
13005 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
13006 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
13008 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
13009 (floatunsdidf2): Ditto.
13011 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
13013 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
13014 (fixuns_trunc<mode>si2_avx512f): Ditto.
13015 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
13016 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
13017 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
13019 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
13021 PR rtl-optimization/79985
13022 * df-scan.c (df_insn_refs_collect): Remove special case for
13023 global registers and asm statements.
13025 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
13027 * extend.texi (Global Register Variables): Rewrite the bullet list.
13028 Note that the register is available for allocation. Note that access
13029 via inline asm must use constraints. Add note about async-signal
13030 handlers. Remove paragraph about automagic register selection.
13032 2018-05-23 Richard Biener <rguenther@suse.de>
13034 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
13035 of fixed offset from memset VN.
13037 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
13039 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
13040 first_interp field.
13041 (alloc_cand_and_find_basis): Initialize first_interp field.
13042 (slsr_process_mul): Modify first_interp field.
13043 (slsr_process_add): Likewise.
13044 (slsr_process_cast): Modify first_interp field for each new
13046 (slsr_process_copy): Likewise.
13047 (dump_candidate): Dump first_interp field.
13048 (replace_mult_candidate): Process all interpretations, not just
13050 (replace_rhs_if_not_dup): Likewise.
13051 (replace_one_candidate): Likewise.
13053 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
13055 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
13057 (aarch64_needs_frame_chain): New function.
13058 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
13060 2018-05-23 Sudakshina Das <sudi.das@arm.com>
13063 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
13064 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
13065 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
13066 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
13067 as true for strict-align.
13068 (aarch64_can_inline_p): Perform checks even when callee has no
13069 attributes to check for strict alignment.
13070 * doc/extend.texi (AArch64 Function Attributes): Document
13072 * doc/invoke.texi: (AArch64 Options): Likewise.
13074 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
13076 PR tree-optimization/85853
13077 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
13078 the handling of the root of the node to...
13079 (vect_slp_analyze_node_operations_1): ...this new function,
13080 and run the whole thing with the child nodes' def types
13081 set according to their SLP node's def type.
13083 2018-05-23 Richard Biener <rguenther@suse.de>
13085 PR middle-end/85874
13086 * tree-data-ref.c (create_runtime_alias_checks): Defer
13087 and ignore overflow warnings.
13089 2018-05-23 Yury Gribov <tetra2005@gmail.com>
13091 PR tree-optimization/85822
13092 * tree-vrp.c (is_masked_range_test): Fix handling of negative
13095 2018-05-23 Richard Biener <rguenther@suse.de>
13097 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13098 memset constants via native_interpret_expr.
13100 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
13103 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
13105 (cgraph_node::create_alias): Likewise.
13106 (cgraph_node::get_availability): Check ifunc_resolver instead
13107 of looking up ifunc attribute.
13108 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
13109 * varasm.c (do_assemble_alias): Likewise.
13110 (assemble_alias): Likewise.
13111 (default_binds_local_p_3): Likewise.
13112 * cgraph.h (cgraph_node): Add ifunc_resolver.
13113 (cgraph_node::only_called_directly_or_aliased_p): Return false
13114 for IFUNC resolver.
13115 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
13117 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
13118 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
13119 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
13120 instead of looking up ifunc attribute.
13122 2018-05-22 Luis Machado <luis.machado@linaro.org>
13124 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
13126 2018-05-22 Martin Sebor <msebor@redhat.com>
13128 PR middle-end/85359
13129 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
13130 only when expasion succeeds.
13131 (expand_builtin_strcmp): Same.
13132 (expand_builtin_strncmp): Same.
13134 2018-05-22 Martin Sebor <msebor@redhat.com>
13136 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
13138 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
13139 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13141 * config/aarch64/aarch64-ldpstp.md: Replace uses of
13142 aarch64_mem_pair_operand with memory_operand and delete operand swapping
13144 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13145 Add check for legitimate_address.
13146 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
13147 (aarch64_swap_ldrstr_operands): New.
13148 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
13151 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
13152 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13154 * config/aarch64/aarch64.md: New patterns to generate stp
13156 (store_pair_sw, store_pair_dw): New patterns to generate stp for
13157 single words and double words.
13158 (load_pair_sw, load_pair_dw): Likewise.
13159 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
13161 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
13163 * config/aarch64/aarch64-ldpstp.md: Modify peephole
13164 for different mode ldpstp and add peephole for merged zero stores.
13165 Likewise for loads.
13166 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13168 (aarch64_gen_store_pair): Rename calls to match new patterns.
13169 (aarch64_gen_load_pair): Rename calls to match new patterns.
13170 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
13171 (load_pair<DREG:mode><DREG2:mode>): ... This.
13172 (store_pair<mode>): Rename to...
13173 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
13174 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
13175 New mode iterators.
13176 (V_INT_EQUIV): Handle SImode.
13177 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
13180 2018-05-22 Martin Sebor <msebor@redhat.com>
13183 * calls.c (maybe_warn_nonstring_arg): Use string length to set
13184 or ajust the presumed bound on an operation to avoid unnecessary
13187 2018-05-22 Martin Sebor <msebor@redhat.com>
13189 PR tree-optimization/85826
13190 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
13191 assuming that a DECL necesarily has a constant size.
13193 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
13195 PR middle-end/85862
13196 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
13198 2018-05-22 Richard Biener <rguenther@suse.de>
13200 PR tree-optimization/85834
13201 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
13202 non-constant and non-zero memset arguments.
13204 2018-05-22 Martin Liska <mliska@suse.cz>
13207 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
13209 2018-05-22 Richard Biener <rguenther@suse.de>
13211 PR tree-optimization/85863
13212 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
13213 comparisons when vectype is specified.
13214 (vectorizable_condition): Do not specify vectype for
13215 vect_is_simple_cond when SLP vectorizing.
13217 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
13220 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
13221 define __ibm128 as long double.
13222 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
13223 as a distinct type when IEEE 128-bit support is enabled.
13224 (init_float128_ieee): Fix up conversions between IFmode and IEEE
13225 128-bit types to use the correct functions.
13226 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
13227 convert between 128-bit floating point types that have different
13228 modes but the same representation, instead of using gen_lowpart to
13230 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
13232 (IFKF_reg): New attributes to give the register constraints for
13234 (extend<mode>tf2_internal): New insns to mark an explicit
13235 conversion between 128-bit floating point types that have a
13236 different mode but share the same representation.
13238 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
13240 PR tree-optimization/85814
13241 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
13242 a null return from get_strinfo when unsharing the next
13243 strinfo in the chain.
13245 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
13248 * varasm.c (weak_finish): Clean up weak_decls.
13250 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13252 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
13253 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
13254 UNSPEC_UADALP values.
13255 * config/aarch64/iterators.md (ABAL): New int iterator.
13258 (sur): Add mappings for the above.
13259 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
13261 (aarch64_<sur>abal<mode>_4): Likewise.
13262 (aarch64_<sur>adalp<mode>_3): Likewise.
13263 (<sur>sadv16qi): New define_expand.
13265 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
13267 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
13268 (*movdf_internal): Ditto.
13269 (*rcpsf2_sse): Ditto.
13270 (*rsqrtsf2_sse): Ditto.
13271 (*sqrt<mode>2_sse): Ditto.
13273 2018-05-21 Tamar Christina <tamar.christina@arm.com>
13275 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
13277 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
13278 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
13279 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
13280 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
13282 * config/aarch64/arm_neon.h: Likewise.
13283 * config/aarch64/iterators.md (VQ_I): New.
13285 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
13287 * config.gcc: Add arc/t-multilib-linux to tmake_file for
13289 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
13292 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13294 * config/nds32/constraints.md (S): New constraint.
13295 * config/nds32/nds32.md (call_internal): Use constraint S.
13296 (call_value_internal): Likewise.
13297 (sibcall_internal): Likewise.
13298 (sibcall_value_internal): Likewise.
13300 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
13301 Chung-Ju Wu <jasonwucj@gmail.com>
13303 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
13304 into consideration.
13306 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
13307 Chung-Ju Wu <jasonwucj@gmail.com>
13309 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
13310 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
13311 (nds32_rtx_costs_impl): Simplify.
13312 (nds32_address_cost_impl): Simplify.
13313 (nds32_init_rtx_costs): New function.
13314 (nds32_rtx_costs_speed_prefer): Likewise.
13315 (nds32_rtx_costs_size_prefer): Likewise.
13316 (nds32_address_cost_speed_prefer): Likewise.
13317 (nds32_address_cost_speed_fwprop): Likewise.
13318 (nds32_address_cost_size_prefer): Likewise.
13319 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
13320 * config/nds32/nds32.c (nds32_option_override): Use
13321 nds32_init_rtx_costs function.
13323 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13325 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
13326 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
13327 (TARGET_PIPELINE_N8): Likewise.
13328 (TARGET_PIPELINE_N10): Likewise.
13329 (TARGET_PIPELINE_N13): Likewise.
13330 (TARGET_PIPELINE_GRAYWOLF): Likewise.
13332 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
13334 * config/nds32/nds32-fpu.md: Update copyright year.
13336 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13338 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
13340 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13342 * config/nds32/nds32.c
13343 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
13344 * config/nds32/nds32.opt (minline-asm-r15): New option.
13346 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13348 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
13350 * config/nds32/nds32.md (abssi2): New pattern.
13352 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
13354 * config/i386/i386.md (rex64namesuffix): New mode attribute.
13355 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
13356 Merge insn pattern from sse_cvtsi2ss<round_name> and
13357 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
13358 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
13359 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
13360 using SWI48 mode iterator.
13361 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
13362 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
13363 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
13364 pattern from sse_cvttss2si<round_saeonly_name>
13365 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
13366 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
13367 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
13368 using SWI48 mode iterator.
13369 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
13370 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
13371 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
13372 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
13373 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
13374 using SWI48 mode iterator.
13375 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
13376 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
13377 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
13378 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
13379 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
13380 SWI48 mode iterator.
13381 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
13382 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
13383 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
13384 pattern from sse_cvttsd2si<round_saeonly_name>
13385 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
13387 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13389 * config/nds32/nds32-md-auxiliary.c
13390 (nds32_valid_smw_lwm_base_p): Refine.
13391 (nds32_output_smw_single_word): Refine.
13392 (nds32_output_smw_double_word): New.
13393 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
13395 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13397 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
13398 (nds32_output_stack_pop): Refine.
13399 (nds32_expand_unaligned_load): Refine.
13400 (nds32_expand_unaligned_store): Refine.
13402 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13403 Chung-Ju Wu <jasonwucj@gmail.com>
13405 * config/nds32/constants.md: Add TP_REGNUM constant.
13406 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
13407 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
13409 * config/nds32/nds32-doubleword.md: Consider flag_pic.
13410 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
13411 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
13412 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
13413 and PIC code generation.
13414 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
13416 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
13418 * config/nds32/nds32.md: Support TLS and PIC.
13419 * config/nds32/nds32.c: Support TLS and PIC.
13420 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
13421 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
13424 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13426 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
13427 mode with E_ prefix.
13429 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13430 Chung-Ju Wu <jasonwucj@gmail.com>
13432 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
13433 * config/nds32/nds32-md-auxiliary.c
13434 (symbolic_reference_mentioned_p): New.
13435 (nds32_legitimize_ict_address): New.
13436 (nds32_expand_ict_move): New.
13437 (nds32_indirect_call_referenced_p): New.
13438 (nds32_symbol_binds_local_p): Delete.
13439 (nds32_long_call_p): Modify.
13440 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
13441 * config/nds32/nds32-protos.h
13442 (symbolic_reference_mentioned_p): Declare.
13443 (nds32_legitimize_ict_address): Declare.
13444 (nds32_expand_ict_move): Declare.
13445 (nds32_indirect_call_referenced_p): Declare.
13446 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
13447 (nds32_relax_group): Use nds32_ict_const_p as condition.
13448 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
13449 (nds32_asm_file_start): Output ict_model directive in asm code.
13450 (nds32_legitimate_address_p): Consider indirect call.
13451 (nds32_print_operand): Consider indirect call.
13452 (nds32_print_operand_address): Consider indirect call.
13453 (nds32_insert_attributes): Handle "indirect_call" attribute.
13454 (TARGET_LEGITIMATE_ADDRESS_P): Define.
13455 (TARGET_LEGITIMATE_CONSTANT_P): Define.
13456 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
13457 (TARGET_DELEGITIMIZE_ADDRESS): Define.
13458 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
13459 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
13460 (TARGET_ICT_MODEL_SMALL): Define.
13461 (TARGET_ICT_MODEL_LARGE): Define.
13462 * config/nds32/nds32.md (movsi): Consider ict model.
13463 (call, call_value): Consider ict model.
13464 (sibcall, sibcall_value): Consider ict model.
13465 * config/nds32/nds32.opt (mict-model): New option.
13466 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
13469 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
13470 Monk Chiang <sh.chiang04@gmail.com>
13471 Jim Wilson <jimw@sifive.com>
13473 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
13474 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
13475 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
13476 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
13477 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
13478 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
13479 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
13480 compute save_libcall_adjustment properly.
13481 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
13482 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
13483 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
13484 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
13485 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
13486 (ABI_SPEC): Handle mabi=ilp32e.
13487 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
13488 (RVE): Add RVE mask.
13489 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
13490 <-march>: Add rv32e as an example.
13492 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
13495 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
13496 (intra_create_variable_infos): Handle C++ constructors.
13498 2018-05-18 Martin Liska <mliska@suse.cz>
13500 * passes.def: Remove a redundant pass.
13502 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
13505 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
13507 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13509 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
13511 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
13513 (armv2, armv2a, armv3, armv3m): Delete architectures.
13514 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
13515 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
13516 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
13518 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
13519 (*mulsidi3adddi): Likewise.
13520 (mulsidi3): Likewise.
13521 (*mulsidi3_nov6): Likewise.
13522 (umulsidi3): Likewise.
13523 (umulsidi3_nov6): Likewise.
13524 (umaddsidi4): Likewise.
13525 (*umulsidi3adddi): Likewise.
13526 (smulsi3_highpart): Likewise.
13527 (*smulsi3_highpart_nov6): Likewise.
13528 (umulsi3_highpart): Likewise.
13529 (*umulsi3_highpart_nov6): Likewise.
13530 * config/arm/arm.h (arm_arch3m): Delete.
13531 * config/arm/arm.c (arm_arch3m): Delete.
13532 (arm_option_override_internal): Update armv3-related comment.
13533 (arm_configure_build_target): Delete use of isa_bit_mode32.
13534 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
13535 (arm_rtx_costs_internal): Delete check of arm_arch3m.
13536 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
13537 (mulsa3): Likewise.
13538 (mulusa3): Likewise.
13539 * config/arm/arm-protos.h (arm_arch3m): Delete.
13540 * config/arm/arm-tables.opt: Regenerate.
13541 * config/arm/arm-tune.md: Likewise.
13542 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
13543 deleted architectures.
13545 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13547 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
13548 (armv5t, armv5te): New features.
13549 (ARMv5, ARMv5e): Delete fgroups.
13550 (ARMv5t, ARMv5te): Adjust for above changes.
13551 (ARMv6m): Likewise.
13552 (armv5, armv5e): Delete arches.
13553 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
13555 (*call_reg_arm): Likewise.
13556 (*call_value_reg_armv5): Likewise.
13557 (*call_value_reg_arm): Likewise.
13558 (*call_symbol): Likewise.
13559 (*call_value_symbol): Likewise.
13560 (*sibcall_insn): Likewise.
13561 (*sibcall_value_insn): Likewise.
13562 (clzsi2): Likewise.
13563 (prefetch): Likewise.
13564 (define_split and define_peephole2 dependent on arm_arch5):
13566 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
13568 (TARGET_ARM_QBIT): Likewise.
13569 (TARGET_DSP_MULTIPLY): Likewise.
13570 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
13571 (arm_arch5, arm_arch5e): Delete.
13572 (arm_arch5t, arm_arch5te): Declare.
13573 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
13574 (arm_arch5t): Declare.
13575 (arm_option_reconfigure_globals): Update for the above.
13576 (arm_options_perform_arch_sanity_checks): Update comment, replace
13577 use of arm_arch5 with arm_arch5t.
13578 (use_return_insn): Likewise.
13579 (arm_emit_call_insn): Likewise.
13580 (output_return_instruction): Likewise.
13581 (arm_final_prescan_insn): Likewise.
13582 (arm_coproc_builtin_available): Likewise.
13583 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
13584 arm_arch5e with arm_arch5t and arm_arch5te.
13585 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
13586 (arm_arch5t, arm_arch5te): Declare.
13587 * config/arm/arm-tables.opt: Regenerate.
13588 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
13589 * config/arm/t-multilib: Likewise.
13590 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
13591 instead of arm_arch5.
13592 (*call_reg_thumb1): Likewise.
13593 (*call_value_reg_thumb1_v5): Likewise.
13594 (*call_value_reg_thumb1): Likewise.
13595 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
13597 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
13599 2018-05-18 Martin Liska <mliska@suse.cz>
13601 PR gcov-profile/84846
13602 * doc/gcov.texi: Document -t option of gcov tool.
13604 2018-05-18 Martin Liska <mliska@suse.cz>
13606 PR gcov-profile/84846
13607 * gcov.c (print_usage): Add new -t option.
13608 (process_args): Handle the option.
13609 (generate_results): Use stdout as output when requested by
13612 2018-05-18 Martin Liska <mliska@suse.cz>
13614 PR gcov-profile/84846
13615 * coverage.c (coverage_init): Write PWD to .gcno file.
13616 * doc/gcov.texi: Document how working directory is printed.
13617 * gcov-dump.c (dump_gcov_file): Print PWD.
13618 * gcov.c (output_intermediate_file): Likewise.
13619 (read_graph_file): Read PWD string.
13620 (output_lines): Print PWD.
13622 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13624 PR middle-end/85817
13625 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
13626 for retval and return false if all args to phi are zero.
13628 2018-05-18 Richard Biener <rguenther@suse.de>
13630 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
13632 (evrp_dom_walker::before_dom_children): Call it.
13634 2018-05-18 Richard Biener <rguenther@suse.de>
13636 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
13637 results when processing array refs with variable index.
13639 2018-05-18 Toon Moene <toon@moene.org>
13641 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
13642 directly after that of -floop-interchange. Indicate that both
13643 options are enabled by default when specifying -O3.
13645 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13647 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
13648 iterator. Delete separate integer-mode vec_set<mode> expander.
13649 (aarch64_simd_vec_setv2di): Delete.
13650 (vec_setv2di): Delete.
13651 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
13652 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
13653 the "w, r" alternative.
13655 2018-05-18 Martin Liska <mliska@suse.cz>
13657 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
13658 * tree-pass.h (make_pass_lower_switch_O0): New function.
13659 * tree-switch-conversion.c (node_has_low_bound): Remove.
13660 (node_has_high_bound): Likewise.
13661 (node_is_bounded): Likewise.
13662 (class pass_lower_switch): Make it a template type and create
13664 (pass_lower_switch::execute): Add template argument.
13665 (make_pass_lower_switch): New function.
13666 (make_pass_lower_switch_O0): New function.
13667 (do_jump_if_equal): Remove.
13668 (emit_case_nodes): Simplify to just handle all 3 cases and leave
13669 all the hard work to tree optimization passes.
13671 2018-05-18 Martin Liska <mliska@suse.cz>
13673 * dbgcnt.c (limit_low): Renamed from limit.
13674 (limit_high): New variable.
13675 (dbg_cnt_is_enabled): Check for upper limit.
13676 (dbg_cnt): Adjust dumping.
13677 (dbg_cnt_set_limit_by_index): Add new argument for high
13679 (dbg_cnt_set_limit_by_name): Likewise.
13680 (dbg_cnt_process_single_pair): Parse new format.
13681 (dbg_cnt_process_opt): Use strtok.
13682 (dbg_cnt_list_all_counters): Remove 'value' and add
13684 * doc/invoke.texi: Document changes.
13686 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
13688 * doc/sourcebuild.texi (scalar_all_fma): Document.
13689 * tree.def (FMA_EXPR): Delete.
13690 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
13691 * internal-fn.c (ternary_direct): New macro.
13692 (expand_ternary_optab_fn): Likewise.
13693 (direct_ternary_optab_supported_p): Likewise.
13694 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
13695 * builtins.c (fold_builtin_fma): Delete.
13696 (fold_builtin_3): Don't call it.
13697 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
13698 * expr.c (expand_expr_real_2): Likewise.
13699 * fold-const.c (operand_equal_p): Likewise.
13700 (fold_ternary_loc): Likewise.
13701 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
13702 * gimple.c (DEFTREECODE): Likewise.
13703 * gimplify.c (gimplify_expr): Likewise.
13704 * optabs-tree.c (optab_for_tree_code): Likewise.
13705 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
13706 * tree-eh.c (operation_could_trap_p): Likewise.
13707 (stmt_could_throw_1_p): Likewise.
13708 * tree-inline.c (estimate_operator_cost): Likewise.
13709 * tree-pretty-print.c (dump_generic_node): Likewise.
13710 (op_code_prio): Likewise.
13711 * tree-ssa-loop-im.c (stmt_cost): Likewise.
13712 * tree-ssa-operands.c (get_expr_operands): Likewise.
13713 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
13714 * fold-const-call.h (fold_fma): Delete.
13715 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
13716 CFN_FNMA and CFN_FNMS.
13717 (fold_fma): Delete.
13718 * genmatch.c (combined_fn): New enum.
13719 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
13720 (commutative_op): New function.
13721 (commutate): Use it. Handle more than 2 operands.
13722 (dt_operand::gen_gimple_expr): Use commutative_op.
13723 (parser::parse_expr): Allow :c to be used with non-binary
13724 operators if the commutative operand is known.
13725 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
13726 CFN_FMS, CFN_FNMA and CFN_FNMS.
13727 (backprop::process_assign_use): Remove FMA_EXPR handling.
13728 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
13729 (gen_hsa_fma): New function.
13730 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
13731 IFN_FNMA and IFN_FNMS.
13732 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
13733 * gimple-fold.h (follow_all_ssa_edges): Declare.
13734 * gimple-fold.c (follow_all_ssa_edges): New function.
13735 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
13736 gimple_build interface and use follow_all_ssa_edges to fold the result.
13737 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
13738 instead of checking for optabs directly.
13739 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
13740 rather than FMA_EXPRs.
13741 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
13742 call to IFN_FMA instead of an FMA_EXPR.
13744 2018-05-17 Jim Wilson <jimw@sifive.com>
13746 * expr.c (do_tablejump): When converting index to Pmode, if we have a
13747 sign extended promoted subreg, and the range does not have the sign bit
13748 set, then do a sign extend.
13750 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
13751 test, check for sign extended subreg and/or constant operands, and
13752 do a sign extend in that case.
13754 2018-05-17 Steve Ellcey <sellcey@cavium.com>
13756 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
13757 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
13759 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
13760 Change logics_shift_reg to logics_shift_imm.
13761 (thunderx2t99_fp_loadpair_basic): Delete.
13762 (thunderx2t99_fp_storepair_basic): Delete.
13763 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13764 (thunderx2t99_asimd_polynomial): Delete.
13765 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13766 and neon_fp_mul_d_scalar_q.
13767 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13768 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13769 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13770 (thunderx2t99_asimd_lut): Add missing tbl types.
13771 (thunderx2t99_asimd_ext): Delete.
13772 (thunderx2t99_asimd_load1_1_mult): Delete.
13773 (thunderx2t99_asimd_load1_2_mult): Delete.
13774 (thunderx2t99_asimd_load1_ldp): New.
13775 (thunderx2t99_asimd_load1): New.
13776 (thunderx2t99_asimd_load2): Add missing *load2* types.
13777 (thunderx2t99_asimd_load3): New.
13778 (thunderx2t99_asimd_load4): New.
13779 (thunderx2t99_asimd_store1_1_mult): Delete.
13780 (thunderx2t99_asimd_store1_2_mult): Delete.
13781 (thunderx2t99_asimd_store2_mult): Delete.
13782 (thunderx2t99_asimd_store2_onelane): Delete.
13783 (thunderx2t99_asimd_store_stp): New.
13784 (thunderx2t99_asimd_store1): New.
13785 (thunderx2t99_asimd_store2): New.
13786 (thunderx2t99_asimd_store3): New.
13787 (thunderx2t99_asimd_store4): New.
13789 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
13791 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13792 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
13794 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
13795 Segher Boessenkool <segher@kernel.crashing.org>
13798 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
13801 2018-05-17 Richard Biener <rguenther@suse.de>
13803 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
13804 for pruning loop and prune defs feeding only already visited PHIs.
13806 2018-05-17 Richard Biener <rguenther@suse.de>
13808 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
13810 2018-05-17 Bin Cheng <bin.cheng@arm.com>
13811 Richard Biener <rguenther@suse.de>
13813 PR tree-optimization/85793
13814 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
13815 for VMAT_ELEMENTWISE.
13817 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13819 * internal-fn.h (lookup_internal_fn): Declare
13820 * internal-fn.c (lookup_internal_fn): New function.
13821 * gimple.c (gimple_build_call_from_tree): Handle calls to
13822 internal functions.
13823 * gimple-pretty-print.c (dump_gimple_call): Print "." before
13824 internal function names.
13825 * tree-pretty-print.c (dump_generic_node): Likewise.
13826 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
13828 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
13830 * gimple-fold.h (gimple_build): Make the function forms take
13831 combined_fn rather than built_in_function.
13832 (gimple_simplify): Likewise.
13833 * gimple-match-head.c (gimple_simplify): Likewise.
13834 * gimple-fold.c (gimple_build): Likewise.
13835 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
13836 rather than gimple_build_call_internal.
13837 (get_initial_defs_for_reduction): Likewise.
13838 (vect_create_epilog_for_reduction): Likewise.
13839 (vectorizable_live_operation): Likewise.
13841 2018-05-17 Martin Liska <mliska@suse.cz>
13843 * gimple-ssa-sprintf.c (format_directive): Do not use
13844 space in between 'G_' and '('.
13846 2018-05-17 Jakub Jelinek <jakub@redhat.com>
13849 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
13850 even if the mask is not all ones.
13853 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
13855 (ix86_gimple_fold_builtin): Likewise.
13858 * config/i386/i386.c: Include tree-vector-builder.h.
13859 (ix86_vector_shift_count): New function.
13860 (ix86_fold_builtin): Fold shift builtins by scalar count.
13861 (ix86_gimple_fold_builtin): Likewise.
13863 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
13864 _mm512_setzero): New intrinsics.
13866 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
13867 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13869 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
13870 code generation for cases where splatting a value is not useful.
13871 * simplify-rtx.c (simplify_ternary_operation): Simplify
13872 vec_merge across a vec_duplicate and a paradoxical subreg forming
13873 a vector mode to a vec_concat.
13875 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13877 * config.gcc: Support "goldmont-plus".
13878 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13880 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13881 PROCESSOR_GOLDMONT_PLUS.
13882 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
13883 (processor_target_table): Add "goldmont-plus".
13884 (PTA_GOLDMONT_PLUS): Define.
13885 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
13886 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
13887 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
13888 (fold_builtin_cpu): Add "goldmont-plus".
13889 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
13890 (ix86_option_override_internal): Add "goldmont-plus".
13891 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
13892 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
13893 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
13894 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
13896 2018-05-17 Richard Biener <rguenther@suse.de>
13898 PR tree-optimization/85757
13899 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
13900 remove defs that only feed that PHI from further processing.
13902 2018-05-16 Jim Wilson <jimw@sifive.com>
13904 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
13906 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
13908 2018-05-16 Mark Wielaard <mark@klomp.org>
13910 * dwarf2out.c (count_index_strings): New function.
13911 (output_indirect_strings): Call count_index_strings and generate
13912 header for dwarf_version >= 5.
13914 2018-05-16 Mark Wielaard <mark@klomp.org>
13916 * dwarf2out.c (dwarf_FORM): New function.
13917 (set_indirect_string): Use dwarf_FORM.
13918 (reset_indirect_string): Likewise.
13919 (size_of_die): Likewise.
13920 (value_format): Likewise.
13921 (output_die): Likewise.
13922 (add_skeleton_AT_string): Likewise.
13923 (output_macinfo_op): Likewise.
13924 (index_string): Likewise.
13925 (output_index_string_offset): Likewise.
13926 (output_index_string): Likewise.
13927 (count_index_strings): Likewise.
13929 2018-05-16 Carl Love <cel@us.ibm.com>
13931 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13932 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
13934 2018-05-16 Martin Jambor <mjambor@suse.cz>
13936 * ipa-prop.c (ipa_free_all_edge_args): Remove.
13937 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
13939 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
13941 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
13942 (fnma<mode>4): Likewise.
13943 (fms<mode>4): Likewise.
13944 (fnms<mode>4): Likewise.
13945 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
13946 (aarch64_fnma<mode>4): Likewise.
13947 (aarch64_fms<mode>4): Likewise.
13948 (aarch64_fnms<mode>4): Likewise.
13949 (aarch64_fnmadd<mode>4): Likewise.
13951 2018-05-16 Jason Merrill <jason@redhat.com>
13953 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
13955 2018-05-16 Richard Biener <rguenther@suse.de>
13957 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
13958 (dump_stmt_cost): Declare.
13959 (add_stmt_cost): Dump cost we add.
13960 (add_stmt_costs): New function.
13961 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
13962 No longer exported.
13963 (vect_analyze_stmt): Adjust prototype.
13964 (vectorizable_condition): Likewise.
13965 (vectorizable_live_operation): Likewise.
13966 (vectorizable_reduction): Likewise.
13967 (vectorizable_induction): Likewise.
13968 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
13969 cost vector to pass to vectorizable_ and record afterwards.
13970 (vect_model_reduction_cost): Take cost vector argument and adjust.
13971 (vect_model_induction_cost): Likewise.
13972 (vectorizable_reduction): Likewise.
13973 (vectorizable_induction): Likewise.
13974 (vectorizable_live_operation): Likewise.
13975 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
13976 SLP_TREE_NUMBER_OF_VEC_STMTS.
13977 (vect_analyze_slp_cost_1): Remove.
13978 (vect_analyze_slp_cost): Likewise.
13979 (vect_slp_analyze_node_operations): Take visited args and
13980 a target cost vector. Avoid processing already visited stmt sets.
13981 (vect_slp_analyze_operations): Use a local cost vector to gather
13982 costs and register those of non-discarded instances.
13983 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
13984 (vect_schedule_slp_instance): Remove copying of
13985 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
13987 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
13988 adding cost. Record cost entry location.
13989 (vect_prologue_cost_for_slp_op): Function to compute cost of
13990 a constant or invariant generated for SLP vect in the prologue,
13991 split out from vect_analyze_slp_cost_1.
13992 (vect_model_simple_cost): Make static. Adjust for SLP costing.
13993 (vect_model_promotion_demotion_cost): Likewise.
13994 (vect_model_store_cost): Likewise, make static.
13995 (vect_model_load_cost): Likewise.
13996 (vectorizable_bswap): Add cost vector arg and adjust.
13997 (vectorizable_call): Likewise.
13998 (vectorizable_simd_clone_call): Likewise.
13999 (vectorizable_conversion): Likewise.
14000 (vectorizable_assignment): Likewise.
14001 (vectorizable_shift): Likewise.
14002 (vectorizable_operation): Likewise.
14003 (vectorizable_store): Likewise.
14004 (vectorizable_load): Likewise.
14005 (vectorizable_condition): Likewise.
14006 (vectorizable_comparison): Likewise.
14007 (can_vectorize_live_stmts): Likewise.
14008 (vect_analyze_stmt): Likewise.
14009 (vect_transform_stmt): Adjust calls to vectorizable_*.
14010 * tree-vectorizer.c: Include gimple-pretty-print.h.
14011 (dump_stmt_cost): New function.
14013 2018-05-16 Richard Biener <rguenther@suse.de>
14015 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
14016 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
14017 * tree-ssa-dse.c: Include tree-ssa-loop.h.
14018 (check_name): New callback.
14019 (dse_classify_store): Track cycles via a visited bitmap of PHI
14020 defs and simplify handling of in-loop and across loop dead stores
14021 and properly fail for loop-variant refs. Handle byte-tracking with
14022 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
14025 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
14027 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
14028 (vect_get_mask_type_for_stmt): Likewise.
14029 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
14031 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
14032 to determine the statement's vector type and the vector type that
14033 should be used for calculating nunits. Deal with cases in which
14034 the type has to be deferred.
14035 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
14036 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
14037 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
14038 (vect_determine_vf_for_stmt): New functions, split out from...
14039 (vect_determine_vectorization_factor): ...here.
14040 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
14041 (vect_get_mask_type_for_stmt): New functions, split out from
14042 vect_determine_vectorization_factor.
14044 2018-05-16 Richard Biener <rguenther@suse.de>
14046 * tree-cfg.c (verify_gimple_assign_ternary): Properly
14047 verify the [VEC_]COND_EXPR embedded comparison.
14049 2018-05-15 Martin Sebor <msebor@redhat.com>
14051 PR tree-optimization/85753
14052 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
14053 RECORD_TYPE in addition to ARRAY_TYPE.
14055 2018-05-15 Martin Sebor <msebor@redhat.com>
14057 PR middle-end/85643
14058 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
14060 2018-05-15 Richard Biener <rguenther@suse.de>
14062 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
14063 add by_clobber_p one. Change algorithm to collect all defs
14064 representing uses we need to walk and try reducing them to
14065 a single one before failing.
14066 (dse_dom_walker::dse_optimize_stmt): Adjust.
14068 2018-05-13 Mark Wielaard <mark@klomp.org>
14070 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
14071 (size_of_loc_descr): Likewise.
14072 (output_loc_operands): Likewise.
14073 (output_loc_operands_raw): Likewise.
14074 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
14075 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
14076 (hash_loc_operands): Likewise.
14077 (compare_loc_operands): Likewise.
14079 2018-05-14 Mark Wielaard <mark@klomp.org>
14081 * dwarf2out.c (count_index_addrs): New function.
14082 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
14084 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14086 PR tree-optimization/83648
14087 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
14088 return value as malloc candidate.
14090 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14093 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
14094 param as true in call to suggest_attribute.
14096 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
14098 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
14099 -mreadonly-in-sdata.
14101 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14103 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
14105 (aarch64_crypto_aesd_fused): Likewise.
14107 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
14109 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
14110 (movsi_aarch64): Likewise.
14111 (load_pairsi): Likewise.
14112 (load_pairdi): Likewise.
14113 (store_pairsi): Likewise.
14114 (store_pairdi): Likewise.
14115 (load_pairsf): Likewise.
14116 (load_pairdf): Likewise.
14117 (store_pairsf): Likewise.
14118 (store_pairdf): Likewise.
14119 (zero_extend): Likewise.
14120 (trunc): Swap alternatives.
14121 (fcvt_target): Add '?' to prefer w over r.
14123 2018-05-14 Jakub Jelinek <jakub@redhat.com>
14126 * config/i386/i386.md: Disallow non-commutative arithmetics in
14127 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
14128 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
14129 in the peephole2 before it.
14131 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
14133 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
14134 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
14135 (ix86_handle_option): Handle -mcldemote.
14136 * config.gcc: New header.
14137 * config/i386/cldemoteintrin.h: New file.
14138 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
14139 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14141 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14142 OPTION_MASK_ISA_CLDEMOTE.
14143 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
14144 (ix86_valid_target_attribute_inner_p): Ditto.
14145 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
14146 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
14147 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
14148 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
14149 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
14151 * config/i386/i386.opt: Add -mcldemote.
14152 * config/i386/x86intrin.h: New header.
14153 * doc/invoke.texi: Add -mcldemote.
14155 2018-05-14 Richard Biener <rguenther@suse.de>
14157 * doc/match-and-simplify.texi: Adjust :s documentation.
14159 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
14161 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
14162 intended memcpy size.
14163 (REORDER_45): Likewise.
14165 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
14167 * sort.cc: New file.
14168 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
14169 * vec.c (qsort_chk): Use gcc_qsort.
14170 * Makefile.in (OBJS-libcommon): Add sort.o.
14171 (build/sort.o): New target. Use it...
14172 (BUILD_RTL): ... here, and...
14173 (build/gencfn-macros): ... here, and...
14174 (build/genmatch): ... here.
14176 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14177 Chung-Ju Wu <jasonwucj@gmail.com>
14179 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
14180 * config/nds32/nds32-graywolf.md: New file.
14181 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
14182 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
14184 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
14185 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
14186 * config/nds32/nds32.md (pipeline_model): Add graywolf.
14187 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
14188 * config/nds32/pipelines.md: Include n15 settings.
14190 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14191 Chung-Ju Wu <jasonwucj@gmail.com>
14193 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
14194 * config/nds32/nds32-n13.md: New file.
14195 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
14196 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
14198 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
14199 * config/nds32/nds32.md (pipeline_model): Add n13.
14200 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
14201 * config/nds32/pipelines.md: Include n13 settings.
14203 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14204 Chung-Ju Wu <jasonwucj@gmail.com>
14206 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
14207 * config/nds32/nds32-n10.md: New file.
14208 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
14209 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
14211 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
14212 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
14213 * config/nds32/nds32.md (pipeline_model): Add n10.
14214 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
14215 * config/nds32/pipelines.md: Include n10 settings.
14217 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
14218 Kito Cheng <kito.cheng@gmail.com>
14219 Chung-Ju Wu <jasonwucj@gmail.com>
14221 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
14222 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14223 Add enum values for DSP extension instructions.
14224 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
14226 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
14227 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
14228 New code iterators.
14229 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
14230 * config/nds32/nds32-dspext.md: New file for DSP implementation.
14231 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
14232 * config/nds32/nds32-intrinsic.md: Likewise.
14233 * config/nds32/nds32_intrinsic.h: Likewise.
14234 * config/nds32/nds32-md-auxiliary.c: Likewise.
14235 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
14236 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
14237 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
14238 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
14239 * config/nds32/nds32-protos.h: New declarations for DSP extension.
14240 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
14241 TYPE_DMAC in switch statement.
14242 * config/nds32/nds32.c: New checking and implementation for DSP
14243 extension instructions.
14244 * config/nds32/nds32.h: Likewise.
14245 * config/nds32/nds32.md: Likewise.
14246 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
14247 * config/nds32/predicates.md: Implement new predicates for DSP
14250 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14252 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
14253 Reformat alternatives and attributes so it is easier to identify
14254 which constraints/attributes go with which instruction.
14255 (mov<mode>_hardfloat32, FMOVE64): Likewise.
14256 (mov<mode>_softfloat32, FMOVE64): Likewise.
14257 (mov<mode>_hardfloat64, FMOVE64): Likewise.
14258 (mov<mode>_softfloat64, FMOVE64): Likewise.
14260 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
14262 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14264 (Basic PowerPC Built-in Functions): The new name of the
14265 subsection previously known as "PowerPC Built-in Functions".
14266 (Basic PowerPC Built-in Functions Available on all Configurations):
14268 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
14269 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
14270 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
14271 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
14273 2018-05-11 Martin Jambor <mjambor@suse.cz>
14276 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
14279 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
14282 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
14284 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
14286 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
14287 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
14288 (ix86_handle_option): Handle -mwaitpkg.
14289 * config.gcc: New header.
14290 * config/i386/cpuid.h (bit_WAITPKG): New bit.
14291 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
14292 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
14294 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14295 OPTION_MASK_ISA_WAITPKG.
14296 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
14297 (ix86_option_override_internal): Add PTA_WAITPKG.
14298 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
14299 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
14300 IX86_BUILTIN_TPAUSE.
14301 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
14302 __builtin_ia32_umwait and __builtin_ia32_tpause.
14303 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
14304 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
14305 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
14306 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
14307 UNSPECV_TPAUSE): New.
14308 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
14309 * config/i386/i386.opt: Add -mwaitpkg.
14310 * config/i386/waitpkgintrin.h: New file.
14311 * config/i386/x86intrin.h: New header.
14312 * doc/invoke.texi: Add -mwaitpkg.
14314 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
14317 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
14319 (cortex-m0): Use armv6s-m isa.
14320 (cortex-m0plus): Likewise.
14321 (cortex-m1): Likewise.
14322 (cortex-m0.small-multiply): Likewise.
14323 (cortex-m0plus.small-multiply): Likewise.
14324 (cortex-m1.small-multiply): Likewise.
14326 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
14327 Jakub Jelinek <jakub@redhat.com>
14329 PR tree-optimization/85692
14330 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
14331 source permute as well.
14333 2018-05-11 Martin Liska <mliska@suse.cz>
14336 * doc/extend.texi: Document LLVM style format for no_sanitize
14339 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14341 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
14342 mode_supports_vsx_dform_quad to mode_supports_dq_form.
14343 (mode_supports_vsx_dform_quad): Likewise.
14344 (mode_supports_vmx_dform): Move these functions to be next to the
14345 other mode_supports functions.
14346 (mode_supports_dq_form): Likewise.
14347 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
14348 mode_supports_dq_form.
14349 (reg_offset_addressing_ok_p): Likewise.
14350 (offsettable_ok_by_alignment): Likewise.
14351 (rs6000_legitimate_offset_address_p): Likewise.
14352 (legitimate_lo_sum_address_p): Likewise.
14353 (rs6000_legitimize_address): Likewise.
14354 (rs6000_legitimize_reload_address): Likewise.
14355 (rs6000_secondary_reload_inner): Likewise.
14356 (rs6000_preferred_reload_class): Likewise.
14357 (rs6000_output_move_128bit): Likewise.
14359 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14361 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
14362 Generate SImode target register for null target.
14363 <case IX86_BUILTIN_XGETBV>: Ditto.
14364 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
14365 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
14367 2018-05-10 Carl Love <cel@us.ibm.com>
14369 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14370 dcbtt and dcbtstt if operands[2] is 0.
14372 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14375 * config/i386/sse.md (usadv64qi): New expander.
14377 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
14379 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
14380 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
14381 -maltivec=be support.
14382 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
14383 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
14384 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
14385 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
14386 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
14387 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
14388 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
14389 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
14390 altivec_vsumsws): Adjust.
14391 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
14392 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
14393 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
14395 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
14396 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
14397 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
14398 (altivec_lve<VI_char>x): Delete expand.
14399 (*altivec_lve<VI_char>x_internal): Rename to...
14400 (altivec_lve<VI_char>x): ... this.
14401 (altivec_lvxl_<mode>): Delete expand.
14402 (*altivec_lvxl_<mode>_internal): Rename to ...
14403 (altivec_lvxl_<mode>): ... this.
14404 (altivec_stvxl_<mode>): Delete expand.
14405 (*altivec_stvxl_<mode>_internal): Rename to ...
14406 (altivec_stvxl_<mode>): ... this.
14407 (altivec_stve<VI_char>x): Delete expand.
14408 (*altivec_stve<VI_char>x_internal): Rename to ...
14409 (altivec_stve<VI_char>x): ... this.
14410 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
14411 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
14412 reduc_plus_scal_<mode>): Adjust.
14413 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
14415 (rs6000_cpu_cpp_builtins): Adjust.
14416 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
14417 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
14418 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
14419 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
14420 -maltivec=be support.
14421 (rs6000_split_vec_extract_var): Adjust.
14422 (rs6000_split_v4si_init): Adjust.
14423 (swap_selector_for_mode): Delete.
14424 (altivec_expand_lvx_be, altivec_expand_stvx_be,
14425 altivec_expand_stvex_be): Delete.
14426 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
14427 -maltivec=be support.
14428 (rs6000_gimple_fold_builtin): Ditto.
14429 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
14431 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
14432 (TARGET_DIRECT_MOVE_64BIT): Adjust.
14433 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
14434 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
14435 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
14436 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
14437 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
14438 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
14439 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
14440 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
14441 anonymous split): Adjust.
14442 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
14443 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
14445 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
14447 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
14448 when --with-gxx-include-dir is also specified.
14449 * configure: Regenerate.
14451 2018-05-09 Jim Wilson <jimw@sifive.com>
14454 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
14455 * config/riscv/t-withmultilib: New.
14456 * config/riscv/withmultilib.h: New.
14457 * doc/install.texi: Document RISC-V --with-multilib-list support.
14459 2018-05-09 Richard Biener <rguenther@suse.de>
14461 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
14463 (vect_bb_vectorization_profitable_p): Adjust. Compute
14464 actual scalar cost using the cost vector and the add_stmt_cost
14467 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14469 PR rtl-optimization/85645
14470 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
14471 in the REG_CFA_REGISTER note for LR, don't leave it empty.
14473 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14475 PR rtl-optimization/85645
14476 * shrink-wrap.c (spread_components): Return a boolean saying if
14477 anything was changed.
14478 (try_shrink_wrapping_separate): Iterate spread_components until
14479 nothing changes anymore.
14481 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14483 PR rtl-optimization/85645
14484 * regrename.c (build_def_use): Also kill the chains that include the
14485 destination of a REG_CFA_REGISTER note.
14487 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14489 PR rtl-optimization/85645
14490 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
14491 insn that has a REG_CFA_REGISTER note.
14493 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
14495 * cfgexpand.c (expand_clobber): New function.
14496 (expand_gimple_stmt_1): Use it.
14497 * tree-vect-stmts.c (vect_clobber_variable): New function,
14499 (vectorizable_simd_clone_call): ...here.
14500 (vectorizable_store): Emit a clobber either side of an
14501 IFN_STORE_LANES sequence.
14502 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
14504 2018-05-09 Tom de Vries <tom@codesourcery.com>
14507 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
14508 (define_insn "trap_if_false"): Add exit after trap.
14510 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
14512 PR rtl-optimization/85638
14513 * bb-reorder.c: Include common/common-target.h.
14514 (create_forwarder_block): New function extracted from...
14515 (fix_up_crossing_landing_pad): ...here. Rename into...
14516 (dw2_fix_up_crossing_landing_pad): ...this.
14517 (sjlj_fix_up_crossing_landing_pad): New function.
14518 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
14519 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
14520 from both partitions and exit the loop after one iteration.
14522 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14525 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14527 (Basic PowerPC Built-in Functions): The new name of the
14528 subsection previously known as "PowerPC Built-in Functions".
14529 (Basic PowerPC Built-in Functions Available on all Configurations):
14531 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14533 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14535 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14537 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14540 2018-05-08 Jim Wilson <jimw@sifive.com>
14542 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
14543 (LD_EMUL_SUFFIX): New.
14544 (LINK_SPEC): Use it.
14546 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14548 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14550 (Basic PowerPC Built-in Functions): The new name of the
14551 subsection previously known as "PowerPC Built-in Functions".
14552 (Basic PowerPC Built-in Functions Available on all Configurations):
14554 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14556 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14558 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14560 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14563 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14566 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
14567 after cmpelim optimization.
14569 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
14571 * config.gcc: Support "goldmont".
14572 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
14573 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14574 PROCESSOR_GOLDMONT.
14575 * config/i386/i386.c (m_GOLDMONT): Define.
14576 (processor_target_table): Add "goldmont".
14577 (PTA_GOLDMONT): Define.
14578 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
14579 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
14580 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
14581 (fold_builtin_cpu): Add "goldmont".
14582 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
14583 (ix86_option_override_internal): Add "goldmont".
14584 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
14585 (processor_type): Add PROCESSOR_GOLDMONT.
14586 * config/i386/i386.md: Add CPU "glm".
14587 * config/i386/glm.md: New file.
14588 * config/i386/x86-tune.def: Add m_GOLDMONT.
14589 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
14591 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14594 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
14596 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
14597 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
14598 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
14601 * config/i386/i386.c (ix86_fold_builtin): Handle
14602 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
14605 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
14606 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
14608 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
14611 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
14612 (check_arch): Likewise.
14613 (check_fpu): Return the result rather than printing it.
14614 (end arch): Fix operator precedence.
14615 (end cpu): Likewise.
14616 (END): Print the result from check_fpu.
14618 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
14619 Alan Hayward <alan.hayward@arm.com>
14620 David Sherwood <david.sherwood@arm.com>
14622 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
14623 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
14624 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
14625 (*fcmuo<mode>_and): New patterns.
14627 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
14629 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
14630 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
14631 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
14632 (cmp_op, sve_imm_con): New code attributes.
14633 (SVE_COND_INT_CMP, imm_con): Delete.
14634 (cmp_op): Remove above unspecs from int attribute.
14635 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
14637 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
14638 comparison-specific unspecs.
14639 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
14640 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
14641 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
14642 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
14643 (*vec_fcm<cmp_op><mode>): Rename to...
14644 (*fcm<cmp_op><mode>): ...this and adjust likewise.
14645 (*vec_fcmuo<mode>): Rename to...
14646 (*fcmuo<mode>): ...this and adjust likewise.
14647 (*pred_fcm<cmp_op><mode>): New pattern.
14648 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
14649 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
14651 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
14653 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
14654 (aarch64_emit_sve_predicated_cond): New function.
14655 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
14656 (aarch64_emit_unspec_cond_or): Replace with...
14657 (aarch64_emit_sve_or_conds): ...this new function. Use
14658 aarch64_emit_sve_ptrue_op for the individual comparisons and
14659 aarch64_emit_binop to OR them together.
14660 (aarch64_emit_inverted_unspec_cond): Replace with...
14661 (aarch64_emit_sve_inverted_cond): ...this new function. Use
14662 aarch64_emit_sve_ptrue_op for the comparison and
14663 aarch64_emit_unop to invert the result.
14664 (aarch64_expand_sve_vec_cmp_float): Update after the above
14665 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
14667 2018-05-07 Nathan Sidwell <nathan@acm.org>
14669 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
14670 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
14671 (Backwards Compatibility): Likewise.
14673 2018-05-07 Luis Machado <luis.machado@linaro.org>
14677 2018-05-07 Luis Machado <luis.machado@linaro.org>
14679 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14680 <prefetch_dynamic_strides>: New const bool field.
14681 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14682 prefetch_dynamic_strides.
14683 (exynosm1_prefetch_tune): Likewise.
14684 (thunderxt88_prefetch_tune): Likewise.
14685 (thunderx_prefetch_tune): Likewise.
14686 (thunderx2t99_prefetch_tune): Likewise.
14687 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14689 (aarch64_override_options_internal): Update to set
14690 PARAM_PREFETCH_DYNAMIC_STRIDES.
14691 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14692 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14693 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14694 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14695 prefetch-dynamic-strides setting.
14697 2018-05-07 Luis Machado <luis.machado@linaro.org>
14699 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14700 <minimum_stride>: New const int field.
14701 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14702 minimum_stride field.
14703 (exynosm1_prefetch_tune): Likewise.
14704 (thunderxt88_prefetch_tune): Likewise.
14705 (thunderx_prefetch_tune): Likewise.
14706 (thunderx2t99_prefetch_tune): Likewise.
14707 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14708 (aarch64_override_options_internal): Update to set
14709 PARAM_PREFETCH_MINIMUM_STRIDE.
14710 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14711 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14712 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14713 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14714 stride is constant and is below the minimum stride threshold.
14716 2018-05-07 Luis Machado <luis.machado@linaro.org>
14718 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
14721 2018-05-07 Luis Machado <luis.machado@linaro.org>
14723 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14724 <prefetch_dynamic_strides>: New const bool field.
14725 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14726 prefetch_dynamic_strides.
14727 (exynosm1_prefetch_tune): Likewise.
14728 (thunderxt88_prefetch_tune): Likewise.
14729 (thunderx_prefetch_tune): Likewise.
14730 (thunderx2t99_prefetch_tune): Likewise.
14731 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14733 (aarch64_override_options_internal): Update to set
14734 PARAM_PREFETCH_DYNAMIC_STRIDES.
14735 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14736 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14737 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14738 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14739 prefetch-dynamic-strides setting.
14741 2018-05-07 Luis Machado <luis.machado@linaro.org>
14743 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14744 <minimum_stride>: New const int field.
14745 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14746 minimum_stride field.
14747 (exynosm1_prefetch_tune): Likewise.
14748 (thunderxt88_prefetch_tune): Likewise.
14749 (thunderx_prefetch_tune): Likewise.
14750 (thunderx2t99_prefetch_tune): Likewise.
14751 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14752 (aarch64_override_options_internal): Update to set
14753 PARAM_PREFETCH_MINIMUM_STRIDE.
14754 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14755 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14756 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14757 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14758 stride is constant and is below the minimum stride threshold.
14760 2018-05-06 Jakub Jelinek <jakub@redhat.com>
14763 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14764 the type is addressable. Don't force op into register if it has
14767 2018-05-05 Roland McGrath <mcgrathr@google.com>
14770 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14771 any section for which we don't know a specific type it should have,
14772 regardless of name. Previously this was done only for the exact
14773 names ".init_array", ".fini_array", and ".preinit_array".
14774 (default_elf_asm_named_section): Add comment about
14775 relationship with default_section_type_flags and SECTION_NOTYPE.
14776 (get_section): Don't consider it a type conflict if one side has
14777 SECTION_NOTYPE and the other doesn't, as long as neither has the
14778 SECTION_BSS et al used in the default_section_type_flags logic.
14780 2018-05-05 Tom de Vries <tom@codesourcery.com>
14783 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14784 (workaround_barsyncs): New function.
14785 (nvptx_reorg): Use workaround_barsyncs.
14786 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14787 (define_expand "nvptx_membar_cta"): New define_expand.
14788 (define_insn "*nvptx_membar_cta"): New insn.
14790 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
14792 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14793 To improve optimization opportunities.
14794 * builtin-types.def: The new needed builtin types for the above.
14796 2018-05-04 Richard Biener <rguenther@suse.de>
14798 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
14799 * gimple-ssa-store-merging.c
14800 (imm_store_chain_info::output_merged_store): Remove redundant create,
14801 release split_store vector contents on failure.
14802 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
14803 scalar stmt vector on cache hit.
14805 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
14807 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
14809 * config.gcc (powerpc-xilinx-eabi*): Remove.
14810 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
14812 (fusion_addis_mem_combo_load): Ditto.
14813 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
14815 (rs6000_cpu_cpp_builtins): Ditto.
14816 * config/rs6000/rs6000-linux.c
14817 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
14818 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
14819 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
14821 (rs6000_setup_reg_addr_masks): Ditto.
14822 (rs6000_init_hard_regno_mode_ok): Ditto.
14823 (rs6000_option_override_internal): Ditto.
14824 (legitimate_lo_sum_address_p): Ditto.
14825 (rs6000_legitimize_address): Ditto.
14826 (rs6000_legitimize_reload_address): Ditto.
14827 (rs6000_legitimate_address_p): Ditto.
14828 (abi_v4_pass_in_fpr): Ditto.
14829 (setup_incoming_varargs): Ditto.
14830 (rs6000_gimplify_va_arg): Ditto.
14831 (rs6000_split_multireg_move): Ditto.
14832 (rs6000_savres_strategy): Ditto.
14833 (rs6000_emit_prologue_components): Ditto.
14834 (rs6000_emit_epilogue_components): Ditto.
14835 (rs6000_emit_prologue): Ditto.
14836 (rs6000_emit_epilogue): Ditto.
14837 (rs6000_elf_file_end): Ditto.
14838 (rs6000_function_value): Ditto.
14839 (rs6000_libcall_value): Ditto.
14840 * config/rs6000/rs6000.h: Ditto.
14841 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
14842 (TARGET_MINMAX): ... this. New.
14843 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
14844 * config/rs6000/rs6000.md: Remove Xilinx FP support.
14845 (*movsi_internal1_single): Delete.
14846 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
14847 mfpu=, mxilinx-fpu): Delete.
14848 * config/rs6000/singlefp.h: Delete.
14849 * config/rs6000/sysv4.h: Remove Xilinx FP support.
14850 * config/rs6000/t-rs6000: Ditto.
14851 * config/rs6000/t-xilinx: Delete.
14852 * config/rs6000/titan.md: Adjust for fp_type removal.
14853 * config/rs6000/vsx.md: Remove Xilinx FP support.
14854 (VStype_simple): Delete.
14855 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
14856 * config/rs6000/xfpu.h: Delete.
14857 * config/rs6000/xfpu.md: Delete.
14858 * config/rs6000/xilinx.h: Delete.
14859 * config/rs6000/xilinx.opt: Delete.
14860 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
14861 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
14863 2018-05-04 Tom de Vries <tom@codesourcery.com>
14866 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
14869 2018-05-04 Richard Biener <rguenther@suse.de>
14871 PR middle-end/85627
14872 * tree-complex.c (update_complex_assignment): We are always in SSA form.
14873 (expand_complex_div_wide): Likewise.
14874 (expand_complex_operations_1): Likewise.
14875 (expand_complex_libcall): Preserve EH info of the original stmt.
14876 (tree_lower_complex): Handle removed blocks.
14877 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
14878 on complex multiplication and division libcall builtins.
14880 2018-05-04 Richard Biener <rguenther@suse.de>
14882 PR middle-end/85574
14883 * fold-const.c (negate_expr_p): Restrict negation of operand
14884 zero of a division to when we know that can happen without
14886 (fold_negate_expr_1): Likewise.
14888 2018-05-04 Jakub Jelinek <jakub@redhat.com>
14891 * real.h (real_nextafter): Declare.
14892 * real.c (real_nextafter): New function.
14893 * fold-const-call.c (fold_const_nextafter): New function.
14894 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
14895 CASE_CFN_NEXTTOWARD.
14896 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
14897 even when arg1_mode is different from arg0_mode.
14899 2018-05-03 Nathan Sidwell <nathan@acm.org>
14901 * doc/extend.texi (Deprecated Features): Remove
14902 -ffriend-injection.
14903 (Backwards Compatibility): Likewise.
14904 * doc/invoke.texi (C++ Language Options): Likewise.
14905 (C++ Dialect Options): Likewise.
14907 2018-05-03 Jakub Jelinek <jakub@redhat.com>
14910 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
14911 _mm512_mask_mullox_epi64): New intrinsics.
14913 2018-05-03 Tom de Vries <tom@codesourcery.com>
14916 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14917 dump files): Add offload-tree.
14919 2018-05-03 Richard Biener <rguenther@suse.de>
14921 PR tree-optimization/85615
14922 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
14923 to loops not nested in BBs loop father to avoid creating multi-entry
14926 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14928 PR tree-optimization/70291
14929 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
14930 arguments. Change return type to tree. Emit libcall as a new
14931 statement rather than replacing existing one when inplace_p is true.
14932 (expand_complex_multiplication_components): New function.
14933 (expand_complex_multiplication): Expand floating-point complex
14934 multiplication using the above.
14935 (expand_complex_division): Rename inner_type parameter to type.
14936 Update expand_complex_libcall call-site.
14937 (expand_complex_operations_1): Update expand_complex_multiplication
14938 and expand_complex_division call-sites.
14940 2018-05-02 Jakub Jelinek <jakub@redhat.com>
14943 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
14944 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
14945 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
14946 the highest significant bit of the shift count mask is clear. In
14947 check whether and[sq]i3 is needed verify that all significant bits
14948 of the shift count other than the highest are set.
14950 2018-05-02 Tom de Vries <tom@codesourcery.com>
14953 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
14954 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
14955 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
14956 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
14957 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
14958 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
14959 __builtin_goacc_parlevel_size.
14961 2018-05-02 Richard Biener <rguenther@suse.de>
14963 PR tree-optimization/85597
14964 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
14965 do not use split vect_get_vec_defs call but call vect_get_slp_defs
14968 2018-05-02 Tom de Vries <tom@codesourcery.com>
14971 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14972 dump files): Add ltrans-tree.
14974 2018-05-02 Tom de Vries <tom@codesourcery.com>
14977 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
14978 dump files): Add wpa-ipa.
14980 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
14982 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
14983 powerpc*-*-linux*paired* target.
14984 * config/rs6000/750cl.h: Delete.
14985 * config/rs6000/paired.h: Delete.
14986 * config/rs6000/paired.md: Delete.
14987 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
14989 * config/rs6000/rs6000-builtin.def: Remove paired float support.
14990 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
14991 comment. Remove paired float support.
14992 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
14993 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
14995 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
14996 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
14998 * config/rs6000/rs6000.c: Remove paired float support.
14999 (paired_expand_vector_init, paired_expand_vector_move,
15000 paired_emit_vector_compare, paired_emit_vector_cond_expr,
15001 (paired_expand_lv_builtin, paired_expand_stv_builtin,
15002 paired_expand_builtin, paired_expand_predicate_builtin,
15003 paired_init_builtins): Delete.
15004 * config/rs6000/rs6000.h: Remove paired float support.
15005 * config/rs6000/rs6000.md: Remove paired float support.
15006 (move_from_CR_ov_bit): Delete.
15007 * config/rs6000/rs6000.opt (mpaired): Delete.
15008 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
15009 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
15011 2018-05-02 Richard Biener <rguenther@suse.de>
15013 PR middle-end/85567
15014 * gimplify.c (gimplify_save_expr): When in SSA form allow
15015 SAVE_EXPRs to compute to SSA vars.
15017 2018-05-02 Jakub Jelinek <jakub@redhat.com>
15020 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15021 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15022 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
15023 clobber operands[2], instead use a new pseudo. Formatting fixes.
15025 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
15027 PR tree-optimization/85586
15028 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
15029 exit early for statements in the same group if the accesses are
15032 2018-05-02 Tom de Vries <tom@codesourcery.com>
15035 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
15038 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
15040 PR tree-optimization/85143
15041 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
15043 2018-05-01 Tom de Vries <tom@codesourcery.com>
15046 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
15047 not found" error message.
15049 2018-05-01 Tom de Vries <tom@codesourcery.com>
15052 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
15053 * vec.c (test_ordered_remove_if): New function.
15054 (vec_c_tests): Call test_ordered_remove_if.
15055 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
15056 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
15057 * tree-vect-patterns.c (vect_pattern_recog_1): Use
15058 VEC_ORDERED_REMOVE_IF.
15060 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15062 PR tree-optimization/82665
15063 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
15064 pointer subtraction where arguments come from a memchr call.
15066 2018-05-01 Jakub Jelinek <jakub@redhat.com>
15068 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
15069 --push-state --as-needed and --pop-state instead of --as-needed and
15070 --no-as-needed if ld supports it.
15071 * configure: Regenerated.
15074 * doc/install.texi2html: Replace _002d with - and _002a with * in
15075 generated html files using sed.
15077 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15080 * gcc-rich-location.c (blank_line_before_p): New function.
15081 (use_new_line): New function.
15082 (gcc_rich_location::add_fixit_insert_formatted): New function.
15083 * gcc-rich-location.h
15084 (gcc_rich_location::add_fixit_insert_formatted): New function.
15086 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15088 * selftest.c (assert_streq): Rename "expected" and "actual" to
15089 "val1" and "val2". Extend NULL-handling to cover both inputs
15090 symmetrically, while still requiring both to be non-NULL for a pass.
15091 * selftest.h (assert_streq): Rename "expected" and "actual" to
15093 (ASSERT_EQ): Likewise.
15094 (ASSERT_EQ_AT): Likewise.
15095 (ASSERT_KNOWN_EQ): Likewise.
15096 (ASSERT_KNOWN_EQ_AT): Likewise.
15097 (ASSERT_NE): Likewise.
15098 (ASSERT_MAYBE_NE): Likewise.
15099 (ASSERT_MAYBE_NE_AT): Likewise.
15100 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
15101 the assertion to pass.
15102 (ASSERT_STREQ_AT): Likewise.
15104 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
15106 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
15107 interaction with -pie.
15109 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15111 * selftest.h: Fix alphabetization of per-source-file selftest
15114 2018-04-30 Jason Merrill <jason@redhat.com>
15116 PR c++/61982 - dead stores to destroyed objects.
15117 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
15120 2018-04-30 Jason Merrill <jason@redhat.com>
15122 * tree.c (build_clobber): New.
15123 * tree.h: Declare it.
15124 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
15126 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15128 * diagnostic-show-locus.c (layout::layout): Update for
15129 location_get_source_line returning a char_span.
15130 (struct char_span): Move to input.h.
15131 (struct correction): Update for fields in char_span becoming
15133 (struct source_line): Update for location_get_source_line
15134 returning a char_span.
15135 (layout::print_line): Likewise.
15136 * edit-context.c (edited_file::print_content): Likewise.
15137 (edited_file::print_diff_hunk): Likewise.
15138 (edited_file::print_run_of_changed_lines): Likewise.
15139 (edited_file::get_num_lines): Likewise.
15140 (edited_line::edited_line): Likewise.
15141 * final.c (asm_show_source): Likewise.
15142 * input.c (location_get_source_line): Convert return type
15143 from const char * to char_span, losing the final "line_len"
15145 (dump_location_info): Update for the above.
15146 (get_substring_ranges_for_loc): Likewise. Use a char_span
15147 when handling the literal within the line.
15148 (test_reading_source_line): Update for location_get_source_line
15149 returning a char_span.
15150 * input.h (class char_span): Move here from
15151 diagnostic-show-locus.c, converting from a struct to a class.
15152 Make data members private.
15153 (char_span::operator bool): New.
15154 (char_span::length): New.
15155 (char_span::get_buffer): New.
15156 (char_span::operator[]): New.
15157 (char_span::subspan): Make const.
15158 (char_span::xstrdup): New.
15159 (location_get_source_line): Convert return type from const char *
15160 to char_span, losing the final "line_size" param.
15162 2018-04-30 Jan Hubicka <jh@suse.cz>
15164 * lto-wrapper.c (ltrans_priorities): New static var.
15165 (cmp_priority): New.
15166 (run_gcc): Read priorities and if doing parallel build order
15167 the Makefile by them.
15169 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15171 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
15173 2018-04-30 Richard Biener <rguenther@suse.de>
15175 * tree-cfg.c (verify_address): Remove base argument, add
15176 flag whether to check TREE_ADDRESSABLE and do that.
15177 (verify_expr): Remove.
15178 (verify_types_in_gimple_reference): Add pieces from verify_expr.
15179 (verify_gimple_assign_single): Likewise.
15180 (verify_gimple_switch): Likewise.
15181 (verify_expr_location_1): Dereference tp once. Add (disabled)
15182 piece from verify_expr.
15183 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
15185 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15187 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
15189 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15191 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
15192 (small_data_pattern): Likewise.
15193 (arc_rewrite_small_data): Likewise.
15194 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
15195 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
15196 (get_symbol_alignment): New function.
15197 (legitimate_small_data_address_p): Likewise.
15198 (legitimate_scaled_address): Update, call
15199 legitimate_small_data_address_p.
15200 (output_sdata): New static variable.
15201 (arc_print_operand): Update how we handle small data operands.
15202 (arc_print_operand_address): Likewise.
15203 (arc_legitimate_address_p): Update, use
15204 legitimate_small_data_address_p.
15205 (arc_rewrite_small_data_p): Remove.
15206 (arc_rewrite_small_data_1): Likewise.
15207 (arc_rewrite_small_data): Likewise.
15208 (small_data_pattern): Likewise.
15209 (compact_sda_memory_operand): Update to use
15210 legitimate_small_data_address_p and get_symbol_alignment.
15211 (prepare_move_operands): Don't rewite sdata pattern.
15212 (prepare_extend_operands): Remove.
15213 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
15215 (zero_extendqisi2): Likewise.
15216 (zero_extendhisi2): Likewise.
15217 (extendqihi2): Likewise.
15218 (extendqisi2): Likewise.
15219 (extendhisi2): Likewise.
15220 (addsi3): Likewise.
15221 (subsi3): Likewise.
15222 (andsi3): Likewise.
15223 * config/arc/constraints.md (Usd): Change it to memory constraint.
15225 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15227 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
15228 as source of std instructions.
15229 * config/arc/arc.md (movsi_insn): Update pattern predicate to
15230 allow 6-bit constants as source for store instructions.
15231 (movdi_insn): Update instruction pattern to allow 6-bit constants
15232 as source for store instructions.
15234 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
15236 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
15238 2018-04-30 Nathan Sidwell <nathan@acm.org>
15239 Sandra Loosemore <sandra@codesourcery.com>
15241 * dumpfile.c (dump_open): Allow '-' for stdout.
15242 * doc/invoke.texi (Developer Options): Document dump filename
15243 determination early. Document stdin/stdout selection.
15245 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
15247 Microblaze Target: PIC data text relative
15249 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
15250 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
15252 * config/microblaze/microblaze.h (microblaze_constant_address_p):
15253 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
15254 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
15255 New addressing mode for data-text relative position indepenedent code.
15256 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
15257 'ADDRESS_SYMBOLIC_TXT_REL'.
15258 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
15259 (microblaze_legitimate_pic_operand): Exclude function calls from
15260 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
15261 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
15263 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15264 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15265 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
15266 for 'address + offset'.
15267 (microblaze_expand_prologue): Add new function prologue call for
15269 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
15270 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
15271 table in case of TARGET_PIC_DATA_TEXT_REL.
15272 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
15273 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
15274 Add new macros 'UNSPEC_TEXT',
15275 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
15276 + exclude function calls from 'UNSPEC_PLT' in case of data text
15278 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15279 new target hook for generating address diff vector tables in case of
15281 * doc/tm.texi : Regenerate.
15282 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
15283 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
15284 of addr diff vector generation.
15285 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15286 target hook definition.
15287 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
15288 Add default function for generate_pic_addr_diff_vec -> flag_pic.
15289 * doc/invoke.texi (Add new pic option): Add new microblaze pic
15290 option for data text relative.
15292 2018-04-30 Richard Biener <rguenther@suse.de>
15294 * tree-chrec.h (evolution_function_is_constant_p): Remove
15296 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
15298 2018-04-30 Richard Biener <rguenther@suse.de>
15301 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
15303 2018-04-30 Richard Biener <rguenther@suse.de>
15305 PR tree-optimization/28364
15306 PR tree-optimization/85275
15307 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
15308 copying first exit test.
15310 2018-04-28 Mark Wielaard <mark@klomp.org>
15312 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
15313 dwarf_version >= 5.
15314 (dwarf_AT): Handle DW_AT_addr_base.
15315 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
15317 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
15320 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
15321 (*ashl<dwi>3_doubleword_mask_1): Ditto.
15322 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
15323 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
15325 2018-04-28 Richard Biener <rguenther@suse.de>
15327 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
15328 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
15329 to reflect use. Only add interesting stmts.
15331 2018-04-27 Martin Jambor <mjambor@suse.cz>
15334 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
15335 the jump function allows for passing through aggregate values.
15337 2018-04-27 David Malcolm <dmalcolm@redhat.com>
15339 * input.h (in_system_header_at): Convert from macro to inline
15341 (from_macro_expansion_at): Likewise.
15342 (from_macro_definition_at): Likewise.
15344 2018-04-27 Jeff Law <law@redhat.com>
15346 * config.gcc: Mark tile* targets as deprecated/obsolete.
15348 2018-04-27 Richard Biener <rguenther@suse.de>
15350 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
15353 2018-04-27 Richard Biener <rguenther@suse.de>
15355 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
15357 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
15359 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
15360 with Yd constraint. Set "preferred_for_speed" attribute from
15361 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
15362 with Yd constraint.
15363 (*movdi_internal): Ditto.
15364 (movti_interunit splitters): Remove
15365 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
15366 (movdi_interunit splitters): Ditto.
15367 * config/i386/constraints.md (Ye): Remove.
15368 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
15370 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15373 * config/aarch64/constraints.md (Usg): Limit to 31.
15374 (Usj): Limit to 63.
15376 2018-04-27 Jakub Jelinek <jakub@redhat.com>
15378 PR tree-optimization/85529
15379 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
15380 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
15381 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
15382 zero extension or masking of the MSB bit.
15383 (optimize_range_tests): Add FIRST_BB argument, pass it through
15384 to optimize_range_tests_var_bound.
15385 (maybe_optimize_range_tests, reassociate_bb): Adjust
15386 optimize_range_tests callers.
15388 2018-04-26 Richard Biener <rguenther@suse.de>
15389 Jakub Jelinek <jakub@redhat.com>
15391 * cgraph.h (symbol_table): Just declare debug method here.
15392 * symtab.c (symbol_table::debug): Define.
15394 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
15396 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
15398 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
15400 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
15401 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
15402 (*movdi_internal): Substitute Yi and Yj constraint with x
15403 and Ym and Yn constraint with y constraint. Update "isa"
15404 attribute and set "preferred_for_speed" attribute from
15405 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
15406 (*movsi_internal): Ditto.
15407 (*movdf_internal): Ditto.
15408 (*movsf_internal): Ditto.
15409 (*zero_extendsidi2): Ditto.
15410 * config/i386/sse.md (vec_set<mode>_0): Ditto.
15411 (sse2_loadld): Ditto.
15412 (*vec_extract<ssevecmodelower>_0): Ditto.
15413 (*vec_extractv4si_0_zext_sse4): Ditto.
15414 (vec_concatv2di): Ditto.
15415 (*vec_dup<mode>): Ditto.
15416 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
15417 * config/i386/constraints.md (Yi): Remove.
15422 2018-04-26 Nathan Sidwell <nathan@acm.org>
15424 * dumpfile.c (dump_open): New.
15425 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
15426 (dump_finish): Detect stdio/stderr by value not name.
15428 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
15430 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
15432 2018-04-26 Tom de Vries <tom@codesourcery.com>
15435 * config/nvptx/nvptx.c (verify_neutering_jumps)
15436 (verify_neutering_labels): New function
15437 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
15439 2018-04-26 Tom de Vries <tom@codesourcery.com>
15442 * config/nvptx/nvptx.c (needs_neutering_p): New function.
15443 (nvptx_single): Use needs_neutering_p to skip over insns that do not
15446 2018-04-26 Richard Biener <rguenther@suse.de>
15447 Tom de Vries <tom@codesourcery.com>
15450 * lto-streamer-out.c (output_function): Fixup loops if required to match
15451 discovery done in the reader.
15453 2018-04-26 Richard Biener <rguenther@suse.de>
15455 PR tree-optimization/85116
15456 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
15457 have a loop exit from the single latch predecessor. Remove
15458 case of header with just condition.
15459 (ch_base::copy_headers): Exclude infinite loops from any
15461 (pass_ch::execute): Record exits.
15463 2018-04-26 Richard Biener <rguenther@suse.de>
15465 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
15466 prologue cost vector and pass it to vect_get_load_cost.
15467 (vect_get_peeling_costs_all_drs): Likewise.
15468 (vect_peeling_hash_get_lowest_cost): Likewise.
15469 (vect_enhance_data_refs_alignment): Likewise.
15471 2018-04-26 Richard Biener <rguenther@suse.de>
15473 PR middle-end/85450
15474 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
15475 checking of integer<->pointer conversions.
15476 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
15477 sign-/zero-extending pointer types.
15478 (expand_omp_for_static_chunk): Likewise.
15480 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
15481 Jean Lee <xiaoyur347@gmail.com>
15483 * config/mips/mips.c (mips_asan_shadow_offset): New function.
15484 (TARGET_ASAN_SHADOW_OFFSET): Define.
15485 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
15486 true for -fsanitize=address.
15488 2018-04-25 Mark Wielaard <mark@klomp.org>
15490 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
15493 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15495 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
15496 than "alu", remove explicit "memory" and "imm_disp" attributes.
15497 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
15499 PR middle-end/85414
15500 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
15501 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
15502 gen_lowpart_no_emit.
15504 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
15507 * config/i386/i386.c (ix86_expand_builtin): Change memory
15508 operand to XI, extend p0 to Pmode.
15509 * config/i386/i386.md: Change unspec volatile and operand
15510 1 mode to XI, change operand 0 mode to P.
15512 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15514 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
15515 GET_MODE_MASK before any checking.
15516 (nds32_can_use_bset_p): Likewise.
15517 (nds32_can_use_btgl_p): Likewise.
15519 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15521 * config/nds32/nds32-doubleword.md: New define_split pattern for
15522 illegal register number.
15524 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15526 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
15528 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15530 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
15532 2018-04-25 Richard Biener <rguenther@suse.de>
15534 * lto-streamer.h (LTO_major_version): Bump to 8.
15536 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15538 * BASE-VER: Set to 9.0.0.
15540 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
15542 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
15543 in __abskf2 and __powikf2.
15545 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15548 * config/aarch64/constraints.md (Usg, Usj): New constraints.
15549 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
15550 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
15551 Use the above on operand 2. Reindent.
15552 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
15554 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15557 * common/config/i386/i386-common.c (ix86_handle_option): Don't
15559 * config/i386/i386.opt (mcet): Removed.
15560 * doc/install.texi: Remove -mcet documentation.
15561 * doc/invoke.texi: Likewise.
15563 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15566 * doc/install.texi: Remove -mcet from bootstrap-cet.
15568 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15571 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
15572 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
15576 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
15577 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
15578 containing a CONST_VECTOR.
15580 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
15582 * doc/install.texi: Update newlib dependency for nvptx.
15584 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15587 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
15588 instead of INTVAL when shifting x left.
15590 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
15592 PR tree-optimization/85478
15593 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
15594 vect_grouped_store_supported for single element vectors.
15596 2018-04-24 Richard Biener <rguenther@suse.de>
15599 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
15600 load cost increase to the case of non-constant step.
15602 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15605 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
15606 destination if any_malformed_asm.
15608 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
15610 PR middle-end/85496
15611 * expr.c (store_field): In the bitfield case, if the value comes from
15612 a function call and is returned in registers by means of a PARALLEL,
15613 do not change the mode of the temporary unless BLKmode and VOIDmode.
15615 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
15617 PR rtl-optimization/85423
15618 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
15619 dependencies to debug insns when the previous insn is non-debug.
15621 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
15623 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
15624 enums into a single definition.
15625 (fls): Fix predicates and printing.
15628 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
15630 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
15631 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
15632 and short u6 immediate.
15633 (check_if_valid_sleep_operand): Remove.
15634 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
15636 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15638 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
15639 flag_always_save_lp condition.
15640 * config/nds32/nds32.opt (malways-save-lp): New option.
15642 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15644 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
15645 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
15646 * config/nds32/nds32.h
15647 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
15648 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
15650 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15652 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
15653 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
15655 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15656 Chung-Ju Wu <jasonwucj@gmail.com>
15658 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
15660 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
15661 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
15663 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15665 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
15667 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15669 * config/nds32/nds32-protos.h (nds32_data_alignment,
15670 nds32_local_alignment): Declare.
15671 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
15672 nds32_local_alignment): New functions.
15673 (TARGET_CONSTANT_ALIGNMENT): Define.
15674 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
15676 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15678 * config/nds32/nds32.c
15679 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
15680 (TARGET_MODES_TIEABLE_P): Likewise.
15682 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15684 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
15685 level Ofast and Og.
15687 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
15688 Chung-Ju Wu <jasonwucj@gmail.com>
15690 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15691 for unaligned access.
15692 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15694 * config/nds32/nds32-intrinsic.md: Likewise.
15695 * config/nds32/nds32_intrinsic.h: Likewise.
15696 * config/nds32/nds32.h (nds32_builtins): Likewise.
15697 * config/nds32/nds32.opt (munaligned-access): New option.
15698 * config/nds32/nds32.c (nds32_asm_file_start): Display
15699 flag_unaligned_access status.
15701 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
15703 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
15704 -mno-relax is present.
15705 * config/riscv/linux.h (LINK_SPEC): Ditto.
15707 2018-04-20 Martin Sebor <msebor@redhat.com>
15710 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
15712 (gimple_fold_builtin_stxcpy_chk): Same.
15713 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
15715 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
15718 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
15719 __powikf2 when long double is IEEE 128-bit.
15721 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
15723 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
15724 step to make sure stack always aligned.
15726 2018-04-20 Carl Love <cel@us.ibm.com>
15729 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
15730 size check for arg0.
15732 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
15733 Tom de Vries <tom@codesourcery.com>
15736 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
15737 Emit insns for calls too.
15738 (nvptx_find_par): Always look for worker-level predecessor insn.
15739 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
15741 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
15742 (nvptx_process_pars): Propagate frames for calls.
15744 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
15747 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
15749 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
15750 (ix86_handle_option): Don't handle OPT_mibt.
15751 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
15753 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
15755 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
15756 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
15757 (ix86_target_macros): Define __CET__ with flag_cf_protection
15758 for -fcf-protection.
15759 * config/i386/i386.c (isa2_opts): Remove -mibt.
15760 * config/i386/i386.h (TARGET_IBT): Removed.
15761 (TARGET_IBT_P): Likewise.
15762 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
15763 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15764 * config/i386/i386.opt (mcet): Update help message.
15765 (mshstk): Likewise.
15767 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
15768 -mcet as an alias for -mshstk.
15770 2018-04-20 Richard Biener <rguenther@suse.de>
15772 PR middle-end/85475
15773 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15774 complexity by forcing a single use of the multiply operand.
15776 2018-04-20 Martin Jambor <mjambor@suse.cz>
15779 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15780 recursion dependency to only apply to non-clones.
15782 2018-04-20 Martin Jambor <mjambor@suse.cz>
15785 * ipa-cp.c (create_specialized_node): Check that clones of
15786 self-recursive edges exist during IPA-CP.
15788 2018-04-19 Toon Moene <toon@moene.org>
15790 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15793 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15795 PR tree-optimization/85467
15796 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
15797 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
15798 VECTOR_CST element to type.
15800 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15803 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
15804 * config/i386/i386.md (builtin_setjmp_setup): Removed.
15805 (builtin_longjmp): Likewise.
15806 (save_stack_nonlocal): New pattern.
15807 (restore_stack_nonlocal): Likewise.
15809 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15812 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15813 Replace ASM_OUTPUT_LABEL with fprintf.
15815 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
15818 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
15819 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
15820 * config/i386/i386-c.c (ix86_target_macros_internal): Also
15821 define __IBT__ and __SHSTK__ for -fcf-protection.
15822 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
15824 (ix86_trampoline_init): Likewise.
15825 (x86_output_mi_thunk): Likewise.
15826 (ix86_notrack_prefixed_insn_p): Likewise.
15827 (ix86_option_override_internal): Don't disallow -fcf-protection.
15828 * config/i386/i386.md (rdssp<mode>): Also enable for
15830 (incssp<mode>): Likewise.
15831 (nop_endbr): Likewise.
15832 * config/i386/i386.opt (mcet): Change help message to built-in
15835 (mshstk): Likewise.
15836 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
15837 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
15838 enable CET built-in functions.
15840 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
15842 * common/config/i386/i386-common.c
15843 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
15844 OPTION_MASK_ISA_MOVDIRI_UNSET,
15845 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
15846 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
15847 * config.gcc (movdirintrin.h): New header.
15848 * config/i386/cpuid.h (bit_MOVDIRI,
15849 bit_MOVDIR64B): New bits.
15850 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
15852 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
15853 (VOID, PVOID, PCVOID)): New function types.
15854 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
15855 __builtin_ia32_directstoreu_u64,
15856 __builtin_ia32_movdir64b): New builtins.
15857 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
15858 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
15860 (ix86_valid_target_attribute_inner_p): Ditto.
15861 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
15862 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
15863 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
15864 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
15865 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
15866 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
15867 (movdiri<mode>, movdir64b_<mode>): New.
15868 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
15869 * config/i386/immintrin.h: Include movdirintrin.h.
15870 * config/i386/movdirintrin.h: New file.
15871 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
15873 2018-04-19 Richard Biener <rguenther@suse.de>
15875 PR middle-end/85455
15876 * cfg.c (clear_bb_flags): When loop state says we have
15877 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
15879 2018-04-19 Richard Biener <rguenther@suse.de>
15881 PR tree-optimization/84737
15882 * tree-vect-data-refs.c (vect_copy_ref_info): New function
15883 copying restrict info.
15884 (vect_setup_realignment): Use it.
15885 * tree-vectorizer.h (vect_copy_ref_info): Declare.
15886 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
15887 the first DR to all generated stores.
15888 (vectorizable_load): Likewise for loads.
15890 2018-04-19 Jakub Jelinek <jakub@redhat.com>
15892 PR tree-optimization/85446
15893 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
15894 the integral and pointer types to have the same precision.
15896 * doc/install.texi: Document --disable-cet being the default and
15899 2018-04-18 Martin Liska <mliska@suse.cz>
15901 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
15904 2018-04-18 Martin Liska <mliska@suse.cz>
15907 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15910 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15911 arguments if they are comparable.
15913 2018-04-18 Martin Liska <mliska@suse.cz>
15916 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15919 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15922 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
15925 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
15926 ENDBR after calling __morestack.
15928 2018-04-18 David Malcolm <dmalcolm@redhat.com>
15931 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
15932 by using gcc_base_ver to generate a gcc_driver_version, and use
15933 it when generating GCC_DRIVER_NAME.
15934 * configure: Regenerate.
15936 2018-04-18 Jakub Jelinek <jakub@redhat.com>
15939 * config.gcc: Obsolete powerpc*-*-*spe*.
15941 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15944 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
15945 (stabstr_D): Change type of unum from unsigned int to
15946 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
15949 2018-04-17 Jim Wilson <jimw@sifive.com>
15952 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
15953 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
15954 Set arg_pointer_offset after using pretend_args_size.
15956 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15958 PR rtl-optimization/85431
15959 * dse.c (record_store): Ignore zero width stores.
15962 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15963 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
15964 __builtin_stack_restore rather than after it.
15965 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
15966 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
15967 argument instead of virtual_dynamic_stack_rtx.
15969 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
15971 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
15973 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15974 Add note to error message to explain internal mapping of overloaded
15975 built-in function name to non-overloaded built-in function name.
15976 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
15979 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15982 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
15983 where the inputs overlap with the output.
15985 2018-04-17 Jakub Jelinek <jakub@redhat.com>
15987 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
15988 (=v, v) alternative and explicit "memory" attribute.
15989 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
15990 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15992 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
15993 "sselog1" type instead of "sselog".
15994 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
15995 "sselog". Remove explicit "memory" attribute.
15996 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
15997 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
15999 (vec_extract_hi_v32hi): Merge all alternatives into one, use
16000 "sselog1" type instead of "sselog". Remove explicit "memory"
16002 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
16003 use "sselog1" type instead of "sselog". Remove explicit "memory"
16005 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
16006 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16008 (vec_extract_hi_v64qi): Merge all alternatives into one, use
16009 "sselog1" type instead of "sselog". Remove explicit "memory"
16011 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
16012 use "sselog1" type instead of "sselog". Remove explicit "memory"
16016 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
16018 PR middle-end/85414
16019 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
16022 2018-04-17 Martin Jambor <mjambor@suse.cz>
16025 * ipa-cp.c (create_specialized_node): Call
16026 expand_all_artificial_thunks if necessary.
16028 2018-04-17 Martin Liska <mliska@suse.cz>
16031 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
16032 in message, remote space in between '_G' and '('.
16034 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16037 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
16038 avx512f_vmcmp<mode>3<round_saeonly_name>,
16039 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
16040 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
16041 avx512f_rndscale<mode><round_saeonly_name>,
16042 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
16043 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
16044 Use %<iptr>2 instead of %2 for -masm=intel.
16045 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
16046 avx512f_vcvttss2usi<round_saeonly_name>,
16047 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
16049 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
16050 avx512f_vcvttsd2usi<round_saeonly_name>,
16051 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
16052 Use %q1 instead of %1 for -masm=intel.
16053 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
16054 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
16055 of %3 for -masm=intel.
16056 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
16058 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
16060 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
16062 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
16063 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
16065 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
16067 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
16068 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
16069 %g1 and one with %0 and %1.
16070 (avx512er_vmrcp28<mode><round_saeonly_name>,
16071 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
16072 %1 for -masm=intel.
16073 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
16074 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
16075 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
16076 of %0 and %{%4%} for -masm=intel.
16077 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
16078 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
16079 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
16080 order of %0 and %{%5%}%{z%} for -masm=intel.
16082 2018-04-17 Jan Hubicka <jh@suse.cz>
16085 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
16087 2018-04-17 Martin Liska <mliska@suse.cz>
16090 * multiple_target.c (create_dispatcher_calls): Set apostrophes
16091 for target_clone error message. Make default implementation
16092 clone to be a local declaration.
16093 (separate_attrs): Add new argument and check for an empty
16095 (expand_target_clones): Handle it.
16096 (ipa_target_clone): Make redirection just for target_clones
16099 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
16100 Tom de Vries <tom@codesourcery.com>
16102 PR middle-end/84955
16103 * omp-expand.c (expand_oacc_for): Add dummy false branch for
16104 tiled basic blocks without omp continue statements.
16106 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
16109 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
16110 vec_extract expression as having side effects to make sure it gets
16113 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
16116 * config/i386/i386.c (get_builtin_code_for_version): Check
16119 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
16122 * config.gcc: Support "skylake".
16123 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16125 * config/i386/i386.c (m_SKYLAKE): Define.
16126 (processor_target_table): Add "skylake".
16127 (ix86_option_override_internal): Add "skylake".
16128 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
16129 PROCESSOR_CANNONLAKE.
16130 (get_builtin_code_for_version): Fix priority for
16131 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
16132 PROCESSOR_SKYLAKE-AVX512.
16133 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
16134 (processor_type): Add PROCESSOR_SKYLAKE.
16136 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
16137 Jason Merrill <jason@redhat.com>
16140 * convert.c (convert_to_integer_1): Use direct recursion for
16141 enumeral types and types with a precision less than the number
16142 of bits in their mode.
16144 2018-04-16 Julia Koval <julia.koval@intel.com>
16147 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
16148 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
16150 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
16153 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
16154 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
16155 and -mno-direct-move.
16157 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
16160 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
16161 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
16162 Ensure negative shifts result in {0}.
16164 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
16166 PR rtl-optimization/79916
16167 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
16168 regs (if any) to define how to gnerate SD moves when LRA is in
16171 2018-04-13 Jakub Jelinek <jakub@redhat.com>
16173 PR rtl-optimization/85393
16174 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
16175 * except.c (expand_dw2_landing_pad_for_region): Make static.
16176 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
16177 a label and unconditional jump to old_bb, rather than
16178 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
16181 PR rtl-optimization/85376
16182 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
16183 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
16184 instead of a specific value.
16186 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
16187 Bin Cheng <bin.cheng@arm.com>
16189 PR tree-optimization/82965
16190 PR tree-optimization/83991
16191 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
16192 by_profile_only parameter.
16193 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
16194 information if the loop was predicted to iterate too many times.
16195 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
16197 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
16200 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
16203 2018-04-13 Martin Liska <mliska@suse.cz>
16204 Jakub Jelinek <jakub@redhat.com>
16206 PR middle-end/81657
16207 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
16208 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
16209 * builtins.c (expand_builtin_memory_copy_args): Use
16210 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
16211 handle dest_addr == pc_rtx.
16213 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
16216 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
16217 asked to not generate direct moves.
16218 (fix_trunc<mode>si2_stfiwx): Similar.
16219 (fix_trunc<mode>si2_internal): Similar.
16221 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16224 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
16225 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
16226 lookup if dest in some wider mode is known to be const0_rtx and
16227 if so, record permanent equivalence for it to be ZERO_EXTEND of
16228 the narrower mode destination.
16230 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
16232 * lto-streamer-out.c (output_function): Revert 259346.
16233 * omp-expand.c (expand_oacc_for): Likewise.
16235 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
16237 PR rtl-optimization/85354
16238 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
16239 * sel-sched.c (sel_global_init): ... here.
16241 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
16244 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
16245 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
16246 mode for PE-COFF targets.
16247 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
16248 (i386_pe_asm_lto_end): Likewise.
16249 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
16250 (TARGET_ASM_LTO_END): Likewise.
16251 * config/i386/winnt.c (saved_debug_info_level): New static variable.
16252 (i386_pe_asm_lto_start): New function.
16253 (i386_pe_asm_lto_end): Likewise.
16255 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
16256 Richard Biener <rguenther@suse.de>
16258 PR middle-end/84955
16259 * lto-streamer-out.c (output_function): Fix CFG loop state before
16261 * omp-expand.c (expand_oacc_for): Handle calls to internal
16262 functions like regular functions.
16264 2018-04-12 Richard Biener <rguenther@suse.de>
16267 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
16268 for the early LTO debug to properly generate references to it
16269 during DIE emission. Do not re-use that for the skeleton for
16271 (dwarf2out_early_finish): Likewise.
16273 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16276 * config/i386/sse.md
16277 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
16278 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
16279 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
16280 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
16281 and output is a reg, avoid creating invalid lowpart subreg, but
16282 instead split into a 512-bit move. Don't split if not AVX512VL,
16283 input is xmm16+ reg and output is a mem.
16284 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
16285 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
16286 xmm16+ reg and output is a mem.
16288 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16290 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
16291 also for flag_dwarf2_cfi_asm.
16293 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16295 PR rtl-optimization/85342
16296 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
16297 a bool scalar var inside of the loop instead. Don't try to update
16298 recog_data.operand after failed apply_change_group.
16300 2018-04-12 Tom de Vries <tom@codesourcery.com>
16303 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
16304 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
16305 array with flexible array member as array without given dimension.
16306 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
16307 argument for undefined param to true.
16309 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
16312 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16313 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
16314 from PowerPC section.
16315 * config/rs6000/sysv4.opt (mcall-): Improve help text.
16316 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
16317 help text that is too long.
16318 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
16319 help text that is too long.
16320 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
16321 help text that is too long.
16323 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
16325 * config/alpha/alpha.md (stack_probe_internal): Rename
16326 from "probe_stack". Update all callers.
16328 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16330 PR rtl-optimization/84566
16331 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
16332 sched_macro_fuse_insns.
16334 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16337 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
16338 (compute_block_dependences): ... from here.
16340 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16342 PR tree-optimization/85331
16343 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
16344 from int to HOST_WIDE_INT.
16346 2018-04-11 Martin Jambor <mjambor@suse.cz>
16349 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
16350 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
16351 not the same as the source val.
16352 (cgraph_edge_brings_value_p): New parameter.
16353 (gather_edges_for_value): Pass destination value to
16354 cgraph_edge_brings_value_p.
16355 (perhaps_add_new_callers): Likewise.
16356 (get_info_about_necessary_edges): Likewise and exclude values brought
16357 only by self-recursive edges.
16358 (create_specialized_node): Redirect only clones of self-calling edges.
16359 (+self_recursive_pass_through_p): New function.
16360 (find_more_scalar_values_for_callers_subset): Use it.
16361 (find_aggregate_values_for_callers_subset): Likewise.
16362 (known_aggs_to_agg_replacement_list): Removed.
16363 (decide_whether_version_node): Re-calculate known constants for all
16364 remaining context clones.
16366 2018-04-11 Richard Biener <rguenther@suse.de>
16369 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
16370 from early DWARF output.
16371 (dwarf2out_early_finish): Output line info unconditionally into
16372 early DWARF and add reference to it.
16374 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16377 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
16378 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
16379 other than V2DFmode using iptr mode attribute.
16380 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
16382 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16384 PR rtl-optimization/84659
16385 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
16387 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16390 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
16392 (output_loc_list): Pass address of a dummy size variable even in the
16393 locview handling loop.
16394 (index_location_lists): Add comment on why skip_loc_list_entry can't
16397 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
16400 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
16403 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
16406 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16407 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
16408 and -mstring-compare-inline-limit.
16410 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16413 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
16414 for stack clash protection in a register whenever we need it to be in
16417 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16419 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
16420 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
16422 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16425 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
16427 (mlong-double-): Ditto.
16428 * config/rs6000/sysv4.opt (msdata=): Ditto.
16429 (mtls-size=): Ditto.
16431 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16433 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16434 erroneous entries for
16435 "vector int vec_ldl (int, long int *)", and
16436 "vector unsigned int vec_ldl (int, unsigned long int *)".
16437 Add comments and entries for
16438 "vector bool char vec_ldl (int, bool char *)",
16439 "vector bool short vec_ldl (int, bool short *)",
16440 "vector bool int vec_ldl (int, bool int *)",
16441 "vector bool long long vec_ldl (int, bool long long *)",
16442 "vector pixel vec_ldl (int, pixel *)",
16443 "vector long long vec_ldl (int, long long *)",
16444 "vector unsigned long long vec_ldl (int, unsigned long long *)".
16445 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
16446 type tree bool_long_long_type_node and correct definition of
16447 bool_V2DI_type_node to make reference to this new type tree.
16448 (rs6000_mangle_type): Replace erroneous reference to
16449 bool_long_type_node with bool_long_long_type_node.
16450 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
16451 comments to emphasize sign distinctions for char and int types and
16452 replace RS6000_BTI_bool_long constant with
16453 RS6000_BTI_bool_long_long constant. Also add comment to restrict
16454 use of RS6000_BTI_pixel.
16455 (bool_long_type_node): Remove this macro definition.
16456 (bool_long_long_type_node): New macro definition
16458 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16460 PR rtl-optimization/85300
16461 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
16462 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
16463 simplify_unary_operation fails.
16465 2018-04-10 Martin Liska <mliska@suse.cz>
16467 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
16468 cgraph_edge and ipa_ref.
16470 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16474 * config/i386/sse.md
16475 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
16476 computation of the VEC_MERGE selector from mask.
16477 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
16478 Fix decoding of the VEC_MERGE selector into mask.
16480 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16482 PR tree-optimization/85286
16483 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
16485 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16487 * final.c (final_1): Set insn_last_address as well as
16488 insn_current_address.
16490 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16493 * explow.c (emit_stack_probe): Call validize_mem on memory location
16494 before passing it to gen_probe_stack. Create address operand and
16495 legitimize it for the probe_stack_address case.
16497 2018-04-09 Jan Hubicka <jh@suse.cz>
16500 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
16501 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
16502 * tree.c (free_lang_data_in_type): Fix handling of binfos;
16504 (free_lang_data): Rebuild type inheritance graph.
16506 2018-04-09 Martin Sebor <msebor@redhat.com>
16508 * invoke.texi (-finline-small-functions): Mention other optimization
16510 (-findirect-inlining, -fpartial-inlining): Same.
16511 (-finline-functions-called-once): Same.
16512 (-freorder-blocks-and-partition): Same.
16514 2018-04-09 Jan Hubicka <jh@suse.cz>
16517 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
16518 jumps; choose last target that matches the criteria (i.e.
16519 no partition changes for non-crossing jumps).
16520 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
16521 support for redirecting crossing jumps to non-crossing.
16523 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
16525 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
16526 also for naked functions.
16528 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
16530 * config/arc/arc.md (add_shift): New pattern.
16531 (add_shift2): Likewise.
16532 (sub_shift): Likewise.
16533 (sub_shift_cmp0_noout): Likewise.
16534 (compare_si_ashiftsi): Likewise.
16535 (xbfu_cmp0_noout): New combine pattern.
16536 (xbfu_cmp0"): Likewise.
16537 (movsi_set_cc_insn): Place the predicable variant first.
16538 (commutative_binary_cmp0_noout): Remove clobber.
16539 (commutative_binary_cmp0): New pattern.
16540 (noncommutative_binary_cmp0): Likewise.
16541 (noncommutative_binary_cmp0_noout): Likewise.
16542 (noncommutative_binary_comparison_result_used): Removed.
16543 (rsub_cmp0): New pattern.
16544 (rsub_cmp0_noout): Likewise.
16545 (extzvsi): Changed, keep only meaningful variants.
16546 (SQH, SEZ): New iterators.
16547 (SQH_postfix): New mode attribute.
16548 (SEZ_prefix): New code attribute.
16549 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
16550 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
16551 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
16552 of numerical value.
16553 (noncommutative_operator): Check the availability of barrel
16556 2018-04-09 Richard Biener <rguenther@suse.de>
16558 PR tree-optimization/85284
16559 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
16560 Only use the niter constraining form of simple_iv when the exit
16561 is always executed.
16563 2018-04-09 Tom de Vries <tom@codesourcery.com>
16566 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16567 (define_expand "*memory_barrier"): New define_expand.
16568 (define_insn "memory_barrier"): New insn.
16570 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16572 PR rtl-optimization/80463
16573 PR rtl-optimization/83972
16574 PR rtl-optimization/83480
16576 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
16577 correct producer for the insn.
16578 (tidy_control_flow): Fixup seqnos in case of debug insns.
16580 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16582 PR rtl-optimization/83913
16584 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
16585 different sched-times when merging exprs.
16587 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16589 PR rtl-optimization/83962
16591 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
16592 tidy_fallthru_edge and tidy_control_flow.
16594 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16596 PR rtl-optimization/83530
16598 * sel-sched.c (force_next_insn): New global variable.
16599 (remove_insn_for_debug): When force_next_insn is true, also leave only
16600 next insn in the ready list.
16601 (sel_sched_region): When the region wasn't scheduled, make another pass
16602 over it with force_next_insn set to 1.
16604 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
16606 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
16608 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16609 for interrupt control.
16610 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16611 functions for interrupt control.
16612 * config/nds32/nds32-intrinsic.md: Likewise.
16613 * config/nds32/nds32_intrinsic.h: Likewise.
16614 * config/nds32/nds32.h (nds32_builtins): Likewise.
16616 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
16618 * config/nds32/nds32.c (nds32_init_machine_status,
16619 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
16620 strict_aligned_p field.
16621 (nds32_expand_to_rtl_hook): New function.
16622 (TARGET_EXPAND_TO_RTL_HOOK): Define.
16623 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
16625 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16626 Chung-Ju Wu <jasonwucj@gmail.com>
16628 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
16629 * config/nds32/nds32-n7.md: New file.
16630 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
16631 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
16633 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
16634 * config/nds32/nds32.md (pipeline_model): Add n7.
16635 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
16636 * config/nds32/pipelines.md: Include n7 settings.
16638 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16639 Chung-Ju Wu <jasonwucj@gmail.com>
16641 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
16642 * config/nds32/nds32-e8.md: New file.
16643 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
16644 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
16646 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
16647 * config/nds32/nds32.md (pipeline_model): Add e8.
16648 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
16649 * config/nds32/pipelines.md: Include e8 settings.
16651 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16652 Chung-Ju Wu <jasonwucj@gmail.com>
16654 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
16655 * config/nds32/nds32-n8.md: New file.
16656 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
16657 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
16659 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
16660 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
16661 * config/nds32/nds32.md (pipeline_model): Add n8.
16662 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
16663 * config/nds32/pipelines.md: Include n8 settings.
16665 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16666 Chung-Ju Wu <jasonwucj@gmail.com>
16668 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
16669 * config/nds32/nds32-n9-2r1w.md: New file.
16670 * config/nds32/nds32-n9-3r2w.md: New file.
16671 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
16672 nds32_register_ports): New or modify for cpu n9.
16673 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
16675 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
16676 * config/nds32/nds32-utils.c: New file.
16677 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
16678 TARGET_MUL_SLOW): Define.
16679 * config/nds32/nds32.md (pipeline_model): New attribute.
16680 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
16681 New options that support cpu n9.
16682 * config/nds32/pipelines.md: Include n9 settings.
16683 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
16685 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
16687 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
16688 information if necessary.
16689 (output_cond_branch_compare_zero): Likewise.
16690 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
16691 (nds32_target_alignment): Refine for alignment.
16692 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
16693 (FUNCTION_BOUNDARY): Modify.
16694 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
16696 * config/nds32/nds32.opt (malways-align, malign-functions): New.
16698 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
16700 * config/nds32/constants.md (unspec_volatile_element): Add values for
16701 TLB operation and data prefetch.
16702 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16703 functions for TLB operation and data prefetch.
16704 * config/nds32/nds32-intrinsic.md: Likewise.
16705 * config/nds32/nds32_intrinsic.h: Likewise.
16706 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
16707 (nds32_print_operand): Likewise.
16708 * config/nds32/nds32.h (nds32_builtins): Likewise.
16710 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
16711 Andrew Pinski <pinsika@gcc.gnu.org>
16713 PR middle-end/82976
16714 * match.pd: Use constant_boolean_node of correct type instead of
16715 boolean_true_node or boolean_false_node for simplifying
16716 pointer comparisons to zero.
16718 2018-04-07 Jakub Jelinek <jakub@redhat.com>
16720 PR tree-optimization/80021
16721 * tree.c (verify_type_variant): Make error call in verify_variant_match
16722 translatable and remove final full stop.
16724 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16726 * config/nds32/constants.md (unspec_volatile_element): Add
16727 UNSPEC_VOLATILE_EH_RETURN.
16728 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
16729 nds32_output_stack_pop): Support dwarf exception handling process.
16730 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
16731 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
16732 exception handling process.
16733 (nds32_compute_stack_frame): Likewise.
16734 (nds32_return_addr_rtx): Likewise.
16735 (nds32_initial_elimination_offset): Likewise.
16736 (nds32_expand_prologue): Likewise.
16737 (nds32_expand_epilogue): Likewise.
16738 (nds32_dynamic_chain_address): New function.
16739 * config/nds32/nds32.h (machine_function): Add fields for dwarf
16740 exception handling.
16741 (DYNAMIC_CHAIN_ADDRESS): Define.
16742 (EH_RETURN_DATA_REGNO): Define.
16743 (EH_RETURN_STACKADJ_RTX): Define.
16744 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
16745 patterns for dwarf exception handling.
16747 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16749 * config/nds32/nds32.h: Clean up obsolete macros.
16751 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16753 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16754 Add enum values for particular instructions.
16755 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16756 particular intrinsic functions.
16757 * config/nds32/nds32-intrinsic.md: Likewise.
16758 * config/nds32/nds32_intrinsic.h: Likewise.
16759 * config/nds32/nds32.h (nds32_builtins): Likewise.
16760 * config/nds32/nds32.md (type): Add pbsad and pbsada.
16761 (btst, ave): New patterns for particular instructions.
16763 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16765 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16766 Add enum values for atomic load/store and memory sync.
16767 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16769 * config/nds32/nds32-intrinsic.md: Likewise.
16770 * config/nds32/nds32_intrinsic.h: Likewise.
16771 * config/nds32/nds32.h (nds32_builtins): Likewise.
16773 2018-04-07 Jakub Jelinek <jakub@redhat.com>
16775 PR tree-optimization/85257
16776 * fold-const.c (native_encode_vector): If not all elts could fit
16777 and off is -1, return 0 rather than offset.
16778 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16779 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16780 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
16781 adjust buffer in native_interpret_expr call.
16783 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16785 * config/nds32/constants.md (unspec_volatile_element): Add cache
16786 control enum values.
16787 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16788 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16789 * config/nds32/nds32.c (nds32_cctl_names): New.
16790 (nds32_print_operand): Handle cache control register names.
16791 * config/nds32/nds32.h (nds32_builtins): New enum values.
16792 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16794 * config/nds32/nds32.md (type): Add mmu.
16795 * config/nds32/pipelines.md (simple_insn): Add mmu.
16797 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16799 * config/nds32/nds32.md (type): Remove call.
16800 * config/nds32/pipelines.md (simple_insn): Likewise.
16802 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16804 * config/nds32/constants.md (unspec_volatile_element): Add
16805 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
16806 UNSPEC_VOLATILE_FMFCFG.
16807 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
16808 description for fmfcfg and fmfcsr.
16809 (bdesc_1arg): Add fmtcsr.
16810 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
16811 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
16812 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
16813 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
16814 unspec_fmfcfg): New patterns.
16815 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
16816 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
16817 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
16818 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
16819 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
16820 __nds32__fmfcfg): Define.
16822 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
16824 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
16825 intrinsic register names.
16826 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
16827 intrinsic register enum values and macros.
16829 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
16831 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
16832 for load/store addressing form.
16833 (nds32_print_operand_address): Likewise.
16835 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
16838 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
16839 based on LABEL_REF. Remove useless assertion.
16840 (pic_address_needs_scratch): Fix formatting.
16841 (sparc_legitimize_pic_address): Minor tweaks.
16842 (sparc_delegitimize_address): Adjust assertion accordingly.
16843 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
16844 into symbolic_operand.
16845 (movsi_high_pic_label_ref): Likewise.
16846 (movsi_lo_sum_pic_label_ref): Likewise.
16847 (movdi_pic_label_ref): Likewise.
16848 (movdi_high_pic_label_ref): Likewise.
16849 (movdi_lo_sum_pic_label_ref): Likewise.
16851 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
16853 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
16854 custom LIB_SPEC setup.
16856 2018-04-06 Ruslan Bukin <br@bsdpad.com>
16857 Kito Cheng <kito.cheng@gmail.com>
16859 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
16860 * config/riscv/freebsd.h: New.
16862 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16864 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
16865 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
16868 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16869 Kito Cheng <kito.cheng@gmail.com>
16871 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
16872 nds32_output_call, nds32_symbol_binds_local_p): New functions.
16873 * config/nds32/nds32-protos.h (nds32_output_call,
16874 nds32_output_return): Declare.
16875 * config/nds32/nds32.md: Refine all the call and return patterns.
16877 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16880 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
16881 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
16883 PR rtl-optimization/84872
16884 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
16885 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
16886 EDGE_CROSSING edge.
16888 2018-04-06 Tamar Christina <tamar.christina@arm.com>
16890 * expr.c (copy_blkmode_to_reg): Revert 254862.
16891 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
16893 2018-04-06 Richard Biener <rguenther@suse.de>
16895 PR middle-end/85244
16896 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
16897 after seeing a component reference with an adjacent field. Treat
16898 refs to arrays at struct end of external decls similar to
16899 refs to unconstrained commons.
16901 2018-04-06 Jakub Jelinek <jakub@redhat.com>
16904 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
16905 look through SAVE_EXPRs with non-side-effects argument. Adjust
16907 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
16910 2018-04-06 Richard Biener <rguenther@suse.de>
16912 PR middle-end/85180
16913 * alias.c (find_base_term): New wrapper around find_base_term
16914 unwinding CSELIB_VAL_PTR changes.
16915 (find_base_term): Do not restore CSELIB_VAL_PTR during the
16918 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16920 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
16922 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
16923 constant definitions.
16924 ("nop"): lr 0,0 -> nopr r0
16925 ("nop_lr0", "nop_lr1"): New insn definitions.
16927 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
16929 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
16930 NDS32_V3PUSH_AVAILABLE_P macro.
16932 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
16933 Chung-Ju Wu <jasonwucj@gmail.com>
16935 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
16936 (nds32*-*-*): Add float and fpu_config into supported_defaults.
16937 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
16938 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
16939 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
16940 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
16941 * config/nds32/constraints.md: New constraints and checking for hard
16942 float configuration.
16943 * config/nds32/iterators.md: New mode iterator and attribute for hard
16944 float configuration.
16945 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
16947 * config/nds32/nds32-fpu.md: New file.
16948 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
16949 deal with hard float code generation.
16950 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
16952 (abi_type, float_reg_number): New enum type.
16953 * config/nds32/nds32-predicates.c: New predicates for hard float.
16954 * config/nds32/nds32-protos.h: Declare functions for hard float.
16955 * config/nds32/nds32.c: Implementation for hard float configuration.
16956 * config/nds32/nds32.h: Definitions for hard float configuration.
16957 * config/nds32/nds32.md: Include hard float machine description and
16958 modify patterns for hard float configuration.
16959 * config/nds32/nds32.opt: New options for hard float configuration.
16960 * config/nds32/predicates.md: New predicates for hard float
16963 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16965 * common/config/nds32/nds32-common.c
16966 (nds32_option_optimization_table): Enable -mreleax-hint by default.
16968 2018-04-05 Jakub Jelinek <jakub@redhat.com>
16970 PR middle-end/85195
16971 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
16972 CONSTRUCTOR_ELT (ctor, ...)->value.
16974 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
16977 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
16979 2018-04-05 Tom de Vries <tom@codesourcery.com>
16982 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
16985 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16986 Kito Cheng <kito.cheng@gmail.com>
16988 * config/nds32/constraints.md (U33): Fine-tune checking condition.
16989 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
16990 * config/nds32/nds32.h (nds32_16bit_address_type): Add
16991 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
16993 2018-04-05 Shiva Chen <shiva0217@gmail.com>
16994 Kito Cheng <kito.cheng@gmail.com>
16996 * config/nds32/constraints.md (Ufe): New memory constraint.
16997 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
16998 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
16999 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
17001 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
17002 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
17004 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17006 * config/nds32/nds32.md: Use optimize_size in the condition for
17007 alu-shift instructions.
17009 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17011 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
17013 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17015 * config/nds32/nds32.md (negsi2): Refine pattern.
17017 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
17018 Chung-Ju Wu <jasonwucj@gmail.com>
17020 * config/nds32/iterators.md (shift_rotate): New code iterator.
17021 (shift): New code attribute.
17022 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
17023 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
17024 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
17025 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
17026 bit-wise operations.
17027 (andsi3, *andsi3): Ditto.
17028 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
17029 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
17030 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
17031 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
17032 nds32_ior_operand, nds32_xor_operand): New predicates.
17034 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17036 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
17037 (addsi3, subsi3): ... this.
17039 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17041 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
17043 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17045 * config/nds32/nds32.md: Adjust indention.
17047 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
17049 * config/nds32/nds32.md (feature): New attribute.
17051 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17053 * config/nds32/nds32.md (subtype): New attribute.
17055 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
17058 * config/arm/arm-builtins.c (arm_expand_builtin): Change
17059 expansion to perform a bitwise AND of the argument followed by a
17060 boolean negation of the result.
17062 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
17064 PR rtl-optimization/84878
17065 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
17066 the basic block. Assert the use reference is not artificial and that
17067 it has an associated insn.
17069 2018-04-04 Michael Matz <matz@suse.de>
17071 * builtins.c (compute_objsize): Pass correct operand
17072 to array_at_struct_end_p.
17074 2018-04-04 Richard Biener <rguenther@suse.de>
17077 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
17078 from contexts for DINFO_LEVEL_TERSE and below.
17080 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17082 * config/nds32/nds32-doubleword.md (move_<mode>): Require
17083 resiter_operand condition.
17084 * config/nds32/nds32.md (*move<mode>): Ditto.
17086 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17087 Monk Chiang <sh.chiang04@gmail.com>
17089 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
17091 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17093 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
17095 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17096 Kito Cheng <kito.cheng@gmail.com>
17098 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
17099 nds32_cond_code_str, output_cond_branch,
17100 output_cond_branch_compare_zero, nds32_expand_cbranch,
17101 nds32_expand_cstore, nds32_expand_movcc,
17102 nds32_output_cbranchsi4_equality_zero,
17103 nds32_output_cbranchsi4_equality_reg,
17104 nds32_output_cbranchsi4_equality_reg_or_const_int,
17105 nds32_output_cbranchsi4_greater_less_zero: New functions.
17106 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
17107 nds32_expand_cstore, nds32_expand_movcc,
17108 nds32_output_cbranchsi4_equality_zero,
17109 nds32_output_cbranchsi4_equality_reg,
17110 nds32_output_cbranchsi4_equality_reg_or_const_int,
17111 nds32_output_cbranchsi4_greater_less_zero): Declare.
17112 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
17113 nds32_rimm11s_operand): New predicates.
17114 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
17115 * config/nds32/nds32.md: Rewrite all the branch and conditional move
17118 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17120 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
17121 * config/nds32/nds32.md: Ditto.
17122 * config/nds32/pipelines.md: Ditto.
17124 2018-04-04 Richard Biener <rguenther@suse.de>
17126 PR tree-optimization/85168
17127 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
17128 propagating abnormals.
17130 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17132 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
17134 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17135 Kito Cheng <kito.cheng@gmail.com>
17137 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
17138 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
17139 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
17140 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
17141 * config/nds32/nds32.md (sibcall_internal): New.
17142 (sibcall_register): Remove.
17143 (sibcall_immediate): Remove.
17144 (sibcall_value_internal): New.
17145 (sibcall_value_register): Remove.
17146 (sibcall_value_immediate): Remove.
17147 * config/nds32/predicates.md (nds32_general_register_operand): New.
17148 (nds32_call_address_operand): New.
17150 2018-04-03 Jakub Jelinek <jakub@redhat.com>
17152 PR rtl-optimization/85167
17153 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
17154 bb_defs if *split_p, instead preinitialize it to NULL.
17156 PR tree-optimization/85156
17157 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
17158 evaluating the argument multiple times.
17160 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
17162 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
17164 (_mm_cvtpd_ps): Likewise.
17165 (_mm_cvttpd_epi32): Likewise.
17166 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
17167 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
17168 vector, pixel, and bool following altivec.h include.
17170 2018-04-03 Martin Sebor <msebor@redhat.com>
17172 * doc/extend.texi (Common Function Attributes): Clarify.
17173 (const attribute): Likewise.
17174 (pure attribute): Likewise.
17176 2018-04-03 Jakub Jelinek <jakub@redhat.com>
17179 * config/i386/i386.c (ix86_expand_vector_set): Use
17180 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
17182 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
17184 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
17185 instructions when changing rounding bits to preserve precision bits
17186 in the x87 control word.
17188 2018-04-03 Martin Liska <mliska@suse.cz>
17190 PR tree-optimization/82491
17191 * rtl.h (strip_offset_and_add): Replace += suboffset with
17192 poly_uint64 () + suboffset.
17194 2018-03-29 Martin Liska <mliska@suse.cz>
17195 Martin Jambor <mjambor@suse.cz>
17198 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
17199 param_type is not an integral or pointer type.
17201 2018-04-03 Richard Biener <rguenther@suse.de>
17203 * sese.h (recompute_all_dominators): Remove.
17205 2018-04-02 Martin Sebor <msebor@redhat.com>
17207 * doc/invoke.texi (-Wrestrict): Fix typos.
17209 2018-04-02 Jim Wilson <jimw@sifive.com>
17211 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
17212 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
17213 (<optab>di3, <optab>si3_extend): Likewise.
17214 (<optab>si3_mask, <optab>si3_mask_1): New.
17215 (<optab>di3_mask, <optab>di3_mask_1): New.
17216 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
17217 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
17218 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
17220 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
17222 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
17225 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
17227 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
17228 (nds32_canonicalize_comparison): New function.
17230 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17231 Kito Cheng <kito.cheng@gmail.com>
17232 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17234 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
17235 * config/nds32/constants.md (unspec_volatile_element): Add
17236 UNSPEC_VOLATILE_RELAX_GROUP.
17237 * config/nds32/nds32-relax-opt.c: New file.
17238 * config/nds32/nds32-predicates.c
17239 (nds32_symbol_load_store_p): New function.
17240 * config/nds32/nds32-protos.h
17241 (nds32_symbol_load_store_p): Declare function.
17242 (make_pass_nds32_relax_opt): Declare new rtl pass function.
17243 * config/nds32/nds32.c
17244 (nds32_register_pass): New function to register pass.
17245 (nds32_register_passes): New function to register passes.
17246 * config/nds32/nds32.md (relax_group): New pattern.
17247 * config/nds32/nds32.opt (mrelax-hint): New option.
17248 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
17250 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
17252 * config/nds32/t-nds32: Modify files dependency.
17254 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17256 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
17257 (PROFILE_HOOK): Define its implementation.
17259 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17261 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
17262 type and 32-bit size.
17264 2018-04-01 Jakub Jelinek <jakub@redhat.com>
17266 PR middle-end/85090
17267 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
17268 (V_128_256): New mode iterator.
17269 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
17270 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
17271 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
17273 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
17274 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
17276 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
17279 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
17280 NaN inputs correctly.
17282 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
17285 * config/rs6000/vsx.md (??r): New mode attribute.
17286 (*vsx_mov<mode>_64bit): Use it.
17287 (*vsx_mov<mode>_32bit): Likewise.
17289 2018-03-30 Martin Sebor <msebor@redhat.com>
17291 PR tree-optimization/84818
17292 * builtins.c (check_access): Use warning_n.
17294 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17297 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
17299 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
17302 2018-03-30 Julia Koval <julia.koval@intel.com>
17305 * x86-tune.def (movx, partial_reg_dependency): Enable for
17308 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
17310 PR inline-asm/84985
17311 * lra-constraints.c (process_alt_operands): Move setting
17312 this_alternative_matches below.
17314 2018-03-29 Martin Liska <mliska@suse.cz>
17317 * doc/invoke.texi: Document how LTO works with debug info.
17318 Describe auto-load support of binutils. Mention 'x86-64'
17319 as valid option value of -march option.
17321 2018-03-29 Jakub Jelinek <jakub@redhat.com>
17323 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
17326 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
17327 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
17328 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
17331 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17334 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
17335 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
17336 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
17337 for RS6000_BTM_POWERPC64.
17338 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
17339 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
17340 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
17343 (DIVDEU): Likewise.
17345 2018-03-28 Carl Love <cel@us.ibm.com>
17348 2017-09-27 Carl Love <cel@us.ibm.com>
17350 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
17351 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
17352 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
17355 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17357 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
17358 instead of __vector bool.
17359 (_mm_max_pu8): Likewise.
17360 (_mm_min_pi16): Likewise.
17362 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17365 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
17366 (DIVWEUO): Likewise.
17367 (DIVDEO): Likewise.
17368 (DIVDEUO): Likewise.
17369 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
17370 DIVWEUO and DIVDEUO.
17371 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
17372 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
17373 (div_extend): Likewise.
17374 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
17376 (__builtin_divweuo): Likewise.
17377 (__builtin_divdeo): Likewise.
17378 (__builtin_divdeuo): Likewise.
17380 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17383 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
17384 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
17386 PR tree-optimization/82004
17387 * gimple-match-head.c (optimize_pow_to_exp): New function.
17388 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
17389 Don't fold to exp if optimize_pow_to_exp is false.
17391 2018-03-28 Martin Liska <mliska@suse.cz>
17394 * calls.c (initialize_argument_information): Fix trailing space.
17395 * common.opt: Fix typo and provide better explanation for
17396 -fsanitize-coverage option.
17397 * config/i386/i386.opt: Fix typo.
17399 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17400 Martin Liska <mliska@suse.cz>
17403 * gimplify.c (asan_poison_variable): Don't do the check for
17404 gimplify_omp_ctxp here.
17405 (gimplify_decl_expr): Do it here.
17406 (gimplify_target_expr): Likewise.
17408 2018-03-28 Martin Liska <mliska@suse.cz>
17411 * config/i386/i386.c (ix86_function_arg_advance): Do not call
17412 chkp_type_bounds_count if MPX is not enabled.
17414 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
17416 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
17418 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
17421 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
17422 function to create the function decl for complex long double
17423 multiply and divide for -mabi=ieeelongdouble.
17424 (init_float128_ieee): Call it.
17426 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
17429 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
17430 -fcf-protection=branch -mibt.
17431 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
17433 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17436 * config/arm/arm.c (arm_valid_symbolic_address): Handle
17437 arm_word_relocations.
17439 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
17442 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
17443 extern array declarations.
17445 2018-03-27 Richard Biener <rguenther@suse.de>
17447 PR middle-end/84067
17448 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
17449 explicit single_use checks.
17451 2018-03-27 Richard Biener <rguenther@suse.de>
17453 PR tree-optimization/85082
17454 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
17457 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17459 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
17460 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
17461 Turn on fasynchronous-unwind-tables and funwind-tables.
17463 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
17466 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
17467 (*bmi_blsr_<mode>_ccz): Ditto.
17469 2018-03-26 Tom de Vries <tom@codesourcery.com>
17471 PR tree-optimization/85063
17472 * omp-general.c (offloading_function_p): New function. Factor out
17474 * omp-offload.c (pass_omp_target_link::gate): ... here.
17475 * omp-general.h (offloading_function_p): Declare.
17476 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
17477 with attribute omp declare target for offloading functions.
17479 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
17481 PR tree-optimization/84005
17482 * tree-data-ref.h (get_base_for_alignment): Declare.
17483 * tree-data-ref.c (get_base_for_alignment_1): New function.
17484 (get_base_for_alignment): Likewise.
17485 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
17486 get_base_for_alignment to find a suitable base object, instead
17487 of always using drb->base_address.
17489 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17491 PR inline-asm/85022
17492 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
17493 known size by default.
17495 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
17497 PR inline-asm/85030
17498 * lra-constraints.c (process_alt_operands): Don't match BLKmode
17499 and non BLKmode operands.
17501 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17504 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
17505 Clean up attributes.
17507 2018-03-23 Richard Biener <rguenther@suse.de>
17510 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
17511 we are going to emit early debug for LTO.
17513 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17515 PR inline-asm/85034
17516 * function.c (match_asm_constraints_1): Don't optimize if input
17517 doesn't satisfy general_operand predicate for output's mode.
17519 PR inline-asm/85022
17520 * alias.c (write_dependence_p): Don't require for x_canonicalized
17521 non-VOIDmode if x has VOIDmode.
17524 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
17525 just don't try to optimize it rather than assert it never happens.
17527 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17529 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
17530 macro expansions for definition of ST_INTERNAL_<mode> and
17531 LD_INTERNAL_<mode> builtins.
17532 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
17534 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
17536 (altivec_expand_st_builtin): Likewise.
17537 (altivec_expand_builtin): Remove calls to deleted functions.
17538 (rs6000_address_for_altivec): Delete this function.
17539 * config/rs6000/vector.md: Remove expands for
17540 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
17542 2018-03-22 Sudakshina Das <sudi.das@arm.com>
17545 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
17546 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
17547 re-computing once computed.
17548 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
17549 (arm_init_machine_status): Initialize
17550 machine->static_chain_stack_bytes.
17552 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17555 * doc/extend.texi: Add four new prototypes for vec_ld.
17556 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
17557 definitions for more logical presentation.
17558 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
17559 entries for V1TI variants of __builtin_altivec_ld builtin.
17560 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
17561 handling of V1TI variant of LVX icode pattern.
17562 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
17563 (rs6000_gimple_fold_builtin): Likewise.
17564 (altivec_init_builtins): Add code to define
17565 __builtin_altivec_lvx_v1ti function.
17567 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17569 PR inline-asm/84941
17570 * function.c (match_asm_constraints_1): Don't do the optimization
17571 if input isn't a REG, SUBREG, MEM or constant.
17573 2018-03-22 Tom de Vries <tom@codesourcery.com>
17575 PR tree-optimization/84956
17576 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
17577 bb_has_abnormal_pred.
17579 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17582 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
17583 DECL_INITIAL (decl) to decl at the end.
17584 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
17585 adjust the comment.
17587 2018-03-21 Joseph Myers <joseph@codesourcery.com>
17589 * doc/extend.texi (__builtin_tgmath): Document when complex
17590 integer types are treated as _Complex _Float64.
17592 2018-03-21 Tom de Vries <tom@codesourcery.com>
17594 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
17596 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17598 PR tree-optimization/84960
17599 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
17600 if it is ENTRY block, move them into single succ of ENTRY in that case.
17602 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
17604 PR tree-optimization/84811
17605 * poly-int.h (poly_span_traits): Remove the T3 parameter and
17606 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
17607 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
17608 (known_subrange_p): Update accordingly. Cast each value involved
17609 in the size comparison, rather than casting the result of the
17612 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17614 PR tree-optimization/84982
17615 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
17616 by flipping the least significant bit rather than all bits from
17617 bitpos to bitpos + bitsize - 1.
17619 2018-03-21 Nathan Sidwell <nathan@acm.org>
17621 * doc/extend.texi (Deprecated Features): Remove mention of
17622 long-deleted deprecations.
17624 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17627 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
17628 * configure: Regenerate.
17630 2018-03-21 Tom de Vries <tom@codesourcery.com>
17632 PR tree-optimization/83126
17633 * tree-parloops.c (num_phis): New function.
17634 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
17636 2018-03-21 Nathan Sidwell <nathan@acm.org>
17638 * doc/extend.texi (Deprecated Features): Update deprecated flags,
17639 mention anon-struct/union members and trailing attributes.
17641 2018-03-21 Bin Cheng <bin.cheng@arm.com>
17643 PR tree-optimization/84969
17644 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
17645 builtin memset partitions if they set different rhs values.
17647 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17649 PR rtl-optimization/84989
17650 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
17651 VEC_DUPLICATE with scalar result mode.
17653 2018-03-21 Martin Liska <mliska@suse.cz>
17656 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
17657 not intended return statement.
17659 2018-03-21 Martin Liska <mliska@suse.cz>
17662 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
17663 (chkp_find_bound_slots_1): Limit number of iterations.
17665 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
17668 * Minor grammar fixes for x86 options.
17670 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17673 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
17674 holding REG_CFA_RESTORE notes, instead turn them into a USE.
17676 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
17679 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
17680 (altivec_lvx_<mode>_1op): Likewise.
17681 (altivec_stvx_<mode>_2op): Likewise.
17682 (altivec_stvx_<mode>_1op): Likewise.
17683 (altivec_lvx_<VM2:mode>): New define_expand.
17684 (altivec_stvx_<VM2:mode>): Likewise.
17685 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
17686 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17687 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
17688 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17689 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
17690 (rs6000_gen_lvx): Likewise.
17691 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
17692 (altivec_expand_stv_builtin): Likewise.
17693 (altivec_expand_builtin): Likewise.
17694 * config/rs6000/vector.md: Likewise.
17696 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17699 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
17702 2018-03-20 Richard Biener <rguenther@suse.de>
17705 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
17706 sign-conversions as zero, fall back to standard scalar_stmt
17709 2018-03-20 Martin Liska <mliska@suse.cz>
17712 * predict.c (rebuild_frequencies): Handle case when we have
17713 PROFILE_ABSENT, but flag_guess_branch_prob is false.
17715 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17718 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
17719 flag_section_anchors.
17720 * varasm.c (use_blocks_for_decl_p): Remove hack for
17721 dw2_force_const_mem.
17724 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
17726 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
17727 be created, use lowpart_subreg of operands[0] rather than operands[0]
17729 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
17730 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
17731 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
17732 and n constraint instead of aarch64_shift_imm_di and Usd.
17733 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
17734 (*aarch64_<optab>_reg_minus<mode>3): ... this.
17736 2018-03-20 Sudakshina Das <sudi.das@arm.com>
17739 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
17740 to favor GPR over NEON registers.
17741 (<shift>di3_neon): Likewise.
17743 2018-03-20 Tom de Vries <tom@codesourcery.com>
17746 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
17747 (nvptx_process_pars): Emit bar.sync asap and alap.
17749 2018-03-20 Tom de Vries <tom@codesourcery.com>
17752 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
17753 seen_label if seen_label is already set.
17755 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17758 * config/i386/i386.c (fold_builtin_cpu): For features above 31
17759 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
17760 Use 1U instead of 1. Formatting fixes.
17763 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17764 instead of TREE_TYPE (s1) for the return value.
17766 2018-03-19 Jakub Jelinek <jakub@redhat.com>
17768 PR tree-optimization/84946
17769 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17770 bitsize + bitsize in poly_uint64 rather than poly_int64.
17773 * dwarf2asm.c: Include fold-const.c.
17774 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17775 of decl rather than decl itself.
17777 PR rtl-optimization/84643
17778 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17780 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
17783 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17784 calling assemble_variable.
17786 2018-03-19 Sudakshina Das <sudi.das@arm.com>
17789 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17790 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17792 2018-03-19 Jim Wilson <jimw@sifive.com>
17795 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
17796 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
17797 (riscv_first_stack_step): Likewise.
17798 (riscv_option_override): Use STACK_BOUNDARY instead of
17799 MIN_STACK_BOUNDARY.
17800 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
17801 MIN_STACK_BOUNDARY.
17802 (BIGGEST_ALIGNMENT): Set to 128.
17803 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
17804 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
17807 2018-03-19 Richard Biener <rguenther@suse.de>
17809 PR tree-optimization/84933
17810 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
17811 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
17813 2018-03-19 Richard Biener <rguenther@suse.de>
17815 PR tree-optimization/84859
17816 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
17817 (cond_if_else_store_replacement): Perform sinking operation on
17818 single-store BBs regardless of MAX_STORES_TO_SINK setting.
17819 Generalize what a BB with a single eligible store is.
17821 2018-03-19 Richard Biener <rguenther@suse.de>
17823 PR tree-optimization/84929
17824 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
17825 chrec_is_positive against non-chrec arg.
17827 2018-03-19 Tamar Christina <tamar.christina@arm.com>
17830 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
17832 2018-03-18 Martin Liska <mliska@suse.cz>
17834 PR rtl-optimization/84635
17835 * regrename.c (build_def_use): Use matches_mode only when
17838 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
17840 PR tree-optimization/84913
17841 * tree-vect-loop.c (vectorizable_reduction): Don't try to
17842 vectorize chains of COND_EXPRs.
17844 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17846 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
17848 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17850 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
17852 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
17854 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
17856 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17857 Kito Cheng <kito.cheng@gmail.com>
17859 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
17860 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
17861 (nds32_adjust_reg_alloc_order): New function.
17862 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
17864 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
17866 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
17867 nds32_print_operand, nds32_print_operand_address): Use
17868 HOST_WIDE_INT_PRINT_DEC instead.
17870 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
17872 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
17874 2018-03-17 Jakub Jelinek <jakub@redhat.com>
17877 * config/i386/i386.c (initial_ix86_tune_features,
17878 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
17879 unsigned long long.
17880 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
17881 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
17882 rather than 1u << ix86_tune. Formatting fix.
17883 (ix86_option_override_internal): Change ix86_arch_mask from
17884 unsigned int to unsigned HOST_WIDE_INT, initialize to
17885 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
17886 (ix86_function_specific_restore): Likewise.
17888 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17891 * postreload.c (reload_combine_recognize_pattern): Perform
17892 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
17893 truncate_int_for_mode the result for the destination's mode.
17896 * hsa-gen.c (mem_type_for_type): Fix comment typo.
17897 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
17899 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17902 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
17905 * lra-assigns.c (lra_split_hard_reg_for): Don't use
17906 regno_allocno_class_array and sorted_pseudos.
17907 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
17908 insns where regno is used.
17910 2018-03-16 Martin Liska <mliska@suse.cz>
17913 * multiple_target.c (create_dispatcher_calls): Redirect
17914 reference in the symbol table.
17916 2018-03-16 Martin Liska <mliska@suse.cz>
17919 * multiple_target.c (create_dispatcher_calls): Redirect also
17922 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17926 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
17927 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
17928 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
17930 2018-03-16 Julia Koval <julia.koval@intel.com>
17932 * doc/invoke.texi (Skylake Server): Add CLWB.
17933 Cannonlake): Remove CLWB.
17935 2018-03-16 Jakub Jelinek <jakub@redhat.com>
17937 PR tree-optimization/84841
17938 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
17940 (FLOAT_ONE_CONST_TYPE): Define.
17941 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
17942 (sort_by_operand_rank): Put entries with higher constant_type last
17943 rather than first to match comments.
17945 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
17947 * config/nios2/nios2.md (movsi_internal): Fix thinko in
17950 2018-03-15 Jakub Jelinek <jakub@redhat.com>
17953 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
17954 check and use address of target always.
17956 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
17959 * config/i386/i386.c (indirect_thunk_needed): Update comments.
17960 (indirect_thunk_bnd_needed): Likewise.
17961 (indirect_thunks_used): Likewise.
17962 (indirect_thunks_bnd_used): Likewise.
17963 (indirect_return_needed): New.
17964 (indirect_return_bnd_needed): Likewise.
17965 (output_indirect_thunk_function): Add a bool argument for
17967 (output_indirect_thunk_function): Don't generate alias for
17968 function return thunk.
17969 (ix86_code_end): Call output_indirect_thunk_function to generate
17970 function return thunks.
17971 (ix86_output_function_return): Set indirect_return_bnd_needed
17972 and indirect_return_needed instead of indirect_thunk_bnd_needed
17973 and indirect_thunk_needed.
17975 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
17977 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
17978 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
17979 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
17981 2018-03-15 David Malcolm <dmalcolm@redhat.com>
17982 Paul Hua <paul.hua.gm@gmail.com>
17985 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
17987 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
17989 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
17990 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
17991 resp. SFmode cases.
17993 2018-03-15 Tamar Christina <tamar.christina@arm.com>
17996 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
17997 instead of GET_MODE_SIZE when comparing Units.
17999 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18002 * varasm.c (hash_section): Return an unchangeble hash value
18003 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
18004 Return !aarch64_can_use_per_function_literal_pools_p ().
18006 2018-03-15 Jakub Jelinek <jakub@redhat.com>
18009 * optabs.c (emit_conditional_move): Pass address of cmode's copy
18010 rather than address of cmode as last argument to prepare_cmp_insn.
18012 2018-03-15 Julia Koval <julia.koval@intel.com>
18014 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
18015 F_AVX512VNNI, F_AVX512BITALG): New.
18017 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
18020 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
18021 insn for floating-point loads and stores.
18023 2018-03-14 Carl Love <cel@us.ibm.com>
18025 * config/rs6000/rs6000-c.c: Add macro definitions for
18026 ALTIVEC_BUILTIN_VEC_PERMXOR.
18027 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
18028 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
18029 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
18030 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
18032 * config/doc/extend.texi: Add prototypes for vec_permxor.
18034 2018-03-14 David Malcolm <dmalcolm@redhat.com>
18037 * diagnostic-show-locus.c (class layout_point): Convert m_line
18038 from int to linenum_type.
18039 (line_span::comparator): Use linenum "compare" function when
18040 comparing line numbers.
18041 (test_line_span): New function.
18042 (layout_range::contains_point): Convert param "row" from int to
18044 (layout_range::intersects_line_p): Likewise.
18045 (layout::will_show_line_p): Likewise.
18046 (layout::print_source_line): Likewise.
18047 (layout::should_print_annotation_line_p): Likewise.
18048 (layout::print_annotation_line): Likewise.
18049 (layout::print_leading_fixits): Likewise.
18050 (layout::annotation_line_showed_range_p): Likewise.
18051 (struct line_corrections): Likewise for field m_row.
18052 (line_corrections::line_corrections): Likewise for param "row".
18053 (layout::print_trailing_fixits): Likewise.
18054 (layout::get_state_at_point): Likewise.
18055 (layout::get_x_bound_for_row): Likewise.
18056 (layout::print_line): Likewise.
18057 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
18058 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
18059 * input.c (selftest::test_linenum_comparisons): New function.
18060 (selftest::input_c_tests): Call it.
18061 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
18062 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
18063 * selftest.h (ASSERT_GT): New macro.
18064 (ASSERT_GT_AT): New macro.
18065 (ASSERT_LT): New macro.
18066 (ASSERT_LT_AT): New macro.
18068 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
18070 PR rtl-optimization/84780
18071 * combine.c (distribute_links): Don't make a link based on pc_rtx.
18073 2018-03-14 Martin Liska <mliska@suse.cz>
18075 * tree.c (record_node_allocation_statistics): Use
18076 get_stats_node_kind.
18077 (get_stats_node_kind): New function extracted from
18078 record_node_allocation_statistics.
18079 (free_node): Use get_stats_node_kind.
18081 2018-03-14 Richard Biener <rguenther@suse.de>
18083 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
18084 that the value-set of ANTIC_IN doesn't grow.
18087 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18089 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18090 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18092 2018-03-14 Julia Koval <julia.koval@intel.com>
18094 * config.gcc (icelake-client, icelake-server): New.
18096 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
18097 (initial_ix86_arch_features): Ditto.
18098 (PTA_SKYLAKE): Add SGX.
18099 (PTA_ICELAKE): Remove.
18100 (PTA_ICELAKE_CLIENT): New.
18101 (PTA_ICELAKE_SERVER): New.
18102 (ix86_option_override_internal): Split up icelake on icelake client and
18104 (get_builtin_code_for_version): Ditto.
18105 (fold_builtin_cpu): Ditto.
18106 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
18107 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
18108 * config/i386/i386.h (processor_type): Ditto.
18109 * doc/invoke.texi: Ditto.
18111 2018-03-14 Jakub Jelinek <jakub@redhat.com>
18114 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
18115 INTEGER_CST offset, add it together with bitpos / 8 and
18116 sign extend based on POINTER_SIZE.
18120 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
18123 * config/i386/constraints.md (Yc): New register constraint.
18124 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
18125 Use Yc constraint for alternative 2 of operand 0. Remove
18126 preferred_for_speed attribute.
18128 2018-03-14 Richard Biener <rguenther@suse.de>
18130 PR tree-optimization/84830
18131 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
18132 with the old one to avoid oscillations.
18134 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
18137 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
18139 (assign_by_spills): Return a flag of reload assignment failure.
18140 Do not process the reload assignment failures. Do not spill other
18141 reload pseudos if they has the same reg class. Update n if
18143 (lra_assign): Add a return arg. Set up from the result of
18144 assign_by_spills call.
18145 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18146 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18147 usage_insns if it is not NULL.
18148 (spill_hard_reg_in_range): New function.
18149 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18150 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18151 function prototypes.
18152 (lra_assign): Change prototype.
18153 * lra.c (lra): Add code to deal with fails by splitting hard reg
18156 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
18158 * config/riscv/riscv.opt (mrelax): New option.
18159 * config/riscv/riscv.c (riscv_file_start): Emit ".option
18160 "norelax" when riscv_mrelax is disabled.
18161 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
18163 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18166 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
18167 reassociation for int modes.
18169 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18171 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
18172 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
18174 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
18175 * config/aarch64/aarch64-sve.md
18176 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
18177 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
18178 (*extend<mode><Vwide>2): Rename to...
18179 (aarch64_sve_extend<mode><Vwide>2): ...this.
18180 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
18181 renaming the old pattern to...
18182 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
18184 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
18185 define_expand, renaming the old pattern to...
18186 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
18187 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
18188 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
18189 account when deciding which SVE instruction the optab should use.
18190 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
18192 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18194 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
18195 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
18196 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
18197 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
18198 (tlsdesc_small_<mode>): Turn a define_expand and use
18199 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
18200 (tlsdesc_small_advsimd_<mode>): ...this.
18201 (tlsdesc_small_sve_<mode>): New pattern.
18203 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18205 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
18206 (UNSPEC_UMUL_HIGHPART): New constants.
18207 (MUL_HIGHPART): New int iteraor.
18208 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
18209 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
18211 (*<su>mul<mode>3_highpart): New define_insn.
18213 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
18216 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18219 2018-03-13 Martin Liska <mliska@suse.cz>
18222 * (sem_item_optimizer::sem_item_optimizer): Initialize new
18224 (sem_item_optimizer::~sem_item_optimizer): Release it.
18225 (sem_item_optimizer::merge_classes): Register variable aliases.
18226 (sem_item_optimizer::fixup_pt_set): New function.
18227 (sem_item_optimizer::fixup_points_to_sets): Likewise.
18228 * ipa-icf.h: Declare new variables and functions.
18230 2018-03-13 Jakub Jelinek <jakub@redhat.com>
18232 PR middle-end/84834
18233 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
18234 integer_pow2p@2 and test integer_pow2p in condition.
18235 (A < 0 ? C : 0): Similarly for @1.
18237 PR middle-end/84831
18238 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
18239 characters starting at p contain '\0' character, don't look beyond
18243 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
18244 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
18247 * reg-stack.c (change_stack): Change update_end var from int to
18248 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
18249 also call set_block_for_insn on the newly added insns and rescan.
18252 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
18253 on the last operand.
18256 * tree.c (stabilize_reference_1): Return save_expr (e) for
18257 STATEMENT_LIST even if it doesn't have side-effects.
18259 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
18261 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
18263 2018-03-12 Renlin Li <renlin.li@arm.com>
18265 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
18266 aarch64_output_scalar_simd_mov_immediate.
18268 2018-03-12 Martin Sebor <msebor@redhat.com>
18270 PR tree-optimization/83456
18271 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
18272 for perfectly overlapping calls to memcpy.
18273 (gimple_fold_builtin_memory_chk): Same.
18274 (gimple_fold_builtin_strcpy): Handle no-warning.
18275 (gimple_fold_builtin_stxcpy_chk): Same.
18276 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
18278 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
18280 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
18281 parameter. Use it for SFmode.
18282 (rs6000_function_arg_advance_1): Adjust.
18283 (rs6000_function_arg): Adjust.
18284 (rs6000_gimplify_va_arg): Pass false for that new parameter.
18286 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
18288 PR rtl-optimization/84169
18289 PR rtl-optimization/84780
18290 * combine.c (can_combine_p): Check for a 2-insn combination whether
18291 the destination register is used between the two insns, too.
18293 2018-03-12 Richard Biener <rguenther@suse.de>
18295 PR tree-optimization/84803
18296 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
18297 for refs DR analysis didn't process.
18299 2018-03-12 Richard Biener <rguenther@suse.de>
18301 PR tree-optimization/84777
18302 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
18303 force-vectorize loops ignore whether we are optimizing for size.
18305 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
18307 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
18308 (TARGET_MD_ASM_ADJUST): Define.
18310 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
18311 Kito Cheng <kito.cheng@gmail.com>
18312 Chung-Ju Wu <jasonwucj@gmail.com>
18314 * config/nds32/nds32.c (nds32_compute_stack_frame,
18315 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
18316 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
18317 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
18318 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
18319 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
18320 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
18321 * config/nds32/nds32.md (prologue, epilogue): Use macro
18322 NDS32_V3PUSH_AVAILABLE_P to do checking.
18324 2018-03-11 Jakub Jelinek <jakub@redhat.com>
18327 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
18328 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
18329 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
18330 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
18331 addition of most attributes on !orig_type_die or the attribute not
18332 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
18334 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18335 Chung-Ju Wu <jasonwucj@gmail.com>
18337 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
18338 __NDS32_VH__ macro.
18339 * config/nds32/nds32.opt (mvh): New option.
18341 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18342 Chung-Ju Wu <jasonwucj@gmail.com>
18344 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
18346 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
18347 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
18350 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18351 Chung-Ju Wu <jasonwucj@gmail.com>
18353 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
18355 * config/nds32/nds32-multiple.md (strlensi): New pattern.
18356 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
18358 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
18359 Kito Cheng <kito.cheng@gmail.com>
18360 Chung-Ju Wu <jasonwucj@gmail.com>
18362 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
18363 UNSPEC_FFMISM and UNSPEC_FLMISM.
18364 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
18365 for ffb, ffmism and flmism.
18366 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
18367 (unspec_ffmism): Ditto.
18368 (unspec_flmism): Ditto.
18369 (nds32_expand_builtin_impl): Check if string extension is available.
18370 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
18371 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
18373 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
18376 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18379 * lra-assigns.c (assign_by_spills): Return a flag of reload
18380 assignment failure. Do not process the reload assignment
18381 failures. Do not spill other reload pseudos if they has the same
18383 (lra_assign): Add a return arg. Set up from the result of
18384 assign_by_spills call.
18385 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18386 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18387 usage_insns if it is not NULL.
18388 (spill_hard_reg_in_range): New function.
18389 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18390 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18391 function prototypes.
18392 (lra_assign): Change prototype.
18393 * lra.c (lra): Add code to deal with fails by splitting hard reg
18396 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
18399 * config/i386/i386.opt: Replace Enforcment with Enforcement.
18401 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
18404 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
18405 (dw_val_node): Add val_symbolic_view.
18406 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
18407 (symview_upper_bound): New.
18408 (new_line_info_table): Initialize symviews_since_reset.
18409 (dwarf2out_source_line): Count symviews_since_reset and set
18410 symview_upper_bound.
18411 (dw_val_equal_p): Handle symview.
18412 (add_AT_symview): New.
18413 (print_dw_val): Handle symview.
18414 (attr_checksum, attr_checksum_ordered): Likewise.
18415 (same_dw_val_p, size_of_die): Likewise.
18416 (value_format, output_die): Likewise.
18417 (add_high_low_attributes): Use add_AT_symview for entry_view.
18418 (dwarf2out_finish): Reset symview_upper_bound, clear
18421 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
18424 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
18425 Add strict argument and use it.
18426 (rs6000_split_multireg_move): Update for new strict argument.
18427 (mem_operand_gpr): Disallow all non-offsettable addresses.
18428 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
18430 2018-03-09 Jakub Jelinek <jakub@redhat.com>
18433 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
18434 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
18435 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
18438 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
18439 decl, use remap_type if we want to use the type.
18441 2018-03-09 Martin Sebor <msebor@redhat.com>
18443 PR tree-optimization/84526
18444 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18446 (builtin_access::generic_overlap): Be prepared to handle non-array
18449 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
18451 PR rtl-optimization/84682
18452 * lra-constraints.c (process_address_1): Check is_address flag
18453 for address constraints.
18454 (process_alt_operands): Likewise.
18455 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
18456 preprocess_constraints.
18457 * recog.h (preprocess_constraints): Add oploc parameter.
18459 * recog.c (preprocess_constraints): Test address_operand for
18460 CT_ADDRESS constraints.
18462 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18465 * lra-assigns.c (assign_by_spills): Return a flag of reload
18466 assignment failure. Do not process the reload assignment
18467 failures. Do not spill other reload pseudos if they has the same
18469 (lra_assign): Add a return arg. Set up from the result of
18470 assign_by_spills call.
18471 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18472 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18473 usage_insns if it is not NULL.
18474 (spill_hard_reg_in_range): New function.
18475 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18476 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18477 function prototypes.
18478 (lra_assign): Change prototype.
18479 * lra.c (lra): Add code to deal with fails by splitting hard reg
18482 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18485 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
18486 Accept complain bool parameter. Only emit errors if it is true.
18487 (arm_parse_cpu_option_name): Likewise.
18488 (arm_target_thumb_only): Adjust callers of the above.
18489 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
18490 prototype to take a default true bool parameter.
18491 (arm_parse_arch_option_name): Likewise.
18493 2018-03-09 David Malcolm <dmalcolm@redhat.com>
18494 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
18498 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
18499 * configure: Regenerate.
18500 * configure.ac ("linker --version-script option"): New.
18501 ("linker soname option"): New.
18503 2018-03-09 Richard Biener <rguenther@suse.de>
18505 PR tree-optimization/84775
18506 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
18507 immediate uses of predicate stmts and mark them modified.
18510 PR tree-optimization/84178
18511 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18513 (version_loop_for_if_conversion): Delay update_ssa call.
18514 (tree_if_conversion): Delay update_ssa until after predicate
18517 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
18520 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
18521 when the function accesses prior frames.
18523 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18526 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
18527 gen_llsym, otherwise call maybe_gen_llsym.
18529 PR inline-asm/84742
18530 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
18531 has ',' character inside of it.
18533 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18536 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
18537 as clobbering CC_REGNUM.
18539 2018-03-08 Richard Biener <rguenther@suse.de>
18541 PR middle-end/84552
18542 * tree-scalar-evolution.c: Include tree-into-ssa.h.
18543 (follow_copies_to_constant): Do not follow SSA names registered
18546 2018-03-08 Richard Biener <rguenther@suse.de>
18548 PR tree-optimization/84178
18549 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18551 (version_loop_for_if_conversion): Delay update_ssa call.
18552 (tree_if_conversion): Delay update_ssa until after predicate
18555 2018-03-08 David Malcolm <dmalcolm@redhat.com>
18557 PR tree-optimization/84178
18558 * tree-if-conv.c (release_bb_predicate): Remove the
18559 the assertion that the stmts have NULL use_ops.
18560 Discard the statements, asserting that they haven't
18561 yet been added to a BB.
18563 2018-03-08 Richard Biener <rguenther@suse.de>
18565 PR tree-optimization/84746
18566 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
18567 (phi_translate): Pass in destination ANTIC_OUT set.
18568 (phi_translate_1): Likewise. For a simplified result lookup
18569 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
18570 (phi_translate_set): Adjust.
18571 (do_pre_regular_insertion): Likewise.
18572 (do_pre_partial_partial_insertion): Likewise.
18574 2018-03-08 Martin Liska <mliska@suse.cz>
18576 PR gcov-profile/84735
18577 * doc/gcov.texi: Document usage of profile files.
18578 * gcov-io.h: Document changes in the format.
18580 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
18584 * dwarf2out.c (struct dw_line_info_table): Update comments for
18586 (FORCE_RESET_NEXT_VIEW): New.
18587 (FORCE_RESETTING_VIEW_P): New.
18588 (RESETTING_VIEW_P): Check for -1 too.
18589 (ZERO_VIEW_P): Likewise.
18590 (new_line_info_table): Force-reset next view.
18591 (dwarf2out_begin_function): Likewise.
18592 (dwarf2out_source_line): Simplify zero_view_p initialization.
18593 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
18594 view directly. Omit view when omitting .loc at line 0.
18596 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18598 PR tree-optimization/84740
18599 * tree-switch-conversion.c (process_switch): Call build_constructors
18600 only if info.phi_count is non-zero.
18602 PR tree-optimization/84739
18603 * tree-tailcall.c (find_tail_calls): Check call arguments against
18604 DECL_ARGUMENTS (current_function_decl) rather than
18605 DECL_ARGUMENTS (func) when checking for tail recursion.
18607 2018-03-07 Jakub Jelinek <jakub@redhat.com>
18609 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
18610 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
18611 Volker Reichelt's entry and add entries for people that perform
18612 GCC fuzzy testing and report numerous bugs.
18614 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
18617 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
18618 readonly data in sdata, if that is disabled.
18619 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
18620 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
18621 -mreadonly-in-sdata option.
18623 2018-03-07 Martin Sebor <msebor@redhat.com>
18625 PR tree-optimization/84468
18626 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
18627 basic block when looking for nul assignment.
18629 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
18632 * except.h (output_function_exception_table): Adjust prototype.
18633 * except.c (output_function_exception_table): Remove FNNAME parameter
18634 and add SECTION parameter. Ouput one part of the table at a time.
18635 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
18636 the first part of the exception table and emit unwind directives.
18637 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
18638 (i386_pe_seh_cold_init): Likewise.
18639 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
18640 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
18641 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
18642 (ix86_output_call_insn): Emit a nop in one more case for SEH.
18643 * config/i386/winnt.c: Include except.h.
18644 (struct seh_frame_state): Add reg_offset, after_prologue and
18645 in_cold_section fields.
18646 (i386_pe_seh_end_prologue): Set seh->after_prologue.
18647 (i386_pe_seh_cold_init): New function.
18648 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
18649 to seh->in_cold_section.
18650 (seh_emit_push): Record the offset of the push.
18651 (seh_emit_save): Record the offet of the save.
18652 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
18653 Test seh->after_prologue to disregard the epilogue.
18654 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
18655 (i386_pe_end_cold_function): New function.
18657 2018-03-07 Jakub Jelinek <jakub@redhat.com>
18660 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
18661 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
18664 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
18666 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
18667 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
18669 PR middle-end/84723
18670 * multiple_target.c: Include tree-inline.h and intl.h.
18671 (expand_target_clones): Diagnose and fail if node->definition and
18672 !tree_versionable_function_p (node->decl).
18674 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
18676 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
18678 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
18679 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
18680 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
18682 2018-03-06 Jakub Jelinek <jakub@redhat.com>
18685 * combine.c (try_combine): Use reg_or_subregno instead of handling
18686 just paradoxical SUBREGs and REGs.
18688 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
18690 * config/arc/arc.c (arc_finalize_pic): Remove function.
18691 (arc_must_save_register): We use single base PIC register, remove
18692 checks to save/restore the PIC register.
18693 (arc_expand_prologue): Likewise.
18694 * config/arc/arc-protos.h (arc_set_default_type_attributes):
18696 (arc_verify_short): Likewise.
18697 (arc_attr_type): Likewise.
18698 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
18699 (walk_stores): Likewise.
18700 (arc_address_cost): Make it static.
18701 (arc_verify_short): Likewise.
18702 (branch_dest): Likewise.
18703 (arc_attr_type): Likewise.
18704 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
18705 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
18706 (arc_final_prescan_insn): Remove inserting the nops due to
18707 hardware hazards. It is done in reorg step.
18708 (insn_length_variant_t): Remove.
18709 (insn_length_parameters_t): Likewise.
18710 (arc_insn_length_parameters): Likewise.
18711 (arc_get_insn_variants): Likewise.
18712 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
18714 2018-03-06 Jakub Jelinek <jakub@redhat.com>
18716 PR inline-asm/84683
18717 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
18720 PR tree-optimization/84687
18721 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
18723 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
18725 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18727 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
18728 Rename to ppc_speculation_barrier.
18729 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
18730 __builtin_ppc_speculation_barrier.
18732 2018-03-05 Jakub Jelinek <jakub@redhat.com>
18735 * combine.c (combine_simplify_rtx): Don't try to simplify if
18736 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
18739 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
18741 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
18742 to 32 bytes when compiling for POWER9.
18744 2018-03-05 Jakub Jelinek <jakub@redhat.com>
18747 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
18748 regparm >= 3 with no arg reg available also for calls with
18749 flag_force_indirect_call. Pass decl to ix86_function_regparm.
18752 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
18754 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
18756 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
18759 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
18761 2018-03-05 Richard Biener <rguenther@suse.de>
18763 PR tree-optimization/84486
18764 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18765 When inserting a __builtin_assume_aligned call set the LHS
18766 SSA name alignment info accordingly.
18768 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
18770 PR tree-optimization/84114
18771 * config/aarch64/aarch64.c (aarch64_reassociation_width)
18772 Avoid reassociation of FLOAT_MODE addition.
18774 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
18776 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18777 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18778 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18779 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18780 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18781 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18782 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18784 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18785 __builtin_ia32_wbinvd): New builtins.
18786 (SPECIAL_ARGS2): New.
18787 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18788 (SPECIAL_ARGS2): New.
18789 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18790 (ix86_valid_target_attribute_inner_p): Ditto.
18791 (ix86_init_mmx_sse_builtins): Add special_args2.
18792 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18793 TARGET_WBNOINVD_P): New.
18794 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18795 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
18796 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
18797 * config/i386/immintrin.h (_wbinvd): New intrinsic.
18798 * config/i386/pconfigintrin.h: New file.
18799 * config/i386/wbnoinvdintrin.h: Ditto.
18800 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
18802 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
18804 2018-03-05 Richard Biener <rguenther@suse.de>
18806 PR tree-optimization/84670
18807 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18809 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18810 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18811 (compute_antic_aux): Only assert the number of values in ANTIC_IN
18812 doesn't grow if all successors (recursively) were visited at least
18815 2018-03-05 Richard Biener <rguenther@suse.de>
18817 PR tree-optimization/84650
18818 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
18819 if executed in the loop pipeline.
18821 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
18823 * doc/configfiles.texi (Configuration Files): Move info about
18824 conditionalizing $target-protos.h to...
18825 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
18826 differs from $target-protos.h.
18828 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
18829 Chung-Ju Wu <jasonwucj@gmail.com>
18831 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
18832 * config/nds32/nds32-multiple.md (setmemsi): Define.
18833 * config/nds32/nds32-memory-manipulation.c
18834 (nds32_gen_dup_4_byte_to_word_value): New.
18835 (emit_setmem_word_loop): New.
18836 (emit_setmem_byte_loop): New.
18837 (nds32_expand_setmem_loop): New.
18838 (nds32_expand_setmem_loop_v3m): New.
18839 (nds32_expand_setmem_unroll): New.
18840 (nds32_expand_setmem): New.
18842 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18843 Chung-Ju Wu <jasonwucj@gmail.com>
18845 * config/nds32/nds32-memory-manipulation.c
18846 (nds32_emit_load_store): New.
18847 (nds32_emit_post_inc_load_store): New.
18848 (nds32_emit_mem_move): New.
18849 (nds32_emit_mem_move_block): New.
18850 (nds32_expand_movmemsi_loop_unknown_size): New.
18851 (nds32_expand_movmemsi_loop_known_size): New.
18852 (nds32_expand_movmemsi_loop): New.
18853 (nds32_expand_movmemsi_unroll): New.
18854 (nds32_expand_movmemqi): Rename ...
18855 (nds32_expand_movmemsi): ... to this.
18856 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
18857 (movmemsi): ... to this.
18858 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
18859 (nds32_expand_movmemsi): ... to this.
18861 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18862 Monk Chiang <sh.chiang04@gmail.com>
18863 Chung-Ju Wu <jasonwucj@gmail.com>
18865 * config/nds32/nds32-protos.h
18866 (nds32_expand_load_multiple): New arguments.
18867 (nds32_expand_store_multiple): Ditto.
18868 (nds32_valid_multiple_load_store): Rename ...
18869 (nds32_valid_multiple_load_store_p): ... to this.
18870 * config/nds32/nds32-memory-manipulation.c
18871 (nds32_expand_load_multiple): Refine implementation.
18872 (nds32_expand_store_multiple): Ditto.
18873 * config/nds32/nds32-multiple.md
18874 (load_multiple): Update nds32_expand_load_multiple interface.
18875 (store_multiple): Update nds32_expand_store_multiple interface.
18876 * config/nds32/nds32-predicates.c
18877 (nds32_valid_multiple_load_store): Rename ...
18878 (nds32_valid_multiple_load_store_p): ... to this and refine
18880 * config/nds32/predicates.md
18881 (nds32_load_multiple_and_update_address_operation): New predicate.
18882 (nds32_store_multiple_and_update_address_operation): New predicate.
18884 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
18885 Chung-Ju Wu <jasonwucj@gmail.com>
18887 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
18888 (combo): New attribute.
18889 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
18891 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
18893 * config/nds32/nds32.opt: Change -mcmodel= default value.
18895 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
18896 Monk Chiang <sh.chiang04@gmail.com>
18897 Chung-Ju Wu <jasonwucj@gmail.com>
18899 * config/nds32/constants.md (unspec_element): New enum.
18900 * config/nds32/constraints.md (Umw): New constraint.
18901 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
18902 * config/nds32/nds32-intrinsic.md: Likewise.
18903 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
18904 (nds32_valid_smw_lwm_base_p): New.
18905 (nds32_output_smw_single_word): New.
18906 (nds32_output_lmw_single_word): New.
18907 (nds32_expand_unaligned_load): New.
18908 (nds32_expand_unaligned_store): New.
18909 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
18910 (nds32_output_smw_single_word): Declare.
18911 (nds32_output_lmw_single_word): Declare.
18912 (nds32_expand_unaligned_load): Declare.
18913 (nds32_expand_unaligned_store): Declare.
18914 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
18915 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
18916 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
18917 NDS32_BUILTIN_UASTORE_DW.
18918 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
18921 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
18922 Kito Cheng <kito.cheng@gmail.com>
18923 Chung-Ju Wu <jasonwucj@gmail.com>
18925 * config/nds32/nds32-intrinsic.c
18926 (nds32_expand_builtin_null_ftype_reg): Delete.
18927 (nds32_expand_builtin_reg_ftype_imm): Ditto.
18928 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
18929 (nds32_read_argument): New.
18930 (nds32_legitimize_target): Ditto.
18931 (nds32_legitimize_argument): Ditto.
18932 (nds32_check_constant_argument): Ditto.
18933 (nds32_expand_unop_builtin): Ditto.
18934 (nds32_expand_unopimm_builtin): Ditto.
18935 (nds32_expand_binop_builtin): Ditto.
18936 (nds32_builtin_decl_impl): Ditto.
18937 (builtin_description): Ditto.
18938 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
18939 (nds32_init_builtins_impl): Ditto.
18940 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
18941 (nds32_builtin_decl): New.
18942 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
18943 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
18945 2018-03-02 Jeff Law <law@redhat.com>
18947 * reorg.c (stop_search_p): Handle DEBUG_INSN.
18948 (redundant_insn, fill_simple_delay_slots): Likewise.
18949 (fill_slots_from_thread): Likewise.
18950 * resource.c (mark_referenced_resources): Likewise.
18951 (mark_set_resources, find_dead_or_set_registers): Likewise.
18953 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18955 * substring-locations.h (format_warning_va): Formatting fix for
18956 ATTRIBUTE_GCC_DIAG.
18957 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
18959 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
18960 * substring-locations.c: Include intl.h.
18961 (format_warning_va): Turned into small wrapper around
18962 format_warning_n_va, renamed to ...
18963 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
18964 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
18966 (format_warning_at_substring_n): New function.
18967 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
18968 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
18969 format_warning_at_substring with just a shorter name instead of
18970 const function pointer.
18971 (fmtwarn_n): New function.
18972 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
18973 appropriate, get rid of all the fmtstr temporaries, move conditionals
18974 with G_() wrapped string literals directly into fmtwarn arguments,
18975 cast dir.len to (int), formatting fixes.
18977 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
18979 * doc/invoke.texi: Remove "Cilk Plus" references.
18981 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18982 Richard Biener <rguenther@suse.de>
18985 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
18986 for error or warning attributes if CALL_FROM_THUNK_P is set.
18989 2018-03-02 Jakub Jelinek <jakub@redhat.com>
18992 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
18993 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
18996 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
18997 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
18999 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
19000 instead of -1U in last predictors element's probability member.
19002 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
19005 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19006 arguments if they are comparable.
19008 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
19010 PR tree-optimization/84634
19011 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
19012 masks and masked_loop_p with a single loop_masks, making sure it's
19013 null for bb vectorization.
19015 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
19017 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
19018 (vect_analyze_data_ref_access): Use loop->safe_len rather than
19019 loop->force_vectorize to check whether there is no alias.
19021 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19024 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
19026 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
19028 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
19029 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
19030 instead of a loop around prev_real_insn.
19031 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
19034 PR inline-asm/84625
19035 * config/i386/i386.c (ix86_print_operand): Use conditional
19036 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
19039 2018-03-02 Richard Biener <rguenther@suse.de>
19041 PR tree-optimization/84427
19042 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
19043 (bitmap_set_subtract_values): Rewrite to handle multiple
19046 (prune_clobbered_mems): Likewise.
19047 (phi_translate): Take edge instead of pred/phiblock.
19048 (phi_translate_1): Likewise.
19049 (phi_translate_set): Likewise. Insert all translated
19050 exprs for a value into the set, keeping possibly multiple
19051 expressions per value.
19052 (compute_antic_aux): Adjust for phi_translate changes.
19053 When intersecting union the expressions and prune those
19054 not in the final value set, keeping possibly multiple
19055 expressions per value. Do not use value-insertion
19056 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
19057 all expressions. Add verification that the value-sets
19058 only shrink during iteration.
19059 (compute_partial_antic_aux): Adjust for the phi_translate changes.
19060 (do_pre_regular_insertion): Likewise.
19061 (do_pre_partial_partial_insertion): Likewise.
19063 2018-03-02 Richard Biener <rguenther@suse.de>
19066 * config/darwin.c (saved_debug_info_level): New static global.
19067 (darwin_asm_lto_start): Disable debug info generation for LTO out.
19068 (darwin_asm_lto_end): Restore debug info generation settings.
19070 2018-03-01 Martin Liska <mliska@suse.cz>
19073 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
19074 volatile arguments.
19076 2018-03-01 Richard Biener <rguenther@suse.de>
19079 * dwarf2out.c (gen_variable_die): Properly handle late VLA
19080 type annotation with LTO when debug was disabled at compile-time.
19082 2018-03-01 Matthew Fortune <mfortune@gmail.com>
19084 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
19086 (mips_final_postscan_insn): Likewise.
19088 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
19090 PR rtl-optimization/84528
19091 * alias.c (init_alias_target): Add commentary.
19092 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
19093 a unique base value if the frame pointer is not eliminated
19094 to the stack pointer.
19096 2018-03-01 Tom de Vries <tom@codesourcery.com>
19098 PR rtl-optimization/83327
19099 * lra-int.h (hard_regs_spilled_into): Declare.
19100 * lra.c (hard_regs_spilled_into): Define.
19101 (init_reg_info): Init hard_regs_spilled_into.
19102 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
19103 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
19104 (process_bb_lives): Handle hard_regs_spilled_into.
19105 (lra_create_live_ranges_1): Before doing liveness propagation, clear
19106 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
19108 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
19110 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
19111 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
19112 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
19113 * config/rs6000/aix72.h: New file.
19115 2018-02-28 Jakub Jelinek <jakub@redhat.com>
19117 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
19118 instead of warning_at with conditional singular and plural messages
19122 * stor-layout.c (update_alignment_for_field): For
19123 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
19124 && !DECL_PACKED (field), do the alignment update, just use
19125 only desired_align instead of MAX (type_align, desired_align)
19127 (place_field): Don't do known_align < desired_align handling
19128 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
19129 is non-NULL, instead do it after rli->prev_field handling and
19130 only if not within a bitfield word. For DECL_PACKED (field)
19131 use type_align of BITS_PER_UNIT.
19133 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
19135 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
19136 superfluous parentheses and trailing spaces.
19138 2018-02-28 Richard Biener <rguenther@suse.de>
19140 PR tree-optimization/84584
19141 * graphite-scop-detection.c (scop_detection::add_scop): Discard
19142 SCoPs with fake exit edge.
19144 2018-02-28 Martin Liska <mliska@suse.cz>
19147 * timevar.c (timer::print): Fix format to properly print 100%
19150 2018-02-28 Richard Biener <rguenther@suse.de>
19152 PR middle-end/84607
19153 * genmatch.c (capture_info::walk_match): Do not mark
19154 captured expressions without operands as expr_p given
19155 they act more like predicates and should be subject to
19156 "lost tail" side-effect preserving.
19158 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
19160 PR rtl-optimization/81611
19161 * auto-inc-dec.c (attempt_change): Move dead note from
19162 mem_insn if it's the next use of regno
19163 (find_address): Take address use of reg holding
19164 non-incremented value. Add parm to limit search to the named
19166 (merge_in_block): Attempt to use a mem insn that is the next
19167 use of the original regno.
19169 2018-02-27 Martin Sebor <msebor@redhat.com>
19172 * doc/invoke.texi (-Wmissing-attributes): New option.
19173 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
19175 2018-02-27 Martin Sebor <msebor@redhat.com>
19177 PR translation/84207
19178 * diagnostic-core.h (warning_n, error_n, inform_n): Change
19179 n argument to unsigned HOST_WIDE_INT.
19180 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
19181 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
19182 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
19183 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
19185 2018-02-27 Richard Biener <rguenther@suse.de>
19187 PR tree-optimization/84512
19188 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19189 Do not use the estimate returned from record_stmt_cost for
19190 the scalar iteration cost but sum properly using add_stmt_cost.
19192 2018-02-27 Richard Biener <rguenther@suse.de>
19194 PR tree-optimization/84466
19195 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19196 Adjust last change to less strictly validate use operands.
19198 2018-02-27 Martin Liska <mliska@suse.cz>
19200 PR gcov-profile/84548
19201 * gcov.c (process_file): Allow partial overlap and consider it
19202 also as group functions.
19203 (output_lines): Properly calculate range of lines for a group.
19205 2018-02-27 Martin Liska <mliska@suse.cz>
19207 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
19208 'ggc' suffixes. Change first column width.
19209 (timer::print): Fix formatting of the column.
19211 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
19213 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
19214 preserve inline entry blocks for the sake of debug inline
19215 entry point markers alone.
19216 (remove_unused_locals): Suggest in comments a better place to
19217 force the preservation of inline entry blocks that are
19218 otherwise unused, but do not preserve them.
19220 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19222 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
19224 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19227 * config/i386/constraints.md (Bs): Replace
19228 ix86_indirect_branch_register with
19229 TARGET_INDIRECT_BRANCH_REGISTER.
19231 * config/i386/i386.md (indirect_jump): Likewise.
19232 (tablejump): Likewise.
19233 (*sibcall_memory): Likewise.
19234 (*sibcall_value_memory): Likewise.
19235 Peepholes of indirect call and jump via memory: Likewise.
19236 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
19237 (*sibcall_value_GOT_32): Likewise.
19238 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19239 (GOT_memory_operand): Likewise.
19240 (call_insn_operand): Likewise.
19241 (sibcall_insn_operand): Likewise.
19242 (GOT32_symbol_operand): Likewise.
19243 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
19245 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
19247 PR rtl-optimization/83496
19248 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
19249 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
19250 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
19251 redundant insn, if any.
19252 (relax_delay_slots): Likewise.
19253 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
19255 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
19257 PR tree-optimization/83965
19258 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
19259 that grouped statements are part of a reduction chain. Return
19260 true if the statement is not marked as a reduction itself but
19261 is part of a group.
19262 (vect_recog_dot_prod_pattern): Don't check whether the statement
19263 is part of a group here.
19264 (vect_recog_sad_pattern): Likewise.
19265 (vect_recog_widen_sum_pattern): Likewise.
19267 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
19270 * final.c (rest_of_clean_state): Also look for calls inside sequences.
19272 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19275 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
19277 (ix86_output_indirect_function_return): New prototype.
19278 (ix86_split_simple_return_pop_internal): Likewise.
19279 * config/i386/i386.c (indirect_return_via_cx): New.
19280 (indirect_return_via_cx_bnd): Likewise.
19281 (indirect_thunk_name): Handle return va CX_REG.
19282 (output_indirect_thunk_function): Create alias for
19283 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
19284 (ix86_output_indirect_jmp): Remove the bool argument.
19285 (ix86_output_indirect_function_return): New function.
19286 (ix86_split_simple_return_pop_internal): Likewise.
19287 * config/i386/i386.md (*indirect_jump): Don't pass false
19288 to ix86_output_indirect_jmp.
19289 (*tablejump_1): Likewise.
19290 (simple_return_pop_internal): Change it to define_insn_and_split.
19291 Call ix86_split_simple_return_pop_internal to split it for
19292 -mfunction-return=.
19293 (simple_return_indirect_internal): Call
19294 ix86_output_indirect_function_return instead of
19295 ix86_output_indirect_jmp.
19297 2018-02-26 Jakub Jelinek <jakub@redhat.com>
19300 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
19301 memset and value initialization afterwards.
19303 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
19305 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
19307 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19310 * common/config/aarch64/aarch64-common.c
19311 (aarch_option_optimization_table[]): Switch
19312 off fomit-frame-pointer
19314 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19315 Chung-Ju Wu <jasonwucj@gmail.com>
19317 * config/nds32/nds32-multiple.md (load_multiple): Disallow
19319 (store_multiple): Ditto.
19321 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19323 * config.gcc: Add --with-cpu support for nds32 target.
19324 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
19325 * config/nds32/nds32.opt: Add -mcpu= option.
19327 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
19329 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
19330 isel=yes): Warn for these deprecated options.
19332 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
19334 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
19335 ISA_2_5_MASKS_EMBEDDED.
19337 2018-02-23 Jakub Jelinek <jakub@redhat.com>
19339 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
19340 p->max as pointers rather than using iterative_hash_expr.
19342 2018-02-23 Carl Love <cel@us.ibm.com>
19344 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
19345 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
19347 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
19348 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
19349 P8V_BUILTIN_VEC_VUNSIGNED2.
19351 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
19354 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
19355 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
19357 * lra-constraints.c (curr_insn_transform): Set up
19358 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
19360 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
19361 * lra-eliminations.c (spill_pseudos): Ditto.
19362 (process_insn_for_elimination): Ditto.
19363 * lra-lives.c (reg_early_clobber_p): Use the new macros.
19364 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
19365 LRA_NON_CLOBBERED_ALT.
19367 2018-02-22 Martin Sebor <msebor@redhat.com>
19369 PR tree-optimization/84480
19370 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
19371 to maybe_diag_stxncpy_trunc. Call it.
19372 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
19373 from gimple_fold_builtin_strcpy. Print inlining stack.
19374 (handle_builtin_stxncpy): Print inlining stack.
19375 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
19377 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
19380 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
19381 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
19382 and -fcheck-pointer-bounds are used together.
19383 (indirect_thunk_prefix): New enum.
19384 (indirect_thunk_need_prefix): New function.
19385 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
19386 "_nt" instead of "_bnd" for NOTRACK prefix.
19387 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
19388 (output_indirect_thunk_function): Likewise.
19390 (ix86_code_end): Update output_indirect_thunk_function calls.
19391 (ix86_output_indirect_branch_via_reg): Replace
19392 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
19393 (ix86_output_indirect_branch_via_push): Likewise.
19394 (ix86_output_function_return): Likewise.
19395 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
19396 incompatible with -fcf-protection=branch and
19397 -fcheck-pointer-bounds.
19399 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19402 * config/aarch64/aarch64.c (aarch64_print_address_internal):
19403 Change gcc_assert call to output_operand_lossage.
19405 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19407 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
19409 2018-02-22 DJ Delorie <dj@redhat.com>
19410 Sebastian Perta <sebastian.perta@renesas.com>
19411 Oleg Endo <olegendo@gcc.gnu.org>
19413 * config/rx/rx.c (rx_rtx_costs): New function.
19414 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
19416 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
19418 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
19420 2018-02-22 Martin Liska <mliska@suse.cz>
19423 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
19424 Add "native" as a possible value.
19426 2018-02-22 Martin Liska <mliska@suse.cz>
19429 * config/i386/i386.c (ix86_option_override_internal):
19430 Add "native" as a possible value for -march and -mtune.
19432 2018-02-22 Jakub Jelinek <jakub@redhat.com>
19435 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
19436 to all type variants.
19438 PR tree-optimization/84503
19439 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
19440 width as info->bitpos + info->bitsize - start.
19441 (merged_store_group::merge_overlapping): Simplify width computation.
19442 (check_no_overlap): New function.
19443 (imm_store_chain_info::try_coalesce_bswap): Compute expected
19444 start + width and last_order of the group, fail if check_no_overlap
19446 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
19447 to group if check_no_overlap fails.
19449 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19451 * config/rs6000/altivec.md: Delete contraint arguments to
19452 define_expand, define_split, and define_peephole2, and in
19453 define_insn_and_split if always unused.
19454 * config/rs6000/darwin.md: Ditto.
19455 * config/rs6000/dfp.md: Ditto.
19456 * config/rs6000/rs6000.md: Ditto.
19457 * config/rs6000/sync.md: Ditto.
19458 * config/rs6000/vector.md: Ditto.
19459 * config/rs6000/vsx.md: Ditto.
19461 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19463 * config/rs6000/altivec.md: Write output control strings as braced
19464 blocks instead of double-quoted strings.
19465 * config/rs6000/darwin.md: Ditto.
19466 * config/rs6000/rs6000.md: Ditto.
19467 * config/rs6000/vector.md: Ditto.
19468 * config/rs6000/vsx.md: Ditto.
19470 2018-02-21 Jason Merrill <jason@redhat.com>
19472 PR c++/84314 - ICE with templates and fastcall attribute.
19473 * attribs.c (build_type_attribute_qual_variant): Remove assert.
19475 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19477 * ipa-cp.c (determine_versionability): Fix comment typos.
19479 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19482 * ipa-cp.c (determine_versionability): Do not version functions caling
19485 2018-02-21 Martin Liska <mliska@suse.cz>
19488 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
19489 Add "native" as a possible value.
19490 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
19491 the macro when native cpu detection is available.
19493 2018-02-21 Martin Liska <mliska@suse.cz>
19496 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
19497 Add "native" as a possible value.
19498 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
19499 when native cpu detection is available.
19501 2018-02-21 Jakub Jelinek <jakub@redhat.com>
19502 Martin Sebor <msebor@redhat.com>
19504 PR tree-optimization/84478
19505 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
19507 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
19508 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
19509 support which is conservatively correct, for 2 only stay conservative
19510 for maxlen. Formatting and comment capitalization fixes. Add STRICT
19511 argument to the 2 argument get_range_strlen, adjust 6 arg
19512 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
19514 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
19515 (gimple_fold_builtin_strlen): Pass true as last argument to
19518 2018-02-20 Martin Sebor <msebor@redhat.com>
19520 PR middle-end/84095
19521 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
19522 (builtin_memref::set_base_and_offset): Same. Handle inner references.
19523 (builtin_memref::builtin_memref): Factor out parts into
19524 set_base_and_offset and call it.
19526 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19528 PR middle-end/84406
19529 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
19530 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
19531 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
19532 search at the associated MODE_INT.
19534 2018-02-20 Jeff Law <law@redhat.com>
19536 PR middle-end/82123
19537 PR tree-optimization/81592
19538 PR middle-end/79257
19539 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
19540 for range data rather than using global data.
19541 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
19542 range data rather than using global data.
19543 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
19544 pass it to children as needed.
19545 (struct directive::fmtresult): Similarly.
19546 (struct directive::set_width): Similarly.
19547 (struct directive::set_precision): Similarly.
19548 (format_integer, format_directive, parse_directive): Similarly.
19549 (format_none): Accept unnamed vr_values parameter.
19550 (format_percent, format_floating, format_character): Similarly.
19551 (format_string, format_plain): Similarly.
19552 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
19553 the EVRP range analyzer for range data rather than using global data.
19554 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
19555 gimple-ssa-evrp-analyze.h
19556 (class sprintf_dom_walker): Add after_dom_children member function.
19557 Add evrp_range_analyzer member.
19558 (sprintf_dom_walker::before_dom_children): Call into the EVRP
19559 range analyzer as needed.
19560 (sprintf_dom_walker::after_dom_children): New member function.
19561 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
19563 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
19564 (evrp_range_analyzer::pop_to_marker): Likewise.
19566 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19568 PR tree-optimization/84419
19569 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
19570 with the required type if its current type is compatible but
19573 2018-02-20 Jakub Jelinek <jakub@redhat.com>
19575 PR middle-end/82004
19576 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
19577 after vectorization.
19579 2018-02-20 Martin Liska <mliska@suse.cz>
19582 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
19583 possible values if we don't have a hint.
19585 2018-02-20 Martin Liska <mliska@suse.cz>
19589 * final.c (shorten_branches): Build align_tab array with one
19591 * opts.c (finish_options): Add alignment option limit check.
19592 (MAX_CODE_ALIGN): Likewise.
19593 (MAX_CODE_ALIGN_VALUE): Likewise.
19594 * doc/invoke.texi: Document maximum allowed option value for
19595 all -falign-* options.
19597 2018-02-19 Jakub Jelinek <jakub@redhat.com>
19600 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
19601 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
19602 * var-tracking.c (emit_note_insn_var_location): Remove all references
19603 to NOTE_INSN_CALL_ARG_LOCATION.
19604 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
19605 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
19606 Use copy_rtx_if_shared.
19607 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
19608 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
19609 (dwarf2out_var_location): Remove handling of
19610 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
19612 * final.c (final_scan_insn): Remove all references to
19613 NOTE_INSN_CALL_ARG_LOCATION.
19614 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
19615 before dumping final insns.
19616 * except.c (emit_note_eh_region_end): Remove all references to
19617 NOTE_INSN_CALL_ARG_LOCATION.
19618 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
19619 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
19620 * config/arc/arc.c (hwloop_optimize): Likewise.
19621 * config/arm/arm.c (create_fix_barrier): Likewise.
19622 * config/s390/s390.c (s390_chunkify_start): Likewise.
19623 * config/sh/sh.c (find_barrier): Likewise.
19624 * config/i386/i386.c (rest_of_insert_endbranch,
19625 ix86_seh_fixup_eh_fallthru): Likewise.
19626 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
19627 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
19628 * config/frv/frv.c (frv_function_prologue): Likewise.
19629 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
19631 (note_outside_basic_block_p): Remove all references to
19632 NOTE_INSN_CALL_ARG_LOCATION.
19633 * gengtype.c (adjust_field_rtx_def): Likewise.
19634 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
19636 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
19637 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
19640 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
19643 PR tree-optimization/84452
19644 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
19645 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
19648 2018-02-19 Martin Liska <mliska@suse.cz>
19651 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
19653 2018-02-19 Martin Liska <mliska@suse.cz>
19654 Richard Sandiford <richard.sandiford@linaro.org>
19656 PR tree-optimization/82491
19657 * gimple-fold.c (get_base_constructor): Make earlier bail out
19660 2018-02-19 Carl Love <cel@us.ibm.com>
19662 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
19663 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
19665 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
19666 P8V_BUILTIN_VEC_NEG.
19668 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
19670 * config/rl78/rl78.md (movdf): New define expand.
19672 2018-02-19 Martin Liska <mliska@suse.cz>
19675 * doc/invoke.texi: Fix typo.
19676 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
19678 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
19680 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
19681 handle rs6000_single_float and rs6000_double_float specially for
19684 2018-02-16 Jeff Law <law@redhat.com>
19686 * config/rx/rx.c (add_pop_cfi_notes): New function.;
19687 (pop_regs): Use it.
19689 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19692 * ipa-inline.c (inline_small_functions): Fix a typo.
19694 2018-02-16 Nathan Sidwell <nathan@acm.org>
19696 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
19698 2018-02-16 Carl Love <cel@us.ibm.com>
19700 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
19701 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
19702 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
19703 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
19704 expansion to P8V_BUILTIN_VEC_FLOAT2.
19706 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
19708 PR rtl-optimization/70023
19709 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
19710 src_regno into account.
19712 2018-02-16 Carl Love <cel@us.ibm.com>
19714 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
19715 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
19716 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
19717 * config/rs6000/rs6000.c: Remove case statements for
19718 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
19719 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
19720 and P9V_BUILTIN_VEC_VINSERT4B.
19721 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
19722 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
19723 * config/rs6000/vsx.md:
19724 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
19727 2018-02-16 Carl Love <cel@us.ibm.com>
19729 * config/rs6000/altivec.h: Add builtin names vec_extract4b
19731 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
19733 * config/rs6000/rs6000-c.c: Add the definitions for
19734 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
19735 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
19736 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
19737 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
19738 definition for insert4b and define insn *insert3b_internal.
19739 * doc/extend.texi: Add documentation for vec_extract4b.
19741 2018-02-16 Nathan Sidwell <nathan@acm.org>
19743 * doc/extend.texi (Backwards Compatibility): Mention friend
19744 injection. Note for-scope is deprecated.
19745 * doc/invoke.texi (-ffriend-injection): Deprecate.
19747 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
19749 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
19750 that moved to I2, also allow destinations that are a paradoxical
19751 subreg (instead of a normal reg).
19753 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
19756 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
19759 2018-02-16 Richard Biener <rguenther@suse.de>
19761 PR tree-optimization/84037
19762 PR tree-optimization/84016
19764 * config/i386/i386.c (ix86_builtin_vectorization_cost):
19765 Adjust vec_construct for the fact we need additional higher latency
19766 128bit inserts for AVX256 and AVX512 vector builds.
19767 (ix86_add_stmt_cost): Scale vector construction cost for
19770 2018-02-16 Richard Biener <rguenther@suse.de>
19772 PR tree-optimization/84417
19773 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19774 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19775 (non_rewritable_lvalue_p): Likewise, use poly-ints.
19777 2018-02-16 Martin Liska <mliska@suse.cz>
19780 * internal-fn.def (ASAN_CHECK): Set proper flags.
19781 (ASAN_MARK): Likewise.
19783 2018-02-16 Julia Koval <julia.koval@intel.com>
19785 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19786 from PTA_CANNONLAKE.
19788 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19791 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19792 Use ++iter rather than iter++ for std::list iterators.
19793 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
19794 defer deleting them until all nodes in the forest are processed. Do
19795 free even leaf nodes. Change to_process into auto_vec.
19798 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
19799 * vec.h (vec_default_construct): Use memset instead of placement new
19800 if BROKEN_VALUE_INITIALIZATION is defined.
19801 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
19802 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
19805 PR rtl-optimization/83723
19806 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
19807 * lra.c (lra_substitute_pseudo): Likewise. If true, use
19808 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
19810 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
19812 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
19814 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
19816 PR rtl-optimization/81443
19817 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
19818 from inner REGs to paradoxical SUBREGs.
19820 2018-02-16 Richard Biener <rguenther@suse.de>
19822 PR tree-optimization/84399
19823 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19824 For operands we can analyze at their definition make sure we can
19825 analyze them at each use as well.
19827 2018-02-16 Richard Biener <rguenther@suse.de>
19829 PR tree-optimization/84190
19830 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
19831 volatile accesses if the decl isn't volatile.
19833 2018-02-15 Jason Merrill <jason@redhat.com>
19835 PR c++/84314 - ICE with templates and fastcall attribute.
19836 * attribs.c (build_type_attribute_qual_variant): Don't clobber
19837 TYPE_CANONICAL on an existing type.
19839 2018-02-15 Jakub Jelinek <jakub@redhat.com>
19841 PR tree-optimization/84383
19842 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
19843 dstoff nor call operand_equal_p if dstbase is NULL.
19845 PR tree-optimization/84334
19846 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
19847 also a CONSTANT_CLASS_P, punt.
19849 2018-02-14 Jim Wilson <jimw@sifive.com>
19851 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
19852 first SMALL_OPERAND check. New local min_second_step. Move assert
19853 to where locals are set. Add TARGET_RVC support.
19854 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
19856 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
19858 * doc/invoke.texi: Correct -Wformat-overflow code sample.
19860 2018-02-14 Martin Sebor <msebor@redhat.com>
19862 PR tree-optimization/83698
19863 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
19864 arrays constrain the offset range to their bounds.
19865 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
19866 (builtin_access::overlap): Avoid setting the size of overlap if it's
19868 (maybe_diag_overlap): Also consider arrays when deciding what values
19869 of offsets to include in diagnostics.
19871 2018-02-14 Martin Sebor <msebor@redhat.com>
19874 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
19875 that correspond to the kind of a declaration.
19877 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
19880 * config/pa/pa.md: Load address of PIC label using the linkage table
19881 if the label is nonlocal.
19883 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
19885 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
19886 warning message if user requests -maltivec=be.
19887 * doc/invoke.texi: Document deprecation of -maltivec=be.
19889 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
19892 * config/rs6000/rs6000-c.c: Update definitions for
19893 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
19894 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
19896 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19899 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
19900 add _get_ssp intrinsics. Remove argument from
19901 __builtin_ia32_rdssp[d|q].
19902 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
19903 * config/i386/i386-builtin.def: Remove argument from
19904 __builtin_ia32_rdssp[d|q].
19905 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
19906 ix86_expand_special_args_builtin for _rdssp[d|q].
19907 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
19908 Clear register before usage.
19909 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
19910 Add documentation for new _get_ssp and _inc_ssp intrinsics.
19912 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
19914 PR tree-optimization/84357
19915 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
19916 operand 1 of an ARRAY_REF too.
19918 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
19921 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
19922 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19924 (set_of_reg): New struct.
19925 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
19926 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
19927 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
19929 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
19930 Split into bitclr, bitset, bitinvert patterns if appropriate.
19931 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
19932 use rx_fuse_in_memory_bitop.
19933 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
19934 to named insn, correct maximum insn length.
19936 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
19939 * machmode.def: Define a complex mode for PARTIAL_INT.
19940 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
19942 * doc/rtl.texi: Document CSPImode.
19943 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
19945 (msp430_hard_regno_nregs_with_padding): Likewise.
19947 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
19950 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
19952 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
19954 PR rtl-optimization/84169
19955 * combine.c (try_combine): New variable split_i2i3. Set it to true if
19956 we generated a parallel as new i3 and we split that to new i2 and i3
19957 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
19958 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
19959 those to i2, not i1. Partially rewrite this scan code.
19961 2018-02-13 Jakub Jelinek <jakub@redhat.com>
19964 * stor-layout.c (place_field): For variable length fields, adjust
19965 offset_align afterwards not just based on the field's alignment,
19966 but also on the size.
19968 PR middle-end/84309
19969 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
19970 of exps and logs in the use_exp2 case.
19972 2018-02-13 Jeff Law <law@redhat.com>
19974 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
19975 entry for "vector".
19977 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
19980 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
19984 * common.opt (gas-loc-support, gas-locview-support): New.
19985 (ginline-points, ginternal-reset-location-views): New.
19986 * doc/invoke.texi: Document them. Use @itemx where intended.
19987 (gvariable-location-views): Adjust.
19988 * target.def (reset_location_view): New.
19989 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
19990 (TARGET_RESET_LOCATION_VIEW): New.
19991 * doc/tm.texi: Rebuilt.
19992 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
19993 (dwarf2out_default_as_locview_support): New.
19994 (output_asm_line_debug_info): Use option variables.
19995 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
19996 (output_loc_list): Likewise.
19997 (add_high_low_attributes): Check option variables.
19998 Don't output entry view attribute in strict mode.
19999 (gen_inlined_subroutine_die): Check option variables.
20000 (dwarf2out_inline_entry): Likewise.
20001 (init_sections_and_labels): Likewise.
20002 (dwarf2out_early_finish): Likewise.
20003 (maybe_reset_location_view): New, from...
20004 (dwarf2out_var_location): ... here. Call it.
20005 * debug.h (dwarf2out_default_as_loc_support): Declare.
20006 (dwarf2out_default_as_locview_support): Declare.
20007 * hooks.c (hook_int_rtx_insn_0): New.
20008 * hooks.h (hook_int_rtx_insn_0): Declare.
20009 * toplev.c (process_options): Take -gas-loc-support and
20010 -gas-locview-support from dwarf2out. Enable
20011 -gvariable-location-views by default only with locview
20012 assembler support. Enable -ginternal-reset-location-views by
20013 default only if the target defines the corresponding hook.
20014 Enable -ginline-points by default if location views are
20015 enabled; force it disabled if statement frontiers are
20017 * tree-inline.c (expand_call_inline): Check option variables.
20018 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
20020 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
20022 PR tree-optimization/84321
20023 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
20024 handling. Also check whether the anti-range contains any values
20025 that satisfy the mask; switch to a VR_RANGE if not.
20027 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
20030 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
20032 2018-02-13 Martin Jambor <mjambor@suse.cz>
20035 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
20036 of call statements, also set location of a load to a temporary.
20038 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
20040 * config/rl78/rl78.c (add_vector_labels): New function.
20041 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
20042 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
20043 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
20044 which checks that no arguments are passed.
20045 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
20046 * doc/extend.texi: Documentation for the new attribute.
20048 2018-02-13 Andreas Schwab <schwab@suse.de>
20050 * config/riscv/linux.h (CPP_SPEC): Define.
20052 2018-02-13 Jakub Jelinek <jakub@redhat.com>
20055 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
20056 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
20057 OPTION_MASK_ISA_AES as first argument to def_builtin_const
20058 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
20059 instead of OPTION_MASK_ISA_PCLMUL as first argument to
20060 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
20061 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
20062 temporarily for AES and PCLMUL builtins.
20064 PR tree-optimization/84339
20065 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
20066 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
20069 PR middle-end/84309
20070 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
20071 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
20072 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
20074 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
20076 * omp-simd-clone.h: New file.
20077 * omp-simd-clone.c: Include omp-simd-clone.h.
20078 (expand_simd_clones): No longer static.
20079 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
20080 cgraph.h and omp-simd-clone.h.
20081 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
20082 (vect_recog_widen_shift_pattern): Formatting fix.
20083 (vect_pattern_recog_1): Don't check optab for calls.
20086 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
20087 operands[2] into a REG before using gen_lowpart on it.
20089 2018-02-12 Jeff Law <law@redhat.com>
20092 * config/sh/sh.c (find_barrier): Consider a sibling call
20095 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
20096 successfully back substituting a reg.
20098 2018-02-12 Richard Biener <rguenther@suse.de>
20100 PR tree-optimization/84037
20101 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
20102 parameter, move visited init to caller.
20103 (vect_slp_analyze_operations): Separate cost from validity
20104 check, initialize visited once for all instances.
20105 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
20107 * tree-vect-stmts.c (vect_model_simple_cost): Make early
20109 (vect_model_promotion_demotion_cost): Likewise.
20110 (vectorizable_bswap): Guard cost modeling with !slp_node
20111 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
20113 (vectorizable_call): Likewise.
20114 (vectorizable_conversion): Likewise.
20115 (vectorizable_assignment): Likewise.
20116 (vectorizable_shift): Likewise.
20117 (vectorizable_operation): Likewise.
20118 (vectorizable_store): Likewise.
20119 (vectorizable_load): Likewise.
20120 (vectorizable_condition): Likewise.
20121 (vectorizable_comparison): Likewise.
20123 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
20126 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
20127 (ASAN_MARK): Fix fnspec to account for return value, change pointer
20128 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
20130 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20132 PR middle-end/83665
20133 * params.def (inline-min-speedup): Increase from 8 to 15.
20134 (max-inline-insns-auto): Decrease from 40 to 30.
20135 * ipa-split.c (consider_split): Add some buffer for function to
20136 be considered inlining candidate.
20137 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
20140 2018-02-12 Richard Biener <rguenther@suse.de>
20142 PR tree-optimization/84037
20143 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
20144 matched stmts if we cannot swap the non-matched ones.
20146 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
20148 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
20149 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
20150 _mm_maskz_scalef_round_ss): New intrinsics.
20151 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
20152 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
20153 __builtin_ia32_scalefss_round): Remove.
20154 (__builtin_ia32_scalefsd_mask_round,
20155 __builtin_ia32_scalefss_mask_round): New intrinsics.
20156 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
20157 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
20158 ((match_operand:VF_128 2 "<round_nimm_predicate>"
20159 "<round_constraint>")): Changed to ...
20160 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
20161 "<round_scalar_constraint>")): ... this.
20162 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
20163 %0, %1, %2<round_op3>}"): Changed to ...
20164 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
20165 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
20166 %2<round_scalar_mask_op3>}"): ... this.
20167 * config/i386/subst.md (round_scalar_nimm_predicate): New.
20169 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
20171 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
20172 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
20173 (_mm_maskz_sqrt_round_ss): New intrinsics.
20174 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
20175 (__builtin_ia32_sqrtsd_mask_round)
20176 (__builtin_ia32_sqrtss_mask_round): New builtins.
20177 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
20178 (__builtin_ia32_sqrtss_round): Remove.
20179 (__builtin_ia32_sqrtsd_mask_round)
20180 (__builtin_ia32_sqrtss_mask_round): New builtins.
20181 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
20182 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
20183 ((match_operand:VF_128 1 "vector_operand"
20184 "xBm,<round_constraint>")): Changed to ...
20185 ((match_operand:VF_128 1 "vector_operand"
20186 "xBm,<round_scalar_constraint>")): ... this.
20187 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
20188 %0, %2, %<iptr>1<round_op3>}): Changed to ...
20189 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
20190 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
20191 %<iptr>1<round_scalar_mask_op3>}): ... this.
20192 ((set_attr "prefix" "<round_prefix>")): Changed to ...
20193 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
20195 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
20198 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
20199 Cast vec_cmpeq result to correct type.
20200 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
20201 Cast vec_cmpgt result to correct type.
20203 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
20205 * final.c (final_scan_insn_1): Renamed from...
20206 (final_scan_insn): ... this. New wrapper, to recover
20207 seen from the outermost call in recursive ones.
20208 * config/sparc/sparc.c (output_return): Drop seen from call.
20209 (output_sibcall): Likewise.
20210 * config/visium/visium.c (output_branch): Likewise.
20212 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
20214 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
20217 2018-02-10 Alan Modra <amodra@gmail.com>
20220 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
20221 Specify LR as an input.
20223 2018-02-10 Jakub Jelinek <jakub@redhat.com>
20226 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
20227 remove_member_access_dummy_vars): New functions.
20228 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
20229 lower_omp_1, execute_lower_omp): Use them.
20231 PR rtl-optimization/84308
20232 * shrink-wrap.c (spread_components): Release todo vector.
20234 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
20236 PR rtl-optimization/57193
20237 * ira-color.c (struct allocno_color_data): Add member
20238 conflict_allocno_hard_prefs.
20239 (update_conflict_allocno_hard_prefs): New.
20240 (bucket_allocno_compare_func): Add a preference based on
20241 conflict_allocno_hard_prefs.
20242 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
20243 (color_allocnos): Remove a dead code. Initiate
20244 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
20246 2018-02-09 Jakub Jelinek <jakub@redhat.com>
20249 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
20250 constraint from =wa to wa. Avoid a subreg on the output operand,
20251 instead use a pseudo and subreg it in a move.
20252 (p9_xxbrd_<mode>): Changed to ...
20253 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
20254 (p9_xxbrd_v2df): New expander.
20255 (p9_xxbrw_<mode>): Changed to ...
20256 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
20257 (p9_xxbrw_v4sf): New expander.
20259 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
20261 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
20263 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
20266 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
20267 multiply in 32-bit mode.
20268 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
20269 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
20272 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
20274 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
20275 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
20276 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
20277 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
20279 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
20282 * dwarf2out.c (is_trivial_indirect_ref): New function.
20283 (dwarf2out_late_global_decl): Do not generate a location
20284 attribute for variables that have a non-trivial DECL_VALUE_EXPR
20285 and that are not defined in the current unit.
20287 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
20289 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
20290 instead of a libcall for UNORDERED.
20292 2018-02-09 Tamar Christina <tamar.christina@arm.com>
20295 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
20296 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
20298 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20301 * config/s390/s390.c (s390_set_current_function): Invoke
20302 s390_indirect_branch_settings also if fndecl didn't change.
20304 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20306 * config/rs6000/rs6000.md (blockage): Set length to zero.
20308 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
20310 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
20312 2018-02-09 Jakub Jelinek <jakub@redhat.com>
20315 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
20316 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
20320 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
20321 PARALLEL incoming that failed vt_get_decl_and_offset check.
20323 PR middle-end/84237
20324 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
20325 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
20327 (get_variable_section): For decls in named .bss* sections pass true as
20328 second argument to bss_initializer_p.
20330 2018-02-09 Marek Polacek <polacek@redhat.com>
20331 Jakub Jelinek <jakub@redhat.com>
20334 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
20335 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
20336 Sync some changes from cxx_fold_indirect_ref.
20338 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20340 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
20342 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
20343 (BLOCK_INLINE_ENTRY_LABEL): New.
20344 (dwarf2out_var_location): Disregard inline entry markers.
20345 (inline_entry_data): New struct.
20346 (inline_entry_data_hasher): New hashtable type.
20347 (inline_entry_data_hasher::hash): New.
20348 (inline_entry_data_hasher::equal): New.
20349 (inline_entry_data_table): New variable.
20350 (add_high_low_attributes): Add DW_AT_entry_pc and
20351 DW_AT_GNU_entry_view attributes if a pending entry is found
20352 in inline_entry_data_table. Add old entry_pc attribute only
20353 if debug nonbinding markers are disabled.
20354 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
20355 markers are enabled.
20356 (block_within_block_p, dwarf2out_inline_entry): New.
20357 (dwarf2out_finish): Check that no entries remained in
20358 inline_entry_data_table.
20359 * final.c (reemit_insn_block_notes): Handle inline entry notes.
20360 (final_scan_insn, notice_source_line): Likewise.
20361 (rest_of_clean_state): Skip inline entry markers.
20362 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
20364 * gimple.c (gimple_build_debug_inline_entry): New.
20365 * gimple.h (enum gimple_debug_subcode): Add
20366 GIMPLE_DEBUG_INLINE_ENTRY.
20367 (gimple_build_debug_inline_entry): Declare.
20368 (gimple_debug_inline_entry_p): New.
20369 (gimple_debug_nonbind_marker_p): Adjust.
20370 * insn-notes.def (INLINE_ENTRY): New.
20371 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
20372 inline entry marker notes.
20373 (print_insn): Likewise.
20374 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
20375 (INSN_DEBUG_MARKER_KIND): Likewise.
20376 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
20377 * tree-inline.c (expand_call_inline): Build and insert
20378 debug_inline_entry stmt.
20379 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
20380 inline entry blocks early, if nonbind markers are enabled.
20381 (dump_scope_block): Dump fragment info.
20382 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
20383 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
20384 (gimple_build_debug_inline_entry): New.
20385 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
20386 Enable/disable inline entry points too.
20387 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
20388 (DEBUG_INSN): Describe inline entry markers.
20390 * common.opt (gvariable-location-views): New.
20391 (gvariable-location-views=incompat5): New.
20392 * config.in: Rebuilt.
20393 * configure: Rebuilt.
20394 * configure.ac: Test assembler for view support.
20395 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
20396 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
20397 * dwarf2out.c (var_loc_view): New typedef.
20398 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
20399 (dwarf2out_locviews_in_attribute): New.
20400 (dwarf2out_locviews_in_loclist): New.
20401 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
20402 (enum dw_line_info_opcode): Add LI_adv_address.
20403 (struct dw_line_info_table): Add view.
20404 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
20405 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
20406 (zero_view_p): New variable.
20407 (ZERO_VIEW_P): New macro.
20408 (output_asm_line_debug_info): New.
20409 (struct var_loc_node): Add view.
20410 (add_AT_view_list, AT_loc_list): New.
20411 (add_var_loc_to_decl): Add view param. Test it against last.
20412 (new_loc_list): Add view params. Record them.
20413 (AT_loc_list_ptr): Handle loc and view lists.
20414 (view_list_to_loc_list_val_node): New.
20415 (print_dw_val): Handle dw_val_class_view_list.
20416 (size_of_die): Likewise.
20417 (value_format): Likewise.
20418 (loc_list_has_views): New.
20419 (gen_llsym): Set vl_symbol too.
20420 (maybe_gen_llsym, skip_loc_list_entry): New.
20421 (dwarf2out_maybe_output_loclist_view_pair): New.
20422 (output_loc_list): Output view list or entries too.
20423 (output_view_list_offset): New.
20424 (output_die): Handle dw_val_class_view_list.
20425 (output_dwarf_version): New.
20426 (output_compilation_unit_header): Use it.
20427 (output_skeleton_debug_sections): Likewise.
20428 (output_rnglists, output_line_info): Likewise.
20429 (output_pubnames, output_aranges): Update version comments.
20430 (output_one_line_info_table): Output view numbers in asm comments.
20431 (dw_loc_list): Determine current endview, pass it to new_loc_list.
20432 Call maybe_gen_llsym.
20433 (loc_list_from_tree_1): Adjust.
20434 (add_AT_location_description): Create view list attribute if
20435 needed, check it's absent otherwise.
20436 (convert_cfa_to_fb_loc_list): Adjust.
20437 (maybe_emit_file): Call output_asm_line_debug_info for test.
20438 (dwarf2out_var_location): Reset views as needed. Precompute
20439 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
20440 attribute. Set view.
20441 (new_line_info_table): Reset next view.
20442 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
20443 (dwarf2out_source_line): Likewise. Output view resets and labels to
20444 the assembler, or select appropriate line info opcodes.
20445 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
20446 (optimize_string_length): Catch it. Adjust.
20447 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
20448 dw_val_class_view_list, and remove it if no longer needed.
20449 (hash_loc_list): Hash view numbers.
20450 (loc_list_hasher::equal): Compare them.
20451 (optimize_location_lists): Check whether a view list symbol is
20452 needed, and whether the locview attribute is present, and
20453 whether they match. Remove the locview attribute if no longer
20455 (index_location_lists): Call skip_loc_list_entry for test.
20456 (dwarf2out_finish): Call output_asm_line_debug_info for test.
20457 Use output_dwarf_version.
20458 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
20459 (struct dw_val_node): Add val_view_list.
20460 * final.c (SEEN_NEXT_VIEW): New.
20461 (set_next_view_needed): New.
20462 (clear_next_view_needed): New.
20463 (maybe_output_next_view): New.
20464 (final_start_function): Rename to...
20465 (final_start_function_1): ... this. Take pointer to FIRST,
20466 add SEEN parameter. Emit param bindings in the initial view.
20467 (final_start_function): Reintroduce SEEN-less interface.
20468 (final): Rename to...
20469 (final_1): ... this. Take SEEN parameter. Output final pending
20470 next view at the end.
20471 (final): Reintroduce seen-less interface.
20472 (final_scan_insn): Output pending next view before switching
20473 sections or ending a block. Mark the next view as needed when
20474 outputting variable locations. Notify debug backend of section
20475 changes, and of location view changes.
20476 (rest_of_handle_final): Adjust.
20477 * toplev.c (process_options): Autodetect value for debug variable
20478 location views option. Warn on incompat5 without -gdwarf-5.
20479 * doc/invoke.texi (gvariable-location-views): New.
20480 (gvariable-location-views=incompat5): New.
20481 (gno-variable-location-views): New.
20483 2018-02-08 David Malcolm <dmalcolm@redhat.com>
20485 PR tree-optimization/84136
20486 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
20487 that the result of find_edge is non-NULL.
20489 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20492 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
20493 storing integer register in SImode. Fix cost of 256 and 512
20494 byte aligned SSE register store.
20496 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20498 * config/i386/i386.c (ix86_multiplication_cost): Fix
20499 multiplication cost for TARGET_AVX512DQ.
20501 2018-02-08 Marek Polacek <polacek@redhat.com>
20503 PR tree-optimization/84238
20504 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
20507 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20509 PR tree-optimization/84265
20510 * tree-vect-stmts.c (vectorizable_store): Don't treat
20511 VMAT_CONTIGUOUS accesses as grouped.
20512 (vectorizable_load): Likewise.
20514 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20516 PR tree-optimization/81635
20517 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
20518 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
20519 (test_round_for_mask): New functions.
20520 (wide_int_cc_tests): Call test_round_for_mask.
20521 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
20522 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
20523 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
20524 range returned by get_range_info.
20526 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20529 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
20530 * symtab.c: Include builtins.h
20531 (symtab_node::output_to_lto_symbol_table_p): Move here
20532 from lto-streamer-out.c:output_symbol_p.
20533 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
20534 (output_symbol_p): Move all logic to symtab.c
20535 (produce_symtab): Update.
20537 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20539 * config/s390/s390-opts.h (enum indirect_branch): Define.
20540 * config/s390/s390-protos.h (s390_return_addr_from_memory)
20541 (s390_indirect_branch_via_thunk)
20542 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
20543 (enum s390_indirect_branch_type): Define.
20544 * config/s390/s390.c (struct s390_frame_layout, struct
20545 machine_function): Remove.
20546 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
20547 (indirect_branch_table_label_no, indirect_branch_table_name):
20549 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
20550 (enum s390_indirect_branch_option): Define.
20551 (s390_return_addr_from_memory): New function.
20552 (s390_handle_string_attribute): New function.
20553 (s390_attribute_table): Add new attribute handler.
20554 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
20555 (s390_indirect_branch_via_thunk): New function.
20556 (s390_indirect_branch_via_inline_thunk): New function.
20557 (s390_function_ok_for_sibcall): When jumping via thunk disallow
20558 sibling call optimization for non z10 compiles.
20559 (s390_emit_call): Force indirect branch target to be a single
20560 register. Add r1 clobber for non-z10 compiles.
20561 (s390_emit_epilogue): Emit return jump via return_use expander.
20562 (s390_reorg): Handle JUMP_INSNs as execute targets.
20563 (s390_option_override_internal): Perform validity checks for the
20564 new command line options.
20565 (s390_indirect_branch_attrvalue): New function.
20566 (s390_indirect_branch_settings): New function.
20567 (s390_set_current_function): Invoke s390_indirect_branch_settings.
20568 (s390_output_indirect_thunk_function): New function.
20569 (s390_code_end): Implement target hook.
20570 (s390_case_values_threshold): Implement target hook.
20571 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
20573 * config/s390/s390.h (struct s390_frame_layout)
20574 (struct machine_function): Move here from s390.c.
20575 (TARGET_INDIRECT_BRANCH_NOBP_RET)
20576 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
20577 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
20578 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
20579 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
20580 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
20581 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
20582 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
20583 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
20584 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
20585 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
20586 (mnemonic attribute): Add values which aren't recognized
20588 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
20589 pattern for branch conversion. Fix mnemonic attribute.
20590 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
20591 indirect branch via thunk if requested.
20592 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
20593 ("*indirect_jump"): Disable for branch conversion using out of
20595 ("indirect_jump_via_thunk<mode>_z10")
20596 ("indirect_jump_via_thunk<mode>")
20597 ("indirect_jump_via_inlinethunk<mode>_z10")
20598 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
20599 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
20600 ("casesi_jump_via_inlinethunk<mode>_z10")
20601 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
20602 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
20603 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
20604 ("*indirect2_jump"): Disable for branch conversion.
20605 ("casesi_jump"): Turn into expander and expand patterns for branch
20607 ("return_use"): New expander.
20608 ("*return"): Emit return via thunk and rename it to ...
20609 ("*return<mode>"): ... this one.
20610 * config/s390/s390.opt: Add new options and and enum for the
20613 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20615 * lra-constraints.c (match_reload): Unconditionally use
20616 gen_lowpart_SUBREG, rather than selecting between that
20617 and equivalent gen_rtx_SUBREG code.
20619 2018-02-08 Richard Biener <rguenther@suse.de>
20621 PR tree-optimization/84233
20622 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
20623 changed flag instead of boguously re-using phi_inserted.
20625 2018-02-08 Martin Jambor <mjambor@suse.cz>
20627 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
20628 static local variables.
20630 2018-02-08 Richard Biener <rguenther@suse.de>
20632 PR tree-optimization/84278
20633 * tree-vect-stmts.c (vectorizable_store): When looking for
20634 smaller vector types to perform grouped strided loads/stores
20635 make sure the mode is supported by the target.
20636 (vectorizable_load): Likewise.
20638 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
20640 * config/aarch64/aarch64.c (aarch64_components_for_bb):
20641 Increase LDP/STP opportunities by adding adjacent callee-saves.
20643 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
20645 PR rtl-optimization/84068
20646 PR rtl-optimization/83459
20647 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
20649 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
20651 PR tree-optimization/84224
20652 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
20653 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
20654 non-zero arguments.
20656 2018-02-07 Iain Sandoe <iain@codesourcery.com>
20659 * config/rs6000/altivec.md (*restore_world): Remove LR use.
20660 * config/rs6000/predicates.md (restore_world_operation): Adjust op
20661 count, remove one USE.
20663 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
20665 * doc/install.texi (Configuration): Document the
20666 --with-long-double-format={ibm,ieee} PowerPC configuration
20670 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
20671 Convert from define_expand to be define_insn_and_split. Rework
20672 float/double/_Float128 conversions to QI/HI/SImode to work with
20673 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
20674 conversions to QI/HImode types did a store and then a load to
20675 truncate the value. For conversions to VSX registers, don't split
20676 the insn, instead emit the code directly. Use the code iterator
20677 any_fix to combine signed and unsigned conversions.
20678 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
20679 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
20680 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20681 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
20682 (fix_<mode>di2_hw): Likewise.
20683 (fixuns_<mode>di2_hw): Likewise.
20684 (fix_<mode>si2_hw): Likewise.
20685 (fixuns_<mode>si2_hw): Likewise.
20686 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
20687 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20688 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
20689 fix<uns>_trunc<SFDF:mode>si2_p8.
20690 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
20691 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
20692 (fix<uns>_<mode>_mem): Likewise.
20693 (fctiw<u>z_<mode>_mem): Likewise.
20694 (fix<uns>_<mode>_mem): Likewise.
20695 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
20696 the register allocator from doing a direct move to the GPRs to do
20697 a store, and instead use the ISA 3.0 store byte/half-word from
20698 vector register instruction. For IEEE 128-bit floating point,
20699 also optimize stores of 32-bit ints.
20700 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
20702 2018-02-07 Alan Hayward <alan.hayward@arm.com>
20704 * genextract.c (push_pathstr_operand): New function to support
20706 (walk_rtx): Call push_pathstr_operand.
20707 (print_path): Support [a-zA-Z].
20709 2018-02-07 Richard Biener <rguenther@suse.de>
20711 PR tree-optimization/84037
20712 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
20713 (cse_and_gimplify_to_preheader): Declare.
20714 (vect_get_place_in_interleaving_chain): Likewise.
20715 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20717 (_loop_vec_info::~_loop_vec_info): Delete it.
20718 (cse_and_gimplify_to_preheader): New function.
20719 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
20720 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
20721 (vectorizable_load): Likewise. For grouped stores always base
20722 the IV on the first element.
20723 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
20724 condition before gimplifying.
20726 2018-02-07 Jakub Jelinek <jakub@redhat.com>
20728 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
20729 *DIV_EXPR and *MOD_EXPR.
20731 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
20734 * config/i386/i386.c (ix86_option_override_internal): Mask out
20735 the CF_SET bit when checking -fcf-protection.
20737 2018-02-07 Tom de Vries <tom@codesourcery.com>
20740 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
20743 2018-02-07 Richard Biener <rguenther@suse.de>
20745 PR tree-optimization/84204
20746 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
20749 PR tree-optimization/84205
20750 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
20751 special-case isl_ast_op_zdiv_r.
20753 PR tree-optimization/84223
20754 * graphite-scop-detection.c (gather_bbs::before_dom_children):
20755 Only add conditions from within the region.
20756 (gather_bbs::after_dom_children): Adjust.
20758 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
20761 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
20762 * config/avr/avr.md: Only post-reload split REG-REG moves if
20763 either register is GENERAL_REG_P.
20765 2018-02-07 Jakub Jelinek <jakub@redhat.com>
20767 PR tree-optimization/84235
20768 * tree-ssa-scopedtables.c
20769 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20770 if the subtraction is performed in floating point type where NaNs are
20771 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20772 build 1. Formatting fix.
20774 2018-02-06 Jakub Jelinek <jakub@redhat.com>
20777 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20778 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20779 and skip it regardless of bb boundaries. Use CALL_P macro,
20780 don't test INSN_P (insn) together with CALL_P or JUMP_P check
20781 unnecessarily, formatting fix.
20783 2018-02-06 Michael Collison <michael.collison@arm.com>
20785 * config/arm/thumb2.md:
20786 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20787 (*thumb_mov_notscc): Ditto.
20789 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
20792 * config/rs6000/rs6000.md (su code attribute): Use "u" for
20793 unsigned_fix, not "s".
20795 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20797 * configure.ac (gcc_fn_eh_frame_ro): New function.
20798 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
20799 correct .eh_frame permissions.
20800 * configure: Regenerate.
20802 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
20804 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
20805 irrelevant options.
20807 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20809 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20810 Display warning message for -mno-speculate-indirect-jumps.
20812 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
20814 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
20816 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
20818 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
20820 PR tree-optimization/84225
20821 * tree-eh.c (find_trapping_overflow): Only call
20822 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
20824 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20827 * config/i386/i386.c: Reimplement the check of possible options
20828 -mibt/-mshstk conbination. Change error messages.
20829 * doc/invoke.texi: Fix a typo: remove extra '='.
20831 2018-02-06 Marek Polacek <polacek@redhat.com>
20833 PR tree-optimization/84228
20834 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
20836 2018-02-06 Tamar Christina <tamar.christina@arm.com>
20839 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
20840 emitted arch directives.
20841 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
20842 __ARM_FEATURE_COPROC before changing architectures.
20844 2018-02-06 Richard Biener <rguenther@suse.de>
20846 * config/i386/i386.c (print_reg): Fix typo.
20847 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
20849 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
20851 * configure: Regenerate.
20853 2018-02-05 Martin Sebor <msebor@redhat.com>
20855 PR tree-optimization/83369
20856 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
20859 2018-02-05 Martin Liska <mliska@suse.cz>
20861 * doc/invoke.texi: Cherry-pick upstream r323995.
20863 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
20865 * ira.c (ira_init_register_move_cost): Adjust comment.
20867 2018-02-05 Martin Liska <mliska@suse.cz>
20869 PR gcov-profile/84137
20870 * doc/gcov.texi: Fix typo in documentation.
20872 2018-02-05 Martin Liska <mliska@suse.cz>
20874 PR gcov-profile/83879
20875 * doc/gcov.texi: Document necessity of --dynamic-list-data when
20876 using dlopen functionality.
20878 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
20880 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
20881 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
20882 _mm_maskz_range_ss, _mm_mask_range_round_ss,
20883 _mm_maskz_range_round_ss): New intrinsics.
20884 (__builtin_ia32_rangesd128_round)
20885 (__builtin_ia32_rangess128_round): Remove.
20886 (__builtin_ia32_rangesd128_mask_round,
20887 __builtin_ia32_rangess128_mask_round): New builtins.
20888 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
20889 __builtin_ia32_rangess128_round): Remove.
20890 (__builtin_ia32_rangesd128_mask_round,
20891 __builtin_ia32_rangess128_mask_round): New builtins.
20892 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
20893 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
20894 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
20895 "<round_saeonly_constraint>")): Changed to ...
20896 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
20897 "<round_saeonly_scalar_constraint>")): ... this.
20898 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
20899 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
20900 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
20901 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
20902 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
20904 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
20906 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
20908 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
20909 Remove all values except native, 8540 and 8548.
20911 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
20913 * config/i386/i386.c (ix86_output_function_return): Pass
20914 INVALID_REGNUM, instead of -1, as invalid register number to
20915 indirect_thunk_name and output_indirect_thunk.
20917 2018-02-02 Julia Koval <julia.koval@intel.com>
20919 * config.gcc: Add -march=icelake.
20920 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
20921 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
20922 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
20923 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
20924 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
20925 (processor_target_table): Add icelake.
20926 (ix86_option_override_internal): Handle new PTAs.
20927 (get_builtin_code_for_version): Handle icelake.
20928 (M_INTEL_COREI7_ICELAKE): New.
20929 (fold_builtin_cpu): Handle icelake.
20930 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
20931 * doc/invoke.texi: Add -march=icelake.
20933 2018-02-02 Julia Koval <julia.koval@intel.com>
20935 * config/i386/i386.c (ix86_option_override_internal): Change flags type
20936 to wide_int_bitmask.
20937 * wide-int-bitmask.h: New.
20939 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20942 * config/i386/i386.md: Replace Pmode with word_mode in
20943 builtin_setjmp_setup and builtin_longjmp to support x32.
20945 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
20949 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
20951 (rs6000_supported_cpu_names): New static variable.
20952 (linux_cpu_translation_table): Likewise.
20953 (elf_platform) <cpu>: Define new static variable and use it.
20954 Translate kernel AT_PLATFORM name to canonical name if needed.
20955 Error if platform name is unknown.
20957 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
20960 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
20962 2018-02-01 Jeff Law <law@redhat.com>
20965 * config/i386/i386.c (release_scratch_register_on_entry): Add new
20966 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
20967 the scratch if RELEASE_VIA_POP is false.
20968 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
20969 If we have to save a temporary register, decrement SIZE appropriately.
20970 Pass new arguments to release_scratch_register_on_entry.
20971 (ix86_adjust_stack_and_probe): Likewise.
20972 (ix86_emit_probe_stack_range): Pass new arguments to
20973 release_scratch_register_on_entry.
20975 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
20977 PR rtl-optimization/84157
20978 * combine.c (change_zero_ext): Use REG_P predicate in
20979 front of HARD_REGISTER_P predicate.
20981 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
20983 * config/avr/avr.c (avr_option_override): Move disabling of
20984 -fdelete-null-pointer-checks to...
20985 * common/config/avr/avr-common.c (avr_option_optimization_table):
20988 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
20990 PR tree-optimization/81635
20991 * tree-data-ref.c (split_constant_offset_1): For types that
20992 wrap on overflow, try to use range info to prove that wrapping
20995 2018-02-01 Renlin Li <renlin.li@arm.com>
20998 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
20999 TAILCALL_ADDR_REGS.
21000 (aarch64_register_move_cost): Likewise.
21001 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
21002 TAILCALL_ADDR_REGS.
21003 (REG_CLASS_NAMES): Likewise.
21004 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
21005 TAILCALL_ADDR_REGS. Remove IP registers.
21006 * config/aarch64/aarch64.md (Ucs): Update register constraint.
21008 2018-02-01 Richard Biener <rguenther@suse.de>
21010 * domwalk.h (dom_walker::dom_walker): Add additional constructor
21011 for specifying RPO order and allow NULL for that.
21012 * domwalk.c (dom_walker::dom_walker): Likewise.
21013 (dom_walker::walk): Handle NULL RPO order.
21014 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
21016 (rewrite_update_dom_walker): Likewise.
21017 (mark_def_dom_walker): Likewise.
21019 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21021 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
21022 (aarch64_maybe_expand_sve_subreg_move): Declare.
21023 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
21024 * config/aarch64/predicates.md (aarch64_any_register_operand): New
21026 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
21027 that are semantically a reverse operation.
21028 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
21029 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
21030 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
21032 (aarch64_can_change_mode_class): For big-endian, forbid changes
21033 between two SVE modes if they have different element sizes.
21035 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21037 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
21038 the TImode handling for big-endian targets.
21040 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21042 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
21043 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
21045 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
21046 Remove BSWAP handing for big-endian targets and use the form of
21047 LD1RQ appropariate for the mode.
21049 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21051 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
21052 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
21053 duplicated element.
21055 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21058 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
21059 check for operands that need to go through aarch64_sve_reload_be.
21061 2018-02-01 Jakub Jelinek <jakub@redhat.com>
21063 PR tree-optimization/81661
21064 PR tree-optimization/84117
21065 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
21066 * tree-eh.c: Include gimplify.h.
21067 (find_trapping_overflow, replace_trapping_overflow,
21068 rewrite_to_non_trapping_overflow): New functions.
21069 * tree-vect-loop.c: Include tree-eh.h.
21070 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
21071 * tree-data-ref.c: Include tree-eh.h.
21072 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
21074 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
21076 PR rtl-optimization/84123
21077 * combine.c (change_zero_ext): Check if hard register satisfies
21078 can_change_dest_mode before calling gen_lowpart_SUBREG.
21080 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
21083 * ira.c (ira_init_register_move_cost): Remove assert.
21085 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21087 PR rtl-optimization/84071
21088 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
21089 * doc/tm.texi: Regenerate.
21091 2018-01-31 Richard Biener <rguenther@suse.de>
21093 PR tree-optimization/84132
21094 * tree-data-ref.c (analyze_miv_subscript): Properly
21095 check whether evolution_function_is_affine_multivariate_p
21096 before calling gcd_of_steps_may_divide_p.
21098 2018-01-31 Julia Koval <julia.koval@intel.com>
21101 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
21102 * config/i386/i386.md (rdpid_rex64) New.
21103 (rdpid): Make 32bit only.
21105 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
21108 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
21109 an IDENTIFIER_NODE for FUNCTION_TYPE's.
21111 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21114 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21116 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21118 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21120 PR rtl-optimization/84071
21121 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
21122 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
21124 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
21126 * config/arc/arc.c (arc_handle_aux_attribute): New function.
21127 (arc_attribute_table): Add 'aux' attribute.
21128 (arc_in_small_data_p): Consider aux like variables.
21129 (arc_is_aux_reg_p): New function.
21130 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
21131 (arc_get_aux_arg): New function.
21132 (prepare_move_operands): Handle aux-register access.
21133 (arc_handle_aux_attribute): New function.
21134 * doc/extend.texi (ARC Variable attributes): Add subsection.
21136 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
21138 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
21139 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
21140 (arc_attribute_table): Add 'uncached' attribute.
21141 (arc_print_operand): Print '.di' flag for uncached memory
21143 (arc_in_small_data_p): Do not consider for small data the uncached
21145 (arc_is_uncached_mem_p): New function.
21146 * config/arc/predicates.md (compact_store_memory_operand): Check
21147 for uncached memory accesses.
21148 (nonvol_nonimm_operand): Likewise.
21149 * doc/extend.texi (ARC Type Attribute): New subsection.
21151 2018-01-31 Jakub Jelinek <jakub@redhat.com>
21154 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
21155 falign-loops=): Add Optimization flag.
21157 2018-01-30 Jeff Law <law@redhat.com>
21160 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
21161 INT_REGISTERS_SAVED. Check it prior to calling
21162 get_scratch_register_on_entry.
21163 (ix86_adjust_stack_and_probe): Similarly.
21164 (ix86_emit_probe_stack_range): Similarly.
21165 (ix86_expand_prologue): Corresponding changes.
21167 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21170 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
21171 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
21173 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
21176 * lra-constraints.c (curr_insn_transform): Process AND in the
21179 2018-01-30 Jakub Jelinek <jakub@redhat.com>
21181 PR rtl-optimization/83986
21182 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
21183 dependence against last_pending_memory_flush in addition to
21184 pending_jump_insns.
21186 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
21188 PR tree-optimization/81611
21189 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
21192 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21195 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
21198 2018-01-30 Richard Biener <rguenther@suse.de>
21199 Jakub Jelinek <jakub@redhat.com>
21201 PR tree-optimization/84111
21202 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
21203 inner loops added during recursion, as they don't have up-to-date
21206 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
21209 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
21210 (can_inline_edge_by_limits_p): ... here.
21211 (can_early_inline_edge_p, check_callers,
21212 update_caller_keys, update_callee_keys, recursive_inlining,
21213 add_new_edges_to_heap, speculation_useful_p,
21214 inline_small_functions,
21215 inline_small_functions, flatten_function,
21216 inline_to_all_callers_1): Update.
21218 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
21220 * profile-count.c (profile_count::combine_with_ipa_count): Handle
21223 2018-01-30 Richard Biener <rguenther@suse.de>
21225 PR tree-optimization/83008
21226 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
21227 invariant and constant vector uses in stmts when they need
21228 more than one stmt.
21230 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21233 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
21234 * configure: Regenerate.
21236 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21238 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
21240 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
21241 Use gen_rtx_REG rather than gen_lowpart.
21243 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21245 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
21246 rather than 0 when creating partial subregs.
21248 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21250 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
21253 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
21256 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
21257 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
21258 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
21259 flags. This restores the settings used before the 2017-07-24.
21260 Turning off pre increment/decrement/modify allows IVOPTS to
21261 optimize DF/SF loops where the index is an int.
21263 2018-01-29 Richard Biener <rguenther@suse.de>
21264 Kelvin Nilsen <kelvin@gcc.gnu.org>
21267 * tree-vect-stmts.c (vectorizable_call): Don't call
21268 targetm.vectorize_builtin_md_vectorized_function if callee is
21271 2018-01-22 Carl Love <cel@us.ibm.com>
21273 * doc/extend.tex: Fix typo in second arg in
21274 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
21276 2018-01-29 Richard Biener <rguenther@suse.de>
21278 PR tree-optimization/84086
21279 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
21280 (flush_ssaname_freelist): When SSA names were released reset
21281 the SCEV hash table.
21283 2018-01-29 Richard Biener <rguenther@suse.de>
21285 PR tree-optimization/84057
21286 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
21287 removed paths when removing edges.
21289 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
21291 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
21292 -mfunction-return=@var{choice}.
21294 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
21296 PR diagnostic/84034
21297 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
21298 Handle CR like TAB.
21299 (layout::print_source_line): Likewise.
21300 (test_get_line_width_without_trailing_whitespace): Add test cases.
21302 2018-01-27 Jakub Jelinek <jakub@redhat.com>
21304 PR middle-end/84040
21305 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
21308 2018-01-26 Jim Wilson <jimw@sifive.com>
21310 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
21312 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
21315 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21317 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
21318 and CMP + SUB-immediate -> SUBS.
21320 2018-01-26 Martin Sebor <msebor@redhat.com>
21322 PR tree-optimization/83896
21323 * tree-ssa-strlen.c (get_string_len): Rename...
21324 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
21325 Avoid assuming length is constant.
21326 (handle_char_store): Use HOST_WIDE_INT for string length.
21328 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
21331 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
21332 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
21334 2018-01-26 Richard Biener <rguenther@suse.de>
21336 PR rtl-optimization/84003
21337 * dse.c (record_store): Only record redundant stores when
21338 the earlier store aliases at least all accesses the later one does.
21340 2018-01-26 Jakub Jelinek <jakub@redhat.com>
21342 PR rtl-optimization/83985
21343 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
21344 REG_CFA_RESTORE insns.
21345 (delete_unmarked_insns): Don't ignore separate shrink wrapping
21346 REG_CFA_RESTORE insns here.
21349 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
21350 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
21352 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21354 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
21355 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
21356 (arc_init): Likewise.
21357 (arc_override_options): Likewise.
21358 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
21360 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
21362 * config/arc/arc.h (TARGET_DBNZ): Define.
21363 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
21364 properly set the tune attribute.
21365 (dbnz): Use TARGET_DBNZ guard.
21366 * config/arc/arc.opt (mtune): Add core3 option.
21368 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21370 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
21371 recognize new pic like addresses.
21372 (arc_delegitimize_address): Clean up.
21374 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21376 * config/arc/arc-arches.def: Option mrf16 valid for all
21378 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
21379 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
21380 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
21381 * config/arc/arc-tables.opt: Regenerate.
21382 * config/arc/arc.c (arc_conditional_register_usage): Handle
21383 reduced register file case.
21384 (arc_file_start): Set must have build attributes.
21385 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
21386 mrf16 option value.
21387 * config/arc/arc.opt (mrf16): Add new option.
21388 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
21389 * config/arc/genmultilib.awk: Handle new mrf16 option.
21390 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
21391 * config/arc/t-multilib: Regenerate.
21392 * doc/invoke.texi (ARC Options): Document mrf16 option.
21394 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21396 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
21397 * config/arc/arc.c (arc_handle_secure_attribute): New function.
21398 (arc_attribute_table): Add 'secure_call' attribute.
21399 (arc_print_operand): Print secure call operand.
21400 (arc_function_ok_for_sibcall): Don't optimize tail calls when
21402 (arc_is_secure_call_p): New function. * config/arc/arc.md
21403 (call_i): Add support for sjli instruction.
21404 (call_value_i): Likewise.
21405 * config/arc/constraints.md (Csc): New constraint.
21407 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21408 John Eric Martin <John.Martin@emmicro-us.com>
21410 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
21411 * config/arc/arc.c (_arc_jli_section): New struct.
21412 (arc_jli_section): New type.
21413 (rc_jli_sections): New static variable.
21414 (arc_handle_jli_attribute): New function.
21415 (arc_attribute_table): Add jli_always and jli_fixed attribute.
21416 (arc_file_end): New function.
21417 (TARGET_ASM_FILE_END): Define.
21418 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
21419 (arc_add_jli_section): New function.
21420 (jli_call_scan): Likewise.
21421 (arc_reorg): Call jli_call_scan.
21422 (arc_output_addsi): Remove 'S' from printing asm operand.
21423 (arc_is_jli_call_p): New function.
21424 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
21426 (movhi_insn): Likewise.
21427 (movsi_insn): Likewise.
21428 (movsi_set_cc_insn): Likewise.
21429 (loadqi_update): Likewise.
21430 (load_zeroextendqisi_update): Likewise.
21431 (load_signextendqisi_update): Likewise.
21432 (loadhi_update): Likewise.
21433 (load_zeroextendhisi_update): Likewise.
21434 (load_signextendhisi_update): Likewise.
21435 (loadsi_update): Likewise.
21436 (loadsf_update): Likewise.
21437 (movsicc_insn): Likewise.
21438 (bset_insn): Likewise.
21439 (bxor_insn): Likewise.
21440 (bclr_insn): Likewise.
21441 (bmsk_insn): Likewise.
21442 (bicsi3_insn): Likewise.
21443 (cmpsi_cc_c_insn): Likewise.
21444 (movsi_ne): Likewise.
21445 (movsi_cond_exec): Likewise.
21446 (clrsbsi2): Likewise.
21447 (norm_f): Likewise.
21457 (call_i): Remove 'S' asm letter, add jli instruction.
21458 (call_value_i): Likewise.
21459 * config/arc/arc.op (mjli-always): New option.
21460 * config/arc/constraints.md (Cji): New constraint.
21461 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
21463 (subsf3_fpx): Likewise.
21464 (mulsf3_fpx): Likewise.
21465 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
21467 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
21468 function attrbutes.
21469 * doc/invoke.texi (ARC): Document mjli-always option.
21471 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
21473 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
21474 avoid addition with 0 and use incw and decw where possible.
21476 2018-01-26 Richard Biener <rguenther@suse.de>
21478 PR tree-optimization/81082
21479 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
21480 association if it requires casting to unsigned.
21481 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
21482 from fold_plusminus_mult_expr to catch important cases late when
21483 range info is available.
21485 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21487 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
21488 * configure.ac (hidden_linkonce): New test.
21489 * configure: Regenerate.
21490 * config.in: Regenerate.
21492 2018-01-26 Julia Koval <julia.koval@intel.com>
21494 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
21495 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
21496 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
21497 _mm_mask_bitshuffle_epi64_mask): Fix type.
21498 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
21499 USI_FTYPE_V4DI_V4DI_USI): Remove.
21500 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
21501 __builtin_ia32_vpshufbitqmb256_mask,
21502 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
21503 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
21504 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
21506 2018-01-26 Alan Modra <amodra@gmail.com>
21509 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
21510 UNSPEC_VBPERMQ. Sort other unspecs.
21512 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
21514 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
21516 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
21518 PR middle-end/83055
21519 * predict.c (drop_profile): Do not push/pop cfun; update also
21521 (handle_missing_profiles): Fix logic looking for zero profiles.
21523 2018-01-25 Jakub Jelinek <jakub@redhat.com>
21525 PR middle-end/83977
21526 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
21527 on functions with #pragma omp declare simd or functions with simd
21529 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
21530 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
21531 Remove trailing \n from warning_at calls.
21533 2018-01-25 Tom de Vries <tom@codesourcery.com>
21536 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21537 for neutered workers.
21539 2018-01-24 Joseph Myers <joseph@codesourcery.com>
21542 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
21543 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
21545 2018-01-24 Jeff Law <law@redhat.com>
21548 * i386.c (get_probe_interval): Move to earlier point.
21549 (ix86_compute_frame_layout): If -fstack-clash-protection and
21550 the frame is larger than the probe interval, then use pushes
21551 to save registers rather than reg->mem moves.
21552 (ix86_expand_prologue): Remove conditional for int_registers_saved
21555 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
21558 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
21559 min/max for never referenced object.
21561 2018-01-24 Jakub Jelinek <jakub@redhat.com>
21563 PR middle-end/83977
21564 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
21566 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
21567 attributes from DECL_ATTRIBUTES (decl) without affecting
21568 DECL_ATTRIBUTES (current_function_decl).
21569 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
21570 functions with non-NULL DECL_ABSTRACT_ORIGIN.
21572 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
21574 PR tree-optimization/83979
21575 * fold-const.c (fold_comparison): Use constant_boolean_node
21576 instead of boolean_{true,false}_node.
21578 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
21580 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
21583 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21585 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21586 Simplify the clause that sets the length attribute.
21587 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21588 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
21589 clause that sets the length attribute.
21590 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21592 2018-01-24 Tom de Vries <tom@codesourcery.com>
21595 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
21596 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
21597 Add strict parameter.
21598 (prevent_branch_around_nothing): Insert dummy insn between branch to
21599 label and label with no ptx insn inbetween.
21600 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
21602 2018-01-24 Tom de Vries <tom@codesourcery.com>
21605 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21606 for neutered threads in warp.
21607 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
21609 2018-01-24 Richard Biener <rguenther@suse.de>
21611 PR tree-optimization/83176
21612 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
21615 2018-01-24 Richard Biener <rguenther@suse.de>
21617 PR tree-optimization/82819
21618 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
21619 code generating pluses that are no-ops in the target precision.
21621 2018-01-24 Richard Biener <rguenther@suse.de>
21623 PR middle-end/84000
21624 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
21626 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21628 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
21629 to merge probabilities.
21630 * predict.c (probably_never_executed): Also mark as cold functions
21631 with global 0 profile and guessed local profile.
21632 * profile-count.c (profile_probability::combine_with_count): New
21634 * profile-count.h (profile_probability::operator*,
21635 profile_probability::operator*=, profile_probability::operator/,
21636 profile_probability::operator/=): Reduce precision to adjusted
21637 and set value to guessed on contradictory divisions.
21638 (profile_probability::combine_with_freq): Remove.
21639 (profile_probability::combine_wiht_count): Declare.
21640 (profile_count::force_nonzero):: Set to adjusted.
21641 (profile_count::probability_in):: Set quality to adjusted.
21642 * tree-ssa-tail-merge.c (replace_block_by): Use
21643 combine_with_count.
21645 2018-01-23 Andrew Waterman <andrew@sifive.com>
21646 Jim Wilson <jimw@sifive.com>
21648 * config/riscv/riscv.c (riscv_stack_boundary): New.
21649 (riscv_option_override): Set riscv_stack_boundary. Handle
21650 riscv_preferred_stack_boundary_arg.
21651 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
21652 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
21653 (STACK_BOUNDARY): Set to riscv_stack_boundary.
21654 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
21655 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
21656 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
21658 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
21661 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
21662 of struct ix86_frame.
21663 (ix86_expand_epilogue): Likewise. Add a local variable for
21664 the reg_save_offset field in struct ix86_frame.
21666 2018-01-23 Bin Cheng <bin.cheng@arm.com>
21668 PR tree-optimization/82604
21669 * tree-loop-distribution.c (enum partition_kind): New enum item
21670 PKIND_PARTIAL_MEMSET.
21671 (partition_builtin_p): Support above new enum item.
21672 (generate_code_for_partition): Ditto.
21673 (compute_access_range): Differentiate cases that equality can be
21674 proven at all loops, the innermost loops or no loops.
21675 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
21676 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
21677 (finalize_partitions, distribute_loop): Don't fuse partition of
21678 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
21679 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
21680 parloop is enabled.
21682 2018-01-23 Martin Liska <mliska@suse.cz>
21684 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
21685 order to ignore the predictor.
21686 (PRED_POLYMORPHIC_CALL): Likewise.
21687 (PRED_RECURSIVE_CALL): Likewise.
21689 2018-01-23 Martin Liska <mliska@suse.cz>
21691 * tree-profile.c (tree_profiling): Print function header to
21692 aware reader which function we are working on.
21693 * value-prof.c (gimple_find_values_to_profile): Do not print
21694 not interesting value histograms.
21696 2018-01-23 Martin Liska <mliska@suse.cz>
21698 * profile-count.h (enum profile_quality): Add
21699 profile_uninitialized as the first value. Do not number values
21700 as they are zero based.
21701 (profile_count::verify): Update sanity check.
21702 (profile_probability::verify): Likewise.
21704 2018-01-23 Nathan Sidwell <nathan@acm.org>
21706 * doc/invoke.texi (ffor-scope): Deprecate.
21708 2018-01-23 David Malcolm <dmalcolm@redhat.com>
21710 PR tree-optimization/83510
21711 * domwalk.c (set_all_edges_as_executable): New function.
21712 (dom_walker::dom_walker): Convert bool param
21713 "skip_unreachable_blocks" to enum reachability. Move setup of
21714 edge flags to set_all_edges_as_executable and only do it when
21715 reachability is REACHABLE_BLOCKS.
21716 * domwalk.h (enum dom_walker::reachability): New enum.
21717 (dom_walker::dom_walker): Convert bool param
21718 "skip_unreachable_blocks" to enum reachability.
21719 (set_all_edges_as_executable): New decl.
21720 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
21721 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
21723 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
21724 but converting true to REACHABLE_BLOCKS.
21725 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
21727 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
21728 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
21729 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
21731 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
21732 if check_all_array_refs will be called.
21734 2018-01-23 David Malcolm <dmalcolm@redhat.com>
21736 * tree.c (selftest::test_location_wrappers): Add more test
21739 2018-01-23 David Malcolm <dmalcolm@redhat.com>
21741 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
21742 (selftest::test_bit_in_range): Likewise.
21744 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
21747 * doc/sourcebuild.texi (vect_float): Say that the selector
21748 only describes the situation when -funsafe-math-optimizations is on.
21749 (vect_float_strict): Document.
21751 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
21753 PR tree-optimization/83965
21754 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
21755 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
21756 instead of checking only for a reduction.
21757 (vect_recog_widen_sum_pattern): Likewise.
21759 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21761 * predict.c (probably_never_executed): Only use precise profile info.
21762 (compute_function_frequency): Skip after inlining hack since we now
21763 have quality checking.
21765 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21767 * profile-count.h (profile_probability::very_unlikely,
21768 profile_probability::unlikely, profile_probability::even): Set
21769 precision to guessed.
21771 2018-01-23 Richard Biener <rguenther@suse.de>
21773 PR tree-optimization/83963
21774 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21775 Properly terminate dominator walk when crossing the exit edge not
21776 when visiting its source block.
21778 2018-01-23 Jakub Jelinek <jakub@redhat.com>
21781 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21782 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21784 2018-01-22 Jakub Jelinek <jakub@redhat.com>
21786 PR tree-optimization/83957
21787 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
21788 semicolon after for body surrounded by braces.
21790 PR tree-optimization/83081
21791 * profile-count.h (profile_probability::split): New method.
21792 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21793 Use profile_probability::split.
21794 (do_compare_rtx_and_jump): Fix adjustment of probabilities
21795 when splitting a single conditional jump into 2.
21797 2018-01-22 David Malcolm <dmalcolm@redhat.com>
21799 PR tree-optimization/69452
21800 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
21803 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21805 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
21806 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
21807 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
21809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21811 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
21813 * config/rl78/rl78.md (movdi): New define_expand.
21814 * config/rl78/rl78.c (rl78_split_movdi): New function.
21816 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
21819 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
21821 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
21822 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
21823 128-bit to produce an UNSPEC move to get the double word with the
21824 signbit and then a shift directly to do signbit.
21825 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
21826 implementation with a new version that just does either a direct
21827 move or a regular move. Move memory interface to separate insns.
21828 Move insns so they are next to the expander.
21829 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
21830 with signbit move. Split big and little endian case.
21831 (signbit<mode>2_dm_mem_le): Likewise.
21832 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
21833 (signbit<mode>2_dm2): Likewise.
21835 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21837 * config/rl78/rl78.md (anddi3): New define_expand.
21839 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21841 * config/rl78/rl78.md (umindi3): New define_expand.
21843 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21845 * config/rl78/rl78.md (smindi3): New define_expand.
21847 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21849 * config/rl78/rl78.md (smaxdi3): New define_expand.
21851 2018-01-22 Carl Love <cel@us.ibm.com>
21853 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
21854 LVX_V1TI): Add macro expansion.
21855 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
21856 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
21857 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
21858 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
21859 Change check to determine if the instruction is a byte reversing
21860 entry. Fix typo in comment.
21861 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
21862 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
21863 Add def_builtin calls for new builtins.
21864 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
21865 Add define_insn expansion.
21867 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21869 * config/rl78/rl78.md (umaxdi3): New define_expand.
21871 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
21873 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
21874 for non-QImode registers.
21876 2018-01-22 Richard Biener <rguenther@suse.de>
21878 PR tree-optimization/83963
21879 * graphite-scop-detection.c (scop_detection::get_sese): Delay
21880 including the loop exit block.
21881 (scop_detection::merge_sese): Likewise.
21882 (scop_detection::add_scop): Do it here instead.
21884 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21886 * doc/sourcebuild.texi (arm_softfloat): Document.
21888 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
21891 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
21892 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
21893 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
21895 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21896 David Edelsohn <dje.gcc@gmail.com>
21899 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21900 Change "crset eq" to "crset 2".
21901 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21902 (*call_indirect_aix<mode>_nospec): Likewise.
21903 (*call_value_indirect_aix<mode>_nospec): Likewise.
21904 (*call_indirect_elfv2<mode>_nospec): Likewise.
21905 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
21906 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
21907 change assembly output from . to $.
21908 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21909 (indirect_jump<mode>_nospec): Change assembly output from . to $.
21910 (*tablejump<mode>_internal1_nospec): Likewise.
21912 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
21915 * config/sh/sh_optimize_sett_clrt.cc:
21916 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
21918 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
21920 PR tree-optimization/83940
21921 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
21922 offset_dt to vect_constant_def rather than vect_unknown_def_type.
21923 (vect_check_load_store_mask): Add a mask_dt_out parameter and
21924 use it to pass back the definition type.
21925 (vect_check_store_rhs): Likewise rhs_dt_out.
21926 (vect_build_gather_load_calls): Add a mask_dt argument and use
21927 it instead of a call to vect_is_simple_use.
21928 (vectorizable_store): Update calls to vect_check_load_store_mask
21929 and vect_check_store_rhs. Use the dt returned by the latter instead
21930 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
21931 instead of calls to vect_is_simple_use. Pass the scalar rather
21932 than the vector operand to vect_is_simple_use when handling
21933 second and subsequent copies of an rhs value.
21934 (vectorizable_load): Update calls to vect_check_load_store_mask
21935 and vect_build_gather_load_calls. Use the cached mask_dt and
21936 gs_info.offset_dt instead of calls to vect_is_simple_use.
21938 2018-01-20 Jakub Jelinek <jakub@redhat.com>
21940 PR middle-end/83945
21941 * tree-emutls.c: Include gimplify.h.
21942 (lower_emutls_2): New function.
21943 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
21944 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
21945 it before further processing.
21948 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
21949 UINTVAL (trueop1) instead of INTVAL (op1).
21951 2018-01-19 Jakub Jelinek <jakub@redhat.com>
21955 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
21956 INCOMING_FRAME_SP_OFFSET if not defined.
21957 (scan_trace): Add ENTRY argument. If true and
21958 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
21959 emit a note to adjust the CFA offset.
21960 (create_cfi_notes): Adjust scan_trace callers.
21961 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
21962 INCOMING_FRAME_SP_OFFSET in the CIE.
21963 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
21964 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
21966 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
21967 * doc/tm.texi: Regenerated.
21969 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21971 PR rtl-optimization/83147
21972 * lra-constraints.c (remove_inheritance_pseudos): Use
21973 lra_substitute_pseudo_within_insn.
21975 2018-01-19 Tom de Vries <tom@codesourcery.com>
21976 Cesar Philippidis <cesar@codesourcery.com>
21979 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
21981 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
21984 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
21985 spaces for function labels.
21987 2018-01-19 Martin Liska <mliska@suse.cz>
21989 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
21990 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
21991 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
21992 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
21993 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
21994 (PRED_CONST_RETURN): Change from 69 to 65.
21995 (PRED_NULL_RETURN): Change from 91 to 71.
21996 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
21997 (PRED_LOOP_GUARD): Change from 66 to 73.
21999 2018-01-19 Martin Liska <mliska@suse.cz>
22001 * predict.c (predict_insn_def): Add new assert.
22002 (struct branch_predictor): Change type to signed integer.
22003 (test_prediction_value_range): Amend test to cover
22004 PROB_UNINITIALIZED.
22005 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
22006 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
22007 (PRED_LOOP_ITERATIONS_MAX): Likewise.
22008 (PRED_LOOP_IV_COMPARE): Likewise.
22009 * predict.h (PROB_UNINITIALIZED): Define new constant.
22011 2018-01-19 Martin Liska <mliska@suse.cz>
22013 * predict.c (dump_prediction): Add new format for
22014 analyze_brprob.py script which is enabled with -details
22016 * profile-count.h (precise_p): New function.
22018 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
22020 PR tree-optimization/83922
22021 * tree-vect-loop.c (vect_verify_full_masking): Return false if
22022 there are no statements that need masking.
22023 (vect_active_double_reduction_p): New function.
22024 (vect_analyze_loop_operations): Use it when handling phis that
22025 are not in the loop header.
22027 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
22029 PR tree-optimization/83914
22030 * tree-vect-loop.c (vectorizable_induction): Don't convert
22031 init_expr or apply the peeling adjustment for inductions
22032 that are nested within the vectorized loop.
22034 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22036 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
22039 2018-01-18 Jakub Jelinek <jakub@redhat.com>
22043 * function.h (gimplify_parameters): Add gimple_seq * argument.
22044 * function.c: Include gimple.h and options.h.
22045 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
22046 for the added local temporaries if needed.
22047 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
22048 if there are any parameter cleanups, wrap whole body into a
22049 try/finally with the cleanups.
22051 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
22054 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
22055 Use GET_MODE_CLASS for scalar floating point.
22057 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
22061 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
22062 Fix call of call_cgraph_insertion_hooks.
22064 2018-01-18 Martin Sebor <msebor@redhat.com>
22066 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
22068 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
22071 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
22074 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
22077 * common.opt: (-ffile-prefix-map): New option.
22078 * opts.c (common_handle_option): Defer it.
22079 * opts-global.c (handle_common_deferred_options): Handle it.
22080 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
22081 * file-prefix-map.h: New file.
22082 (remap_debug_filename, add_debug_prefix_map): ...here.
22083 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
22084 * final.c (debug_prefix_map, add_debug_prefix_map
22085 remap_debug_filename): Move to...
22086 * file-prefix-map.c: New file.
22087 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
22088 generalize, get rid of alloca(), use strrchr() instead of strchr().
22089 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
22090 Implement in terms of add_prefix_map().
22091 (remap_macro_filename, remap_debug_filename): Implement in term of
22093 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
22094 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
22095 * dbxout.c: Include file-prefix-map.h.
22096 * varasm.c: Likewise.
22097 * vmsdbgout.c: Likewise.
22098 * xcoffout.c: Likewise.
22099 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
22100 * doc/cppopts.texi (-fmacro-prefix-map): Document.
22101 * doc/invoke.texi (-ffile-prefix-map): Document.
22102 (-fdebug-prefix-map): Update description.
22104 2018-01-18 Martin Liska <mliska@suse.cz>
22106 * config/i386/i386.c (indirect_thunk_name): Document that also
22108 (output_indirect_thunk): Document why both instructions
22109 (pause and lfence) are generated.
22111 2018-01-18 Richard Biener <rguenther@suse.de>
22113 PR tree-optimization/83887
22114 * graphite-scop-detection.c
22115 (scop_detection::get_nearest_dom_with_single_entry): Remove.
22116 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
22117 (scop_detection::merge_sese): Re-implement with a flood-fill
22118 algorithm that properly finds a SESE region if it exists.
22120 2018-01-18 Jakub Jelinek <jakub@redhat.com>
22123 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
22124 pointer_diff optimizations use view_convert instead of convert.
22126 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22128 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22129 Generate different code for -mno-speculate-indirect-jumps.
22130 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22131 (*call_indirect_aix<mode>): Disable for
22132 -mno-speculate-indirect-jumps.
22133 (*call_indirect_aix<mode>_nospec): New define_insn.
22134 (*call_value_indirect_aix<mode>): Disable for
22135 -mno-speculate-indirect-jumps.
22136 (*call_value_indirect_aix<mode>_nospec): New define_insn.
22137 (*sibcall_nonlocal_sysv<mode>): Generate different code for
22138 -mno-speculate-indirect-jumps.
22139 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22141 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
22143 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
22144 long double type, set the flags for noting the default long double
22145 type, even if we don't pass or return a long double type.
22147 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
22150 * ipa-inline.c (flatten_function): Do not overwrite final inlining
22153 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
22155 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
22156 support for merge[hl].
22157 (fold_mergehl_helper): New helper function.
22158 (tree-vector-builder.h): New #include for tree_vector_builder usage.
22159 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
22160 (altivec_vmrglw_direct): Add xxmrglw insn.
22162 2018-01-17 Andrew Waterman <andrew@sifive.com>
22164 * config/riscv/riscv.c (riscv_conditional_register_usage): If
22165 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
22167 2018-01-17 David Malcolm <dmalcolm@redhat.com>
22170 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
22171 call the lto_location_cache before reading the
22172 DECL_SOURCE_LOCATION of the types.
22174 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
22175 Richard Sandiford <richard.sandiford@linaro.org>
22177 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
22178 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
22179 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
22180 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
22181 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
22183 * config/aarch64/constraints.md (aarch64_movti_operand):
22185 * config/aarch64/predicates.md (Uti): Add new constraint.
22187 2018-01-17 Carl Love <cel@us.ibm.com>
22189 * config/rs6000/vsx.md (define_expand xl_len_r,
22190 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
22191 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
22193 (define_expand, define_insn): Move the shift left from the
22194 define_insn to the define_expand for lxvl and stxvl instructions.
22195 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
22196 and XL_LEN_R definitions to PURE.
22198 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
22200 * config/i386/i386.c (indirect_thunk_name): Declare regno
22201 as unsigned int. Compare regno with INVALID_REGNUM.
22202 (output_indirect_thunk): Ditto.
22203 (output_indirect_thunk_function): Ditto.
22204 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
22205 in the call to output_indirect_thunk_function.
22207 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
22209 PR middle-end/83884
22210 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
22211 rather than the size of inner_type to determine the stack slot size
22212 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
22214 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
22217 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
22220 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
22222 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
22223 endian Linux systems to optionally enable multilibs for selecting
22224 the long double type if the user configured an explicit type.
22225 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
22226 have no long double multilibs if not defined.
22227 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
22228 warn if the user used -mabi={ieee,ibm}longdouble and we built
22229 multilibs for long double.
22230 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
22231 appropriate multilib option.
22232 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
22234 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
22235 for building long double multilibs.
22236 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
22238 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
22240 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
22243 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
22245 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
22248 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
22251 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
22254 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
22256 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
22257 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
22259 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
22261 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
22262 different rtl trees depending on TARGET_64BIT.
22263 (rs6000_gen_lvx): Likewise.
22265 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
22267 * config/visium/visium.md (nop): Tweak comment.
22268 (hazard_nop): Likewise.
22270 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22272 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
22273 -mspeculate-indirect-jumps.
22274 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
22275 for -mno-speculate-indirect-jumps.
22276 (*call_indirect_elfv2<mode>_nospec): New define_insn.
22277 (*call_value_indirect_elfv2<mode>): Disable for
22278 -mno-speculate-indirect-jumps.
22279 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
22280 (indirect_jump): Emit different RTL for
22281 -mno-speculate-indirect-jumps.
22282 (*indirect_jump<mode>): Disable for
22283 -mno-speculate-indirect-jumps.
22284 (*indirect_jump<mode>_nospec): New define_insn.
22285 (tablejump): Emit different RTL for
22286 -mno-speculate-indirect-jumps.
22287 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
22288 (tablejumpsi_nospec): New define_expand.
22289 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
22290 (tablejumpdi_nospec): New define_expand.
22291 (*tablejump<mode>_internal1): Disable for
22292 -mno-speculate-indirect-jumps.
22293 (*tablejump<mode>_internal1_nospec): New define_insn.
22294 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
22297 2018-01-16 Artyom Skrobov tyomitch@gmail.com
22299 * caller-save.c (insert_save): Drop unnecessary parameter. All
22302 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22303 Richard Biener <rguenth@suse.de>
22306 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
22307 return early, inline manually is_gimple_sizepos. Make sure if we
22308 call gimplify_expr we don't end up with a gimple constant.
22309 * tree.c (variably_modified_type_p): Don't return true for
22310 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
22311 * gimplify.h (is_gimple_sizepos): Remove.
22313 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22315 PR tree-optimization/83857
22316 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
22317 vectorizable_live_operation for pure SLP statements.
22318 (vectorizable_live_operation): Handle PHIs.
22320 2018-01-16 Richard Biener <rguenther@suse.de>
22322 PR tree-optimization/83867
22323 * tree-vect-stmts.c (vect_transform_stmt): Precompute
22324 nested_in_vect_loop_p since the scalar stmt may get invalidated.
22326 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22329 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
22330 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
22331 If off is not INTEGER_CST, issue a may not be aligned warning
22332 rather than isn't aligned. Use isn%'t rather than isn't.
22333 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
22335 <case MULT_EXPR>: Improve the case when bottom and one of the
22336 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
22337 operand, in that case check if the other operand is multiple of
22338 bottom divided by the INTEGER_CST operand.
22340 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22343 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
22344 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
22345 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
22346 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
22347 * config/pa/pa.c (pa_function_arg_advance): Likewise.
22348 (pa_function_arg, pa_arg_partial_bytes): Likewise.
22349 (pa_function_arg_size): New function.
22351 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22353 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
22354 in a separate statement.
22356 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22358 PR tree-optimization/83847
22359 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
22360 group gathers and scatters.
22362 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22364 PR rtl-optimization/86620
22365 * params.def (max-sched-ready-insns): Bump minimum value to 1.
22367 PR rtl-optimization/83213
22368 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
22369 to last if both are JUMP_INSNs.
22371 PR tree-optimization/83843
22372 * gimple-ssa-store-merging.c
22373 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
22374 store_immediate_info for bswap/nop orig_stores.
22376 2018-01-15 Andrew Waterman <andrew@sifive.com>
22378 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
22380 <UDIV>: Increase cost if !TARGET_DIV.
22382 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
22384 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
22385 (define_attr "cr_logical_3op"): New.
22386 (cceq_ior_compare): Adjust.
22387 (cceq_ior_compare_complement): Adjust.
22388 (*cceq_rev_compare): Adjust.
22389 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
22390 (is_cracked_insn): Adjust.
22391 (insn_must_be_first_in_group): Adjust.
22392 * config/rs6000/40x.md: Adjust.
22393 * config/rs6000/440.md: Adjust.
22394 * config/rs6000/476.md: Adjust.
22395 * config/rs6000/601.md: Adjust.
22396 * config/rs6000/603.md: Adjust.
22397 * config/rs6000/6xx.md: Adjust.
22398 * config/rs6000/7450.md: Adjust.
22399 * config/rs6000/7xx.md: Adjust.
22400 * config/rs6000/8540.md: Adjust.
22401 * config/rs6000/cell.md: Adjust.
22402 * config/rs6000/e300c2c3.md: Adjust.
22403 * config/rs6000/e500mc.md: Adjust.
22404 * config/rs6000/e500mc64.md: Adjust.
22405 * config/rs6000/e5500.md: Adjust.
22406 * config/rs6000/e6500.md: Adjust.
22407 * config/rs6000/mpc.md: Adjust.
22408 * config/rs6000/power4.md: Adjust.
22409 * config/rs6000/power5.md: Adjust.
22410 * config/rs6000/power6.md: Adjust.
22411 * config/rs6000/power7.md: Adjust.
22412 * config/rs6000/power8.md: Adjust.
22413 * config/rs6000/power9.md: Adjust.
22414 * config/rs6000/rs64.md: Adjust.
22415 * config/rs6000/titan.md: Adjust.
22417 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22419 * config/i386/predicates.md (indirect_branch_operand): Rewrite
22420 ix86_indirect_branch_register logic.
22422 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22424 * config/i386/constraints.md (Bs): Update
22425 ix86_indirect_branch_register check. Don't check
22426 ix86_indirect_branch_register with GOT_memory_operand.
22428 * config/i386/predicates.md (GOT_memory_operand): Don't check
22429 ix86_indirect_branch_register here.
22430 (GOT32_symbol_operand): Likewise.
22432 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22434 * config/i386/predicates.md (constant_call_address_operand):
22435 Rewrite ix86_indirect_branch_register logic.
22436 (sibcall_insn_operand): Likewise.
22438 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22440 * config/i386/constraints.md (Bs): Replace
22441 ix86_indirect_branch_thunk_register with
22442 ix86_indirect_branch_register.
22444 * config/i386/i386.md (indirect_jump): Likewise.
22445 (tablejump): Likewise.
22446 (*sibcall_memory): Likewise.
22447 (*sibcall_value_memory): Likewise.
22448 Peepholes of indirect call and jump via memory: Likewise.
22449 * config/i386/i386.opt: Likewise.
22450 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22451 (GOT_memory_operand): Likewise.
22452 (call_insn_operand): Likewise.
22453 (sibcall_insn_operand): Likewise.
22454 (GOT32_symbol_operand): Likewise.
22456 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22458 PR middle-end/83837
22459 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
22460 type rather than type addr's type points to.
22461 (expand_omp_atomic_mutex): Likewise.
22462 (expand_omp_atomic): Likewise.
22464 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22467 * config/i386/i386.c (output_indirect_thunk_function): Use
22468 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
22469 for __x86_return_thunk.
22471 2018-01-15 Richard Biener <rguenther@suse.de>
22473 PR middle-end/83850
22474 * expmed.c (extract_bit_field_1): Fix typo.
22476 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22479 * config/arm/iterators.md (VF): New mode iterator.
22480 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
22481 Remove integer-related logic from pattern.
22482 (neon_vabd<mode>_3): Likewise.
22484 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22486 PR middle-end/82694
22487 * common.opt (fstrict-overflow): No longer an alias.
22488 (fwrapv-pointer): New option.
22489 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
22490 also for pointer types based on flag_wrapv_pointer.
22491 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
22492 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
22493 opts->x_flag_wrapv got set.
22494 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
22495 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
22496 POINTER_TYPE_OVERFLOW_UNDEFINED.
22497 * match.pd: Likewise in address comparison pattern.
22498 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
22500 2018-01-15 Richard Biener <rguenther@suse.de>
22503 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
22504 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
22505 Reset type names to their identifier if their TYPE_DECL doesn't
22506 have linkage (and thus is used for ODR and devirt).
22507 (save_debug_info_for_decl): Remove.
22508 (save_debug_info_for_type): Likewise.
22509 (add_tree_to_fld_list): Adjust.
22510 * tree-pretty-print.c (dump_generic_node): Make dumping of
22511 type names more robust.
22513 2018-01-15 Richard Biener <rguenther@suse.de>
22515 * BASE-VER: Bump to 8.0.1.
22517 2018-01-14 Martin Sebor <msebor@redhat.com>
22520 * builtins.c (check_access): Avoid warning when the no-warning bit
22523 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
22525 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
22526 * ira-color (allocno_hard_regs_compare): Likewise.
22528 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
22531 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
22532 Use .pushsection/.popsection.
22534 2018-01-14 Martin Sebor <msebor@redhat.com>
22537 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
22539 2018-01-14 Jakub Jelinek <jakub@redhat.com>
22541 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
22542 entry from extra_headers.
22543 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
22544 extra_headers, make the list bitwise identical to the i?86-*-* one.
22546 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22548 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
22549 -mcmodel=large with -mindirect-branch=thunk,
22550 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
22551 -mfunction-return=thunk-extern.
22552 * doc/invoke.texi: Document -mcmodel=large is incompatible with
22553 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
22554 -mfunction-return=thunk and -mfunction-return=thunk-extern.
22556 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22558 * config/i386/i386.c (print_reg): Print the name of the full
22559 integer register without '%'.
22560 (ix86_print_operand): Handle 'V'.
22561 * doc/extend.texi: Document 'V' modifier.
22563 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22565 * config/i386/constraints.md (Bs): Disallow memory operand for
22566 -mindirect-branch-register.
22568 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22569 (GOT_memory_operand): Likewise.
22570 (call_insn_operand): Likewise.
22571 (sibcall_insn_operand): Likewise.
22572 (GOT32_symbol_operand): Likewise.
22573 * config/i386/i386.md (indirect_jump): Call convert_memory_address
22574 for -mindirect-branch-register.
22575 (tablejump): Likewise.
22576 (*sibcall_memory): Likewise.
22577 (*sibcall_value_memory): Likewise.
22578 Disallow peepholes of indirect call and jump via memory for
22579 -mindirect-branch-register.
22580 (*call_pop): Replace m with Bw.
22581 (*call_value_pop): Likewise.
22582 (*sibcall_pop_memory): Replace m with Bs.
22583 * config/i386/i386.opt (mindirect-branch-register): New option.
22584 * doc/invoke.texi: Document -mindirect-branch-register option.
22586 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22588 * config/i386/i386-protos.h (ix86_output_function_return): New.
22589 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
22590 set function_return_type.
22591 (indirect_thunk_name): Add ret_p to indicate thunk for function
22593 (output_indirect_thunk_function): Pass false to
22594 indirect_thunk_name.
22595 (ix86_output_indirect_branch_via_reg): Likewise.
22596 (ix86_output_indirect_branch_via_push): Likewise.
22597 (output_indirect_thunk_function): Create alias for function
22598 return thunk if regno < 0.
22599 (ix86_output_function_return): New function.
22600 (ix86_handle_fndecl_attribute): Handle function_return.
22601 (ix86_attribute_table): Add function_return.
22602 * config/i386/i386.h (machine_function): Add
22603 function_return_type.
22604 * config/i386/i386.md (simple_return_internal): Use
22605 ix86_output_function_return.
22606 (simple_return_internal_long): Likewise.
22607 * config/i386/i386.opt (mfunction-return=): New option.
22608 (indirect_branch): Mention -mfunction-return=.
22609 * doc/extend.texi: Document function_return function attribute.
22610 * doc/invoke.texi: Document -mfunction-return= option.
22612 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22614 * config/i386/i386-opts.h (indirect_branch): New.
22615 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
22616 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
22617 with local indirect jump when converting indirect call and jump.
22618 (ix86_set_indirect_branch_type): New.
22619 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
22620 (indirectlabelno): New.
22621 (indirect_thunk_needed): Likewise.
22622 (indirect_thunk_bnd_needed): Likewise.
22623 (indirect_thunks_used): Likewise.
22624 (indirect_thunks_bnd_used): Likewise.
22625 (INDIRECT_LABEL): Likewise.
22626 (indirect_thunk_name): Likewise.
22627 (output_indirect_thunk): Likewise.
22628 (output_indirect_thunk_function): Likewise.
22629 (ix86_output_indirect_branch_via_reg): Likewise.
22630 (ix86_output_indirect_branch_via_push): Likewise.
22631 (ix86_output_indirect_branch): Likewise.
22632 (ix86_output_indirect_jmp): Likewise.
22633 (ix86_code_end): Call output_indirect_thunk_function if needed.
22634 (ix86_output_call_insn): Call ix86_output_indirect_branch if
22636 (ix86_handle_fndecl_attribute): Handle indirect_branch.
22637 (ix86_attribute_table): Add indirect_branch.
22638 * config/i386/i386.h (machine_function): Add indirect_branch_type
22639 and has_local_indirect_jump.
22640 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
22642 (tablejump): Likewise.
22643 (*indirect_jump): Use ix86_output_indirect_jmp.
22644 (*tablejump_1): Likewise.
22645 (simple_return_indirect_internal): Likewise.
22646 * config/i386/i386.opt (mindirect-branch=): New option.
22647 (indirect_branch): New.
22650 (thunk-inline): Likewise.
22651 (thunk-extern): Likewise.
22652 * doc/extend.texi: Document indirect_branch function attribute.
22653 * doc/invoke.texi: Document -mindirect-branch= option.
22655 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
22658 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
22660 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
22662 * ipa-inline.c (want_inline_small_function_p): Return false if
22663 inlining has already failed with CIF_FINAL_ERROR.
22664 (update_caller_keys): Call want_inline_small_function_p before
22666 (update_callee_keys): Likewise.
22668 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
22670 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
22672 (rs6000_quadword_masked_address_p): Likewise.
22673 (quad_aligned_load_p): Likewise.
22674 (quad_aligned_store_p): Likewise.
22675 (const_load_sequence_p): Add comment to describe the outer-most loop.
22676 (mimic_memory_attributes_and_flags): New function.
22677 (rs6000_gen_stvx): Likewise.
22678 (replace_swapped_aligned_store): Likewise.
22679 (rs6000_gen_lvx): Likewise.
22680 (replace_swapped_aligned_load): Likewise.
22681 (replace_swapped_load_constant): Capitalize argument name in
22682 comment describing this function.
22683 (rs6000_analyze_swaps): Add a third pass to search for vector loads
22684 and stores that access quad-word aligned addresses and replace
22685 with stvx or lvx instructions when appropriate.
22686 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
22687 New function prototype.
22688 (rs6000_quadword_masked_address_p): Likewise.
22689 (rs6000_gen_lvx): Likewise.
22690 (rs6000_gen_stvx): Likewise.
22691 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
22692 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
22693 when memory address is aligned.
22694 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
22695 this split to select lvx instruction when memory address is aligned.
22696 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
22697 instruction when memory address is aligned.
22698 (*vsx_le_perm_load_v16qi): Likewise.
22699 (four unnamed splitters): Modify to select the stvx instruction
22700 when memory is aligned.
22702 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
22704 * predict.c (determine_unlikely_bbs): Handle correctly BBs
22705 which appears in the queue multiple times.
22707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22708 Alan Hayward <alan.hayward@arm.com>
22709 David Sherwood <david.sherwood@arm.com>
22711 * tree-vectorizer.h (vec_lower_bound): New structure.
22712 (_loop_vec_info): Add check_nonzero and lower_bounds.
22713 (LOOP_VINFO_CHECK_NONZERO): New macro.
22714 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
22715 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
22716 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
22717 fields. Make seg_len the distance travelled, not including the
22719 (dr_direction_indicator): Declare.
22720 (dr_zero_step_indicator): Likewise.
22721 (dr_known_forward_stride_p): Likewise.
22722 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
22724 (runtime_alias_check_p): Allow runtime alias checks with
22726 (operator ==): Compare access_size and align.
22727 (prune_runtime_alias_test_list): Rework for new distinction between
22728 the access_size and seg_len.
22729 (create_intersect_range_checks_index): Likewise. Cope with polynomial
22731 (get_segment_min_max): New function.
22732 (create_intersect_range_checks): Use it.
22733 (dr_step_indicator): New function.
22734 (dr_direction_indicator): Likewise.
22735 (dr_zero_step_indicator): Likewise.
22736 (dr_known_forward_stride_p): Likewise.
22737 * tree-loop-distribution.c (data_ref_segment_size): Return
22738 DR_STEP * (niters - 1).
22739 (compute_alias_check_pairs): Update call to the dr_with_seg_len
22741 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
22742 (vect_preserves_scalar_order_p): New function, split out from...
22743 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
22744 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
22745 (vect_vfa_access_size): New function.
22746 (vect_vfa_align): Likewise.
22747 (vect_compile_time_alias): Take access_size_a and access_b arguments.
22748 (dump_lower_bound): New function.
22749 (vect_check_lower_bound): Likewise.
22750 (vect_small_gap_p): Likewise.
22751 (vectorizable_with_step_bound_p): Likewise.
22752 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
22753 depencies if the vectorization factor is 1. Convert the checks
22754 for nonzero steps into checks on the bounds of DR_STEP. Try using
22755 a bunds check for variable steps if the minimum required step is
22756 relatively small. Update calls to the dr_with_seg_len
22757 constructor and to vect_compile_time_alias.
22758 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
22760 (vect_loop_versioning): Call it.
22761 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
22763 (vect_estimate_min_profitable_iters): Account for any bounds checks.
22765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22766 Alan Hayward <alan.hayward@arm.com>
22767 David Sherwood <david.sherwood@arm.com>
22769 * doc/sourcebuild.texi (vect_scatter_store): Document.
22770 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22772 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22774 * genopinit.c (main): Add supports_vec_scatter_store and
22775 supports_vec_scatter_store_cached to target_optabs.
22776 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22777 IFN_MASK_SCATTER_STORE.
22778 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22780 * internal-fn.h (internal_store_fn_p): Declare.
22781 (internal_fn_stored_value_index): Likewise.
22782 * internal-fn.c (scatter_store_direct): New macro.
22783 (expand_scatter_store_optab_fn): New function.
22784 (direct_scatter_store_optab_supported_p): New macro.
22785 (internal_store_fn_p): New function.
22786 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22787 IFN_MASK_SCATTER_STORE.
22788 (internal_fn_mask_index): Likewise.
22789 (internal_fn_stored_value_index): New function.
22790 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22791 for scatter stores.
22792 * optabs-query.h (supports_vec_scatter_store_p): Declare.
22793 * optabs-query.c (supports_vec_scatter_store_p): New function.
22794 * tree-vectorizer.h (vect_get_store_rhs): Declare.
22795 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
22796 true for scatter stores.
22797 (vect_gather_scatter_fn_p): Handle scatter stores too.
22798 (vect_check_gather_scatter): Consider using scatter stores if
22799 supports_vec_scatter_store_p.
22800 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
22801 scatter stores too.
22802 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22803 internal_fn_stored_value_index.
22804 (check_load_store_masking): Handle scatter stores too.
22805 (vect_get_store_rhs): Make public.
22806 (vectorizable_call): Use internal_store_fn_p.
22807 (vectorizable_store): Handle scatter store internal functions.
22808 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
22809 when deciding whether the end of the group has been reached.
22810 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
22811 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
22812 (mask_scatter_store<mode>): New insns.
22814 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22815 Alan Hayward <alan.hayward@arm.com>
22816 David Sherwood <david.sherwood@arm.com>
22818 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
22819 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
22820 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
22822 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
22823 Use vect_truncate_gather_scatter_offset if we can't treat the
22824 operation as a normal gather load or scatter store.
22825 (get_group_load_store_type): Take the gather_scatter_info
22826 as argument. Try using a gather load or scatter store for
22827 single-element groups.
22828 (get_load_store_type): Update calls to get_group_load_store_type
22829 and vect_use_strided_gather_scatters_p.
22831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22832 Alan Hayward <alan.hayward@arm.com>
22833 David Sherwood <david.sherwood@arm.com>
22835 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
22836 optional tree argument.
22837 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
22839 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
22840 but continue to use the current value as a fallback.
22841 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
22842 to compare the updates.
22843 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
22844 (get_load_store_type): Use it when handling a strided access.
22845 (vect_get_strided_load_store_ops): New function.
22846 (vect_get_data_ptr_increment): Likewise.
22847 (vectorizable_load): Handle strided gather loads. Always pass
22848 a step to vect_create_data_ref_ptr and bump_vector_ptr.
22850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22851 Alan Hayward <alan.hayward@arm.com>
22852 David Sherwood <david.sherwood@arm.com>
22854 * doc/md.texi (gather_load@var{m}): Document.
22855 (mask_gather_load@var{m}): Likewise.
22856 * genopinit.c (main): Add supports_vec_gather_load and
22857 supports_vec_gather_load_cached to target_optabs.
22858 * optabs-tree.c (init_tree_optimization_optabs): Use
22859 ggc_cleared_alloc to allocate target_optabs.
22860 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
22861 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
22863 * internal-fn.h (internal_load_fn_p): Declare.
22864 (internal_gather_scatter_fn_p): Likewise.
22865 (internal_fn_mask_index): Likewise.
22866 (internal_gather_scatter_fn_supported_p): Likewise.
22867 * internal-fn.c (gather_load_direct): New macro.
22868 (expand_gather_load_optab_fn): New function.
22869 (direct_gather_load_optab_supported_p): New macro.
22870 (direct_internal_fn_optab): New function.
22871 (internal_load_fn_p): Likewise.
22872 (internal_gather_scatter_fn_p): Likewise.
22873 (internal_fn_mask_index): Likewise.
22874 (internal_gather_scatter_fn_supported_p): Likewise.
22875 * optabs-query.c (supports_at_least_one_mode_p): New function.
22876 (supports_vec_gather_load_p): Likewise.
22877 * optabs-query.h (supports_vec_gather_load_p): Declare.
22878 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
22879 and memory_type field.
22880 (NUM_PATTERNS): Bump to 15.
22881 * tree-vect-data-refs.c: Include internal-fn.h.
22882 (vect_gather_scatter_fn_p): New function.
22883 (vect_describe_gather_scatter_call): Likewise.
22884 (vect_check_gather_scatter): Try using internal functions for
22885 gather loads. Recognize existing calls to a gather load function.
22886 (vect_analyze_data_refs): Consider using gather loads if
22887 supports_vec_gather_load_p.
22888 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
22889 (vect_get_gather_scatter_offset_type): Likewise.
22890 (vect_convert_mask_for_vectype): Likewise.
22891 (vect_add_conversion_to_patterm): Likewise.
22892 (vect_try_gather_scatter_pattern): Likewise.
22893 (vect_recog_gather_scatter_pattern): New pattern recognizer.
22894 (vect_vect_recog_func_ptrs): Add it.
22895 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
22896 internal_fn_mask_index and internal_gather_scatter_fn_p.
22897 (check_load_store_masking): Take the gather_scatter_info as an
22898 argument and handle gather loads.
22899 (vect_get_gather_scatter_ops): New function.
22900 (vectorizable_call): Check internal_load_fn_p.
22901 (vectorizable_load): Likewise. Handle gather load internal
22903 (vectorizable_store): Update call to check_load_store_masking.
22904 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
22905 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
22906 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
22907 (aarch64_gather_scale_operand_d): New predicates.
22908 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
22909 (mask_gather_load<mode>): New insns.
22911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22912 Alan Hayward <alan.hayward@arm.com>
22913 David Sherwood <david.sherwood@arm.com>
22915 * optabs.def (fold_left_plus_optab): New optab.
22916 * doc/md.texi (fold_left_plus_@var{m}): Document.
22917 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
22918 * internal-fn.c (fold_left_direct): Define.
22919 (expand_fold_left_optab_fn): Likewise.
22920 (direct_fold_left_optab_supported_p): Likewise.
22921 * fold-const-call.c (fold_const_fold_left): New function.
22922 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
22923 * tree-parloops.c (valid_reduction_p): New function.
22924 (gather_scalar_reductions): Use it.
22925 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
22926 (vect_finish_replace_stmt): Declare.
22927 * tree-vect-loop.c (fold_left_reduction_fn): New function.
22928 (needs_fold_left_reduction_p): New function, split out from...
22929 (vect_is_simple_reduction): ...here. Accept reductions that
22930 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
22931 (vect_force_simple_reduction): Also store the reduction type in
22932 the assignment's STMT_VINFO_REDUC_TYPE.
22933 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
22934 (merge_with_identity): New function.
22935 (vect_expand_fold_left): Likewise.
22936 (vectorize_fold_left_reduction): Likewise.
22937 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
22938 scalar phi in place for it. Check for target support and reject
22939 cases that would reassociate the operation. Defer the transform
22940 phase to vectorize_fold_left_reduction.
22941 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
22942 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
22943 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
22945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22947 * tree-if-conv.c (predicate_mem_writes): Remove redundant
22948 call to ifc_temp_var.
22950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22951 Alan Hayward <alan.hayward@arm.com>
22952 David Sherwood <david.sherwood@arm.com>
22954 * target.def (legitimize_address_displacement): Take the original
22955 offset as a poly_int.
22956 * targhooks.h (default_legitimize_address_displacement): Update
22958 * targhooks.c (default_legitimize_address_displacement): Likewise.
22959 * doc/tm.texi: Regenerate.
22960 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
22961 as an argument, moving assert of ad->disp == ad->disp_term to...
22962 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
22963 Try calling targetm.legitimize_address_displacement before expanding
22964 the address rather than afterwards, and adjust for the new interface.
22965 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
22966 Match the new hook interface. Handle SVE addresses.
22967 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
22968 new hook interface.
22970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22972 * Makefile.in (OBJS): Add early-remat.o.
22973 * target.def (select_early_remat_modes): New hook.
22974 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
22975 * doc/tm.texi: Regenerate.
22976 * targhooks.h (default_select_early_remat_modes): Declare.
22977 * targhooks.c (default_select_early_remat_modes): New function.
22978 * timevar.def (TV_EARLY_REMAT): New timevar.
22979 * passes.def (pass_early_remat): New pass.
22980 * tree-pass.h (make_pass_early_remat): Declare.
22981 * early-remat.c: New file.
22982 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
22984 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
22986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22987 Alan Hayward <alan.hayward@arm.com>
22988 David Sherwood <david.sherwood@arm.com>
22990 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
22991 vfm1 with a bound_epilog parameter.
22992 (vect_do_peeling): Update calls accordingly, and move the prologue
22993 call earlier in the function. Treat the base bound_epilog as 0 for
22994 fully-masked loops and retain vf - 1 for other loops. Add 1 to
22995 this base when peeling for gaps.
22996 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
22997 with fully-masked loops.
22998 (vect_estimate_min_profitable_iters): Handle the single peeled
22999 iteration in that case.
23001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23002 Alan Hayward <alan.hayward@arm.com>
23003 David Sherwood <david.sherwood@arm.com>
23005 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
23006 single-element interleaving even if the size is not a power of 2.
23007 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
23008 accesses for single-element interleaving if the group size is
23011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23012 Alan Hayward <alan.hayward@arm.com>
23013 David Sherwood <david.sherwood@arm.com>
23015 * doc/md.texi (fold_extract_last_@var{m}): Document.
23016 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
23017 * optabs.def (fold_extract_last_optab): New optab.
23018 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
23019 * internal-fn.c (fold_extract_direct): New macro.
23020 (expand_fold_extract_optab_fn): Likewise.
23021 (direct_fold_extract_optab_supported_p): Likewise.
23022 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
23023 * tree-vect-loop.c (vect_model_reduction_cost): Handle
23024 EXTRACT_LAST_REDUCTION.
23025 (get_initial_def_for_reduction): Do not create an initial vector
23026 for EXTRACT_LAST_REDUCTION reductions.
23027 (vectorizable_reduction): Leave the scalar phi in place for
23028 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
23029 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
23030 epilogue code for EXTRACT_LAST_REDUCTION and defer the
23031 transform phase to vectorizable_condition.
23032 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
23034 (vect_finish_stmt_generation): ...here.
23035 (vect_finish_replace_stmt): New function.
23036 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
23037 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
23039 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
23041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23042 Alan Hayward <alan.hayward@arm.com>
23043 David Sherwood <david.sherwood@arm.com>
23045 * doc/md.texi (extract_last_@var{m}): Document.
23046 * optabs.def (extract_last_optab): New optab.
23047 * internal-fn.def (EXTRACT_LAST): New internal function.
23048 * internal-fn.c (cond_unary_direct): New macro.
23049 (expand_cond_unary_optab_fn): Likewise.
23050 (direct_cond_unary_optab_supported_p): Likewise.
23051 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
23052 loops using EXTRACT_LAST.
23053 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
23054 (extract_last_<mode>): ...this optab.
23055 (vec_extract<mode><Vel>): Update accordingly.
23057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23058 Alan Hayward <alan.hayward@arm.com>
23059 David Sherwood <david.sherwood@arm.com>
23061 * target.def (empty_mask_is_expensive): New hook.
23062 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
23063 * doc/tm.texi: Regenerate.
23064 * targhooks.h (default_empty_mask_is_expensive): Declare.
23065 * targhooks.c (default_empty_mask_is_expensive): New function.
23066 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
23067 if the target says that empty masks are expensive.
23068 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
23070 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
23072 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23073 Alan Hayward <alan.hayward@arm.com>
23074 David Sherwood <david.sherwood@arm.com>
23076 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
23077 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
23078 (vect_use_loop_mask_for_alignment_p): New function.
23079 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
23080 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
23081 niters_skip argument. Make sure that the first niters_skip elements
23082 of the first iteration are inactive.
23083 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
23084 Update call to vect_set_loop_masks_directly.
23085 (get_misalign_in_elems): New function, split out from...
23086 (vect_gen_prolog_loop_niters): ...here.
23087 (vect_update_init_of_dr): Take a code argument that specifies whether
23088 the adjustment should be added or subtracted.
23089 (vect_update_init_of_drs): Likewise.
23090 (vect_prepare_for_masked_peels): New function.
23091 (vect_do_peeling): Skip prologue peeling if we're using a mask
23092 instead. Update call to vect_update_inits_of_drs.
23093 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23095 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
23096 alignment. Do not include the number of peeled iterations in
23097 the minimum threshold in that case.
23098 (vectorizable_induction): Adjust the start value down by
23099 LOOP_VINFO_MASK_SKIP_NITERS iterations.
23100 (vect_transform_loop): Call vect_prepare_for_masked_peels.
23101 Take the number of skipped iterations into account when calculating
23103 * tree-vect-stmts.c (vect_gen_while_not): New function.
23105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23106 Alan Hayward <alan.hayward@arm.com>
23107 David Sherwood <david.sherwood@arm.com>
23109 * doc/sourcebuild.texi (vect_fully_masked): Document.
23110 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
23111 default value to 0.
23112 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
23114 (vect_analyze_loop_2): ...here. Don't check the vectorization
23115 factor against the number of loop iterations if the loop is
23118 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23119 Alan Hayward <alan.hayward@arm.com>
23120 David Sherwood <david.sherwood@arm.com>
23122 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
23123 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
23124 (dump_groups): Update accordingly.
23125 (iv_use::mem_type): New member variable.
23126 (address_p): New function.
23127 (record_use): Add a mem_type argument and initialize the new
23129 (record_group_use): Add a mem_type argument. Use address_p.
23130 Remove obsolete null checks of base_object. Update call to record_use.
23131 (find_interesting_uses_op): Update call to record_group_use.
23132 (find_interesting_uses_cond): Likewise.
23133 (find_interesting_uses_address): Likewise.
23134 (get_mem_type_for_internal_fn): New function.
23135 (find_address_like_use): Likewise.
23136 (find_interesting_uses_stmt): Try find_address_like_use before
23137 calling find_interesting_uses_op.
23138 (addr_offset_valid_p): Use the iv mem_type field as the type
23139 of the addressed memory.
23140 (add_autoinc_candidates): Likewise.
23141 (get_address_cost): Likewise.
23142 (split_small_address_groups_p): Use address_p.
23143 (split_address_groups): Likewise.
23144 (add_iv_candidate_for_use): Likewise.
23145 (autoinc_possible_for_pair): Likewise.
23146 (rewrite_groups): Likewise.
23147 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
23148 (determine_group_iv_cost): Update after split of USE_ADDRESS.
23149 (get_alias_ptr_type_for_ptr_address): New function.
23150 (rewrite_use_address): Rewrite address uses in calls that were
23151 identified by find_address_like_use.
23153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23154 Alan Hayward <alan.hayward@arm.com>
23155 David Sherwood <david.sherwood@arm.com>
23157 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
23159 * gimple-expr.h (is_gimple_addressable: Likewise.
23160 * gimple-expr.c (is_gimple_address): Likewise.
23161 * internal-fn.c (expand_call_mem_ref): New function.
23162 (expand_mask_load_optab_fn): Use it.
23163 (expand_mask_store_optab_fn): Likewise.
23165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23166 Alan Hayward <alan.hayward@arm.com>
23167 David Sherwood <david.sherwood@arm.com>
23169 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
23170 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
23171 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
23172 (cond_umax@var{mode}): Document.
23173 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
23174 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
23175 (cond_umin_optab, cond_umax_optab): New optabs.
23176 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
23177 (COND_IOR, COND_XOR): New internal functions.
23178 * internal-fn.h (get_conditional_internal_fn): Declare.
23179 * internal-fn.c (cond_binary_direct): New macro.
23180 (expand_cond_binary_optab_fn): Likewise.
23181 (direct_cond_binary_optab_supported_p): Likewise.
23182 (get_conditional_internal_fn): New function.
23183 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
23184 Cope with reduction statements that are vectorized as calls rather
23186 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
23187 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
23188 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
23189 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
23190 (UNSPEC_COND_EOR): New unspecs.
23191 (optab): Add mappings for them.
23192 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
23193 (sve_int_op, sve_fp_op): New int attributes.
23195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23196 Alan Hayward <alan.hayward@arm.com>
23197 David Sherwood <david.sherwood@arm.com>
23199 * optabs.def (while_ult_optab): New optab.
23200 * doc/md.texi (while_ult@var{m}@var{n}): Document.
23201 * internal-fn.def (WHILE_ULT): New internal function.
23202 * internal-fn.h (direct_internal_fn_supported_p): New override
23203 that takes two types as argument.
23204 * internal-fn.c (while_direct): New macro.
23205 (expand_while_optab_fn): New function.
23206 (convert_optab_supported_p): Likewise.
23207 (direct_while_optab_supported_p): New macro.
23208 * wide-int.h (wi::udiv_ceil): New function.
23209 * tree-vectorizer.h (rgroup_masks): New structure.
23210 (vec_loop_masks): New typedef.
23211 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
23212 and fully_masked_p.
23213 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
23214 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
23215 (vect_max_vf): New function.
23216 (slpeel_make_loop_iterate_ntimes): Delete.
23217 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
23218 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
23219 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
23220 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
23221 internal-fn.h, stor-layout.h and optabs-query.h.
23222 (vect_set_loop_mask): New function.
23223 (add_preheader_seq): Likewise.
23224 (add_header_seq): Likewise.
23225 (interleave_supported_p): Likewise.
23226 (vect_maybe_permute_loop_masks): Likewise.
23227 (vect_set_loop_masks_directly): Likewise.
23228 (vect_set_loop_condition_masked): Likewise.
23229 (vect_set_loop_condition_unmasked): New function, split out from
23230 slpeel_make_loop_iterate_ntimes.
23231 (slpeel_make_loop_iterate_ntimes): Rename to..
23232 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
23233 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
23234 (vect_do_peeling): Update call accordingly.
23235 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
23237 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23238 mask_compare_type, can_fully_mask_p and fully_masked_p.
23239 (release_vec_loop_masks): New function.
23240 (_loop_vec_info): Use it to free the loop masks.
23241 (can_produce_all_loop_masks_p): New function.
23242 (vect_get_max_nscalars_per_iter): Likewise.
23243 (vect_verify_full_masking): Likewise.
23244 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
23245 retries, and free the mask rgroups before retrying. Check loop-wide
23246 reasons for disallowing fully-masked loops. Make the final decision
23247 about whether use a fully-masked loop or not.
23248 (vect_estimate_min_profitable_iters): Do not assume that peeling
23249 for the number of iterations will be needed for fully-masked loops.
23250 (vectorizable_reduction): Disable fully-masked loops.
23251 (vectorizable_live_operation): Likewise.
23252 (vect_halve_mask_nunits): New function.
23253 (vect_double_mask_nunits): Likewise.
23254 (vect_record_loop_mask): Likewise.
23255 (vect_get_loop_mask): Likewise.
23256 (vect_transform_loop): Handle the case in which the final loop
23257 iteration might handle a partial vector. Call vect_set_loop_condition
23258 instead of slpeel_make_loop_iterate_ntimes.
23259 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
23260 (check_load_store_masking): New function.
23261 (prepare_load_store_mask): Likewise.
23262 (vectorizable_store): Handle fully-masked loops.
23263 (vectorizable_load): Likewise.
23264 (supportable_widening_operation): Use vect_halve_mask_nunits for
23266 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
23267 (vect_gen_while): New function.
23268 * config/aarch64/aarch64.md (umax<mode>3): New expander.
23269 (aarch64_uqdec<mode>): New insn.
23271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23272 Alan Hayward <alan.hayward@arm.com>
23273 David Sherwood <david.sherwood@arm.com>
23275 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
23276 (reduc_xor_scal_optab): New optabs.
23277 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
23278 (reduc_xor_scal_@var{m}): Document.
23279 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
23280 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
23281 internal functions.
23282 * fold-const-call.c (fold_const_call): Handle them.
23283 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
23284 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
23285 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
23286 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
23287 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
23288 (UNSPEC_XORV): New unspecs.
23289 (optab): Add entries for them.
23290 (BITWISEV): New int iterator.
23291 (bit_reduc_op): New int attributes.
23293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23294 Alan Hayward <alan.hayward@arm.com>
23295 David Sherwood <david.sherwood@arm.com>
23297 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
23298 * internal-fn.def (VEC_SHL_INSERT): New internal function.
23299 * optabs.def (vec_shl_insert_optab): New optab.
23300 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
23301 (duplicate_and_interleave): Likewise.
23302 * tree-vect-loop.c: Include internal-fn.h.
23303 (neutral_op_for_slp_reduction): New function, split out from
23304 get_initial_defs_for_reduction.
23305 (get_initial_def_for_reduction): Handle option 2 for variable-length
23306 vectors by loading the neutral value into a vector and then shifting
23307 the initial value into element 0.
23308 (get_initial_defs_for_reduction): Replace the code argument with
23309 the neutral value calculated by neutral_op_for_slp_reduction.
23310 Use gimple_build_vector for constant-length vectors.
23311 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
23312 but the first group_size elements have a neutral value.
23313 Use duplicate_and_interleave otherwise.
23314 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
23315 Update call to get_initial_defs_for_reduction. Handle SLP
23316 reductions for variable-length vectors by creating one vector
23317 result for each scalar result, with the elements associated
23318 with other scalar results stubbed out with the neutral value.
23319 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
23320 Require IFN_VEC_SHL_INSERT for double reductions on
23321 variable-length vectors, or SLP reductions that have
23322 a neutral value. Require can_duplicate_and_interleave_p
23323 support for variable-length unchained SLP reductions if there
23324 is no neutral value, such as for MIN/MAX reductions. Also require
23325 the number of vector elements to be a multiple of the number of
23326 SLP statements when doing variable-length unchained SLP reductions.
23327 Update call to vect_create_epilog_for_reduction.
23328 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
23329 and remove initial values.
23330 (duplicate_and_interleave): Make public.
23331 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
23332 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
23334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23335 Alan Hayward <alan.hayward@arm.com>
23336 David Sherwood <david.sherwood@arm.com>
23338 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
23339 (can_duplicate_and_interleave_p): New function.
23340 (vect_get_and_check_slp_defs): Take the vector of statements
23341 rather than just the current one. Remove excess parentheses.
23342 Restriction rejectinon of vect_constant_def and vect_external_def
23343 for variable-length vectors to boolean types, or types for which
23344 can_duplicate_and_interleave_p is false.
23345 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
23346 (duplicate_and_interleave): New function.
23347 (vect_get_constant_vectors): Use gimple_build_vector for
23348 constant-length vectors and suitable variable-length constant
23349 vectors. Use duplicate_and_interleave for other variable-length
23350 vectors. Don't defer the update when inserting new statements.
23352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23353 Alan Hayward <alan.hayward@arm.com>
23354 David Sherwood <david.sherwood@arm.com>
23356 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
23357 min_profitable_iters doesn't go negative.
23359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23360 Alan Hayward <alan.hayward@arm.com>
23361 David Sherwood <david.sherwood@arm.com>
23363 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
23364 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
23365 * optabs.def (vec_mask_load_lanes_optab): New optab.
23366 (vec_mask_store_lanes_optab): Likewise.
23367 * internal-fn.def (MASK_LOAD_LANES): New internal function.
23368 (MASK_STORE_LANES): Likewise.
23369 * internal-fn.c (mask_load_lanes_direct): New macro.
23370 (mask_store_lanes_direct): Likewise.
23371 (expand_mask_load_optab_fn): Handle masked operations.
23372 (expand_mask_load_lanes_optab_fn): New macro.
23373 (expand_mask_store_optab_fn): Handle masked operations.
23374 (expand_mask_store_lanes_optab_fn): New macro.
23375 (direct_mask_load_lanes_optab_supported_p): Likewise.
23376 (direct_mask_store_lanes_optab_supported_p): Likewise.
23377 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
23379 (vect_load_lanes_supported): Likewise.
23380 * tree-vect-data-refs.c (strip_conversion): New function.
23381 (can_group_stmts_p): Likewise.
23382 (vect_analyze_data_ref_accesses): Use it instead of checking
23383 for a pair of assignments.
23384 (vect_store_lanes_supported): Take a masked_p parameter.
23385 (vect_load_lanes_supported): Likewise.
23386 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
23387 vect_store_lanes_supported and vect_load_lanes_supported.
23388 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
23389 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
23390 parameter. Don't allow gaps for masked accesses.
23391 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
23392 and vect_load_lanes_supported.
23393 (get_load_store_type): Take a masked_p parameter and update
23394 call to get_group_load_store_type.
23395 (vectorizable_store): Update call to get_load_store_type.
23396 Handle IFN_MASK_STORE_LANES.
23397 (vectorizable_load): Update call to get_load_store_type.
23398 Handle IFN_MASK_LOAD_LANES.
23400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23401 Alan Hayward <alan.hayward@arm.com>
23402 David Sherwood <david.sherwood@arm.com>
23404 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
23406 * config/aarch64/aarch64-protos.h
23407 (aarch64_sve_struct_memory_operand_p): Declare.
23408 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
23409 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
23410 (VPRED, vpred): Handle SVE structure modes.
23411 * config/aarch64/constraints.md (Utx): New constraint.
23412 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
23413 (aarch64_sve_struct_nonimmediate_operand): New predicates.
23414 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
23415 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
23416 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
23417 structure modes. Split into pieces after RA.
23418 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
23419 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
23421 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
23422 SVE structure modes.
23423 (aarch64_classify_address): Likewise.
23424 (sizetochar): Move earlier in file.
23425 (aarch64_print_operand): Handle SVE register lists.
23426 (aarch64_array_mode): New function.
23427 (aarch64_sve_struct_memory_operand_p): Likewise.
23428 (TARGET_ARRAY_MODE): Redefine.
23430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23431 Alan Hayward <alan.hayward@arm.com>
23432 David Sherwood <david.sherwood@arm.com>
23434 * target.def (array_mode): New target hook.
23435 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
23436 * doc/tm.texi: Regenerate.
23437 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
23438 * hooks.c (hook_optmode_mode_uhwi_none): New function.
23439 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
23440 targetm.array_mode.
23441 * stor-layout.c (mode_for_array): Likewise. Support polynomial
23444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23445 Alan Hayward <alan.hayward@arm.com>
23446 David Sherwood <david.sherwood@arm.com>
23448 * fold-const.c (fold_binary_loc): Check the argument types
23449 rather than the result type when testing for a vector operation.
23451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23453 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
23454 * doc/tm.texi: Regenerate.
23456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23457 Alan Hayward <alan.hayward@arm.com>
23458 David Sherwood <david.sherwood@arm.com>
23460 * doc/invoke.texi (-msve-vector-bits=): Document new option.
23461 (sve): Document new AArch64 extension.
23462 * doc/md.texi (w): Extend the description of the AArch64
23463 constraint to include SVE vectors.
23464 (Upl, Upa): Document new AArch64 predicate constraints.
23465 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
23467 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
23468 (msve-vector-bits=): New option.
23469 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
23470 SVE when these are disabled.
23471 (sve): New extension.
23472 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
23473 modes. Adjust their number of units based on aarch64_sve_vg.
23474 (MAX_BITSIZE_MODE_ANY_MODE): Define.
23475 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
23476 aarch64_addr_query_type.
23477 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
23478 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
23479 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
23480 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
23481 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
23482 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
23483 (aarch64_simd_imm_zero_p): Delete.
23484 (aarch64_check_zero_based_sve_index_immediate): Declare.
23485 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23486 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23487 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23488 (aarch64_sve_float_mul_immediate_p): Likewise.
23489 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23490 rather than an rtx.
23491 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
23492 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
23493 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
23494 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
23495 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
23496 (aarch64_regmode_natural_size): Likewise.
23497 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
23498 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
23500 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
23501 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
23502 for VG and the SVE predicate registers.
23503 (V_ALIASES): Add a "z"-prefixed alias.
23504 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
23505 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
23506 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
23507 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
23508 (REG_CLASS_NAMES): Add entries for them.
23509 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
23510 and the predicate registers.
23511 (aarch64_sve_vg): Declare.
23512 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
23513 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
23514 (REGMODE_NATURAL_SIZE): Define.
23515 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
23517 * config/aarch64/aarch64.c: Include cfgrtl.h.
23518 (simd_immediate_info): Add a constructor for series vectors,
23519 and an associated step field.
23520 (aarch64_sve_vg): New variable.
23521 (aarch64_dbx_register_number): Handle VG and the predicate registers.
23522 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
23523 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
23524 (VEC_ANY_DATA, VEC_STRUCT): New constants.
23525 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
23526 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
23527 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
23528 (aarch64_get_mask_mode): New functions.
23529 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
23530 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
23531 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
23532 predicate modes and predicate registers. Explicitly restrict
23533 GPRs to modes of 16 bytes or smaller. Only allow FP registers
23534 to store a vector mode if it is recognized by
23535 aarch64_classify_vector_mode.
23536 (aarch64_regmode_natural_size): New function.
23537 (aarch64_hard_regno_caller_save_mode): Return the original mode
23539 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
23540 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
23541 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
23542 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
23544 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
23545 does not overlap dest if the function is frame-related. Handle
23547 (aarch64_split_add_offset): New function.
23548 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
23549 them aarch64_add_offset.
23550 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
23551 and update call to aarch64_sub_sp.
23552 (aarch64_add_cfa_expression): New function.
23553 (aarch64_expand_prologue): Pass extra temporary registers to the
23554 functions above. Handle the case in which we need to emit new
23555 DW_CFA_expressions for registers that were originally saved
23556 relative to the stack pointer, but now have to be expressed
23557 relative to the frame pointer.
23558 (aarch64_output_mi_thunk): Pass extra temporary registers to the
23560 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
23561 IP0 and IP1 values for SVE frames.
23562 (aarch64_expand_vec_series): New function.
23563 (aarch64_expand_sve_widened_duplicate): Likewise.
23564 (aarch64_expand_sve_const_vector): Likewise.
23565 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
23566 Handle SVE constants. Use emit_move_insn to move a force_const_mem
23567 into the register, rather than emitting a SET directly.
23568 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
23569 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
23570 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
23571 (offset_9bit_signed_scaled_p): New functions.
23572 (aarch64_replicate_bitmask_imm): New function.
23573 (aarch64_bitmask_imm): Use it.
23574 (aarch64_cannot_force_const_mem): Reject expressions involving
23575 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
23576 (aarch64_classify_index): Handle SVE indices, by requiring
23577 a plain register index with a scale that matches the element size.
23578 (aarch64_classify_address): Handle SVE addresses. Assert that
23579 the mode of the address is VOIDmode or an integer mode.
23580 Update call to aarch64_classify_symbol.
23581 (aarch64_classify_symbolic_expression): Update call to
23582 aarch64_classify_symbol.
23583 (aarch64_const_vec_all_in_range_p): New function.
23584 (aarch64_print_vector_float_operand): Likewise.
23585 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
23586 "vN" for FP registers with SVE modes. Handle (const ...) vectors
23587 and the FP immediates 1.0 and 0.5.
23588 (aarch64_print_address_internal): Handle SVE addresses.
23589 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
23590 (aarch64_regno_regclass): Handle predicate registers.
23591 (aarch64_secondary_reload): Handle big-endian reloads of SVE
23593 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
23594 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
23595 (aarch64_convert_sve_vector_bits): New function.
23596 (aarch64_override_options): Use it to handle -msve-vector-bits=.
23597 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23598 rather than an rtx.
23599 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
23600 Handle SVE vector and predicate modes. Accept VL-based constants
23601 that need only one temporary register, and VL offsets that require
23602 no temporary registers.
23603 (aarch64_conditional_register_usage): Mark the predicate registers
23604 as fixed if SVE isn't available.
23605 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
23606 Return true for SVE vector and predicate modes.
23607 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
23608 rather than an unsigned int. Handle SVE modes.
23609 (aarch64_preferred_simd_mode): Update call accordingly. Handle
23611 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
23613 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23614 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23615 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23616 (aarch64_sve_float_mul_immediate_p): New functions.
23617 (aarch64_sve_valid_immediate): New function.
23618 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
23619 Explicitly reject structure modes. Check for INDEX constants.
23620 Handle PTRUE and PFALSE constants.
23621 (aarch64_check_zero_based_sve_index_immediate): New function.
23622 (aarch64_simd_imm_zero_p): Delete.
23623 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
23624 vector modes. Accept constants in the range of CNT[BHWD].
23625 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
23626 ask for an Advanced SIMD mode.
23627 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
23628 (aarch64_simd_vector_alignment): Handle SVE predicates.
23629 (aarch64_vectorize_preferred_vector_alignment): New function.
23630 (aarch64_simd_vector_alignment_reachable): Use it instead of
23632 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
23633 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
23635 (MAX_VECT_LEN): Delete.
23636 (expand_vec_perm_d): Add a vec_flags field.
23637 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
23638 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
23639 (aarch64_evpc_ext): Don't apply a big-endian lane correction
23641 (aarch64_evpc_rev): Rename to...
23642 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
23643 (aarch64_evpc_rev_global): New function.
23644 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
23645 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
23647 (aarch64_evpc_sve_tbl): New function.
23648 (aarch64_expand_vec_perm_const_1): Update after rename of
23649 aarch64_evpc_rev. Handle SVE permutes too, trying
23650 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
23651 than aarch64_evpc_tbl.
23652 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
23653 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
23654 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
23655 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
23656 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
23657 (aarch64_expand_sve_vcond): New functions.
23658 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
23659 of aarch64_vector_mode_p.
23660 (aarch64_dwarf_poly_indeterminate_value): New function.
23661 (aarch64_compute_pressure_classes): Likewise.
23662 (aarch64_can_change_mode_class): Likewise.
23663 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
23664 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
23665 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
23666 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
23667 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
23668 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
23669 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
23670 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
23672 (Dn, Dl, Dr): Accept const as well as const_vector.
23673 (Dz): Likewise. Compare against CONST0_RTX.
23674 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
23675 of "vector" where appropriate.
23676 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
23677 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
23678 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
23679 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
23680 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
23681 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
23682 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
23683 (v_int_equiv): Extend to SVE modes.
23684 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
23686 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
23687 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
23688 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
23689 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
23690 (SVE_COND_FP_CMP): New int iterators.
23691 (perm_hilo): Handle the new unpack unspecs.
23692 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
23694 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
23695 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
23696 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
23697 (aarch64_equality_operator, aarch64_constant_vector_operand)
23698 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
23699 (aarch64_sve_nonimmediate_operand): Likewise.
23700 (aarch64_sve_general_operand): Likewise.
23701 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
23702 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
23703 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
23704 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
23705 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
23706 (aarch64_sve_float_arith_immediate): Likewise.
23707 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
23708 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
23709 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
23710 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
23711 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
23712 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
23713 (aarch64_sve_float_arith_operand): Likewise.
23714 (aarch64_sve_float_arith_with_sub_operand): Likewise.
23715 (aarch64_sve_float_mul_operand): Likewise.
23716 (aarch64_sve_vec_perm_operand): Likewise.
23717 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
23718 (aarch64_mov_operand): Accept const_poly_int and const_vector.
23719 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
23720 as well as const_vector.
23721 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
23722 in file. Use CONST0_RTX and CONSTM1_RTX.
23723 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
23724 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
23725 Use aarch64_simd_imm_zero.
23726 * config/aarch64/aarch64-sve.md: New file.
23727 * config/aarch64/aarch64.md: Include it.
23728 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
23729 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
23730 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
23731 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
23732 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
23733 (sve): New attribute.
23734 (enabled): Disable instructions with the sve attribute unless
23736 (movqi, movhi): Pass CONST_POLY_INT operaneds through
23737 aarch64_expand_mov_immediate.
23738 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
23739 CNT[BHSD] immediates.
23740 (movti): Split CONST_POLY_INT moves into two halves.
23741 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
23742 Split additions that need a temporary here if the destination
23743 is the stack pointer.
23744 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
23745 (*add<mode>3_poly_1): New instruction.
23746 (set_clobber_cc): New expander.
23748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23750 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
23751 parameter and use it instead of GET_MODE_SIZE (innermode). Use
23752 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
23753 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
23754 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
23755 Change innermode from fixed_mode_size to machine_mode.
23756 (simplify_subreg): Update call accordingly. Handle a constant-sized
23757 subreg of a variable-length CONST_VECTOR.
23759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23760 Alan Hayward <alan.hayward@arm.com>
23761 David Sherwood <david.sherwood@arm.com>
23763 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23764 (add_offset_to_base): New function, split out from...
23765 (create_mem_ref): ...here. When handling a scale other than 1,
23766 check first whether the address is valid without the offset.
23767 Add it into the base if so, leaving the index and scale as-is.
23769 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23772 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23773 fold_for_warn before checking if arg2 is INTEGER_CST.
23775 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
23777 * config/rs6000/predicates.md (load_multiple_operation): Delete.
23778 (store_multiple_operation): Delete.
23779 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23780 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23781 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23782 guarded by TARGET_STRING.
23783 (rs6000_output_load_multiple): Delete.
23784 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23785 OPTION_MASK_STRING / TARGET_STRING handling.
23786 (print_operand) <'N', 'O'>: Add comment that these are unused now.
23787 (const rs6000_opt_masks) <"string">: Change mask to 0.
23788 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23789 (MASK_STRING): Delete.
23790 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23792 (load_multiple): Delete.
23799 (store_multiple): Delete.
23806 (movmemsi_8reg): Delete.
23807 (corresponding unnamed define_insn): Delete.
23808 (movmemsi_6reg): Delete.
23809 (corresponding unnamed define_insn): Delete.
23810 (movmemsi_4reg): Delete.
23811 (corresponding unnamed define_insn): Delete.
23812 (movmemsi_2reg): Delete.
23813 (corresponding unnamed define_insn): Delete.
23814 (movmemsi_1reg): Delete.
23815 (corresponding unnamed define_insn): Delete.
23816 * config/rs6000/rs6000.opt (mno-string): New.
23817 (mstring): Replace by deprecation warning stub.
23818 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
23820 2018-01-12 Jakub Jelinek <jakub@redhat.com>
23822 * regrename.c (regrename_do_replace): If replacing the same
23823 reg multiple times, try to reuse last created gen_raw_REG.
23826 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
23827 main to workaround a bug in GDB.
23829 2018-01-12 Tom de Vries <tom@codesourcery.com>
23832 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
23834 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
23836 PR rtl-optimization/80481
23837 * ira-color.c (get_cap_member): New function.
23838 (allocnos_conflict_by_live_ranges_p): Use it.
23839 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
23840 (setup_slot_coalesced_allocno_live_ranges): Ditto.
23842 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
23845 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
23846 (*saddl_se_1): Ditto.
23847 (*ssubsi_1): Ditto.
23848 (*ssubl_se_1): Ditto.
23850 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23852 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
23853 rather than wi::to_widest for DR_INITs.
23854 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
23855 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
23856 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
23858 (vect_analyze_group_access_1): Note that here.
23860 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23862 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
23863 polynomial type sizes.
23865 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
23867 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
23868 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
23869 (gimple_add_tmp_var): Likewise.
23871 2018-01-12 Martin Liska <mliska@suse.cz>
23873 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
23874 (gimple_alloc_sizes): Likewise.
23875 (dump_gimple_statistics): Use PRIu64 in printf format.
23876 * gimple.h: Change uint64_t to int.
23878 2018-01-12 Martin Liska <mliska@suse.cz>
23880 * tree-core.h: Use uint64_t instead of int.
23881 * tree.c (tree_node_counts): Likewise.
23882 (tree_node_sizes): Likewise.
23883 (dump_tree_statistics): Use PRIu64 in printf format.
23885 2018-01-12 Martin Liska <mliska@suse.cz>
23887 * Makefile.in: As qsort_chk is implemented in vec.c, add
23888 vec.o to linkage of gencfn-macros.
23889 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
23890 passing the info to record_node_allocation_statistics.
23891 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
23893 * ggc-common.c (struct ggc_usage): Add operator== and use
23894 it in operator< and compare function.
23895 * mem-stats.h (struct mem_usage): Likewise.
23896 * vec.c (struct vec_usage): Remove operator< and compare
23897 function. Can be simply inherited.
23899 2018-01-12 Martin Jambor <mjambor@suse.cz>
23902 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
23903 * tree-ssa-math-opts.c: Include domwalk.h.
23904 (convert_mult_to_fma_1): New function.
23905 (fma_transformation_info): New type.
23906 (fma_deferring_state): Likewise.
23907 (cancel_fma_deferring): New function.
23908 (result_of_phi): Likewise.
23909 (last_fma_candidate_feeds_initial_phi): Likewise.
23910 (convert_mult_to_fma): Added deferring logic, split actual
23911 transformation to convert_mult_to_fma_1.
23912 (math_opts_dom_walker): New type.
23913 (math_opts_dom_walker::after_dom_children): New method, body moved
23914 here from pass_optimize_widening_mul::execute, added deferring logic
23916 (pass_optimize_widening_mul::execute): Moved most of code to
23917 math_opts_dom_walker::after_dom_children.
23918 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
23919 * config/i386/i386.c (ix86_option_override_internal): Added
23920 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
23922 2018-01-12 Richard Biener <rguenther@suse.de>
23925 * dwarf2out.c (gen_variable_die): Do not reset old_die for
23926 inline instance vars.
23928 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
23931 * config/rx/rx.c (rx_is_restricted_memory_address):
23932 Handle SUBREG case.
23934 2018-01-12 Richard Biener <rguenther@suse.de>
23936 PR tree-optimization/80846
23937 * target.def (split_reduction): New target hook.
23938 * targhooks.c (default_split_reduction): New function.
23939 * targhooks.h (default_split_reduction): Declare.
23940 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
23941 target requests first reduce vectors by combining low and high
23943 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
23944 (get_vectype_for_scalar_type_and_size): Export.
23945 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
23946 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
23947 * doc/tm.texi: Regenerate.
23948 * config/i386/i386.c (ix86_split_reduction): Implement
23949 TARGET_VECTORIZE_SPLIT_REDUCTION.
23951 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23954 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
23955 in PIC mode except for TARGET_VXWORKS_RTP.
23956 * config/sparc/sparc.c: Include cfgrtl.h.
23957 (TARGET_INIT_PIC_REG): Define.
23958 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
23959 (sparc_pic_register_p): New predicate.
23960 (sparc_legitimate_address_p): Use it.
23961 (sparc_legitimize_pic_address): Likewise.
23962 (sparc_delegitimize_address): Likewise.
23963 (sparc_mode_dependent_address_p): Likewise.
23964 (gen_load_pcrel_sym): Remove 4th parameter.
23965 (load_got_register): Adjust call to above. Remove obsolete stuff.
23966 (sparc_expand_prologue): Do not call load_got_register here.
23967 (sparc_flat_expand_prologue): Likewise.
23968 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
23969 (sparc_use_pseudo_pic_reg): New function.
23970 (sparc_init_pic_reg): Likewise.
23971 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23972 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
23974 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23976 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
23977 Add item for branch_cost.
23979 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
23981 PR rtl-optimization/83565
23982 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
23983 not extend the result to a larger mode for rotate operations.
23984 (num_sign_bit_copies1): Likewise.
23986 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23989 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
23991 Use values-Xc.o for -pedantic.
23992 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
23994 2018-01-12 Martin Liska <mliska@suse.cz>
23997 * ipa-devirt.c (final_warning_record::grow_type_warnings):
23999 (possible_polymorphic_call_targets): Use it.
24000 (ipa_devirt): Likewise.
24002 2018-01-12 Martin Liska <mliska@suse.cz>
24004 * profile-count.h (enum profile_quality): Use 0 as invalid
24005 enum value of profile_quality.
24007 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
24009 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
24010 -mext-string options.
24012 2018-01-12 Richard Biener <rguenther@suse.de>
24014 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
24015 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
24016 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
24018 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
24020 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
24022 * configure.ac (--with-long-double-format): Add support for the
24023 configuration option to change the default long double format on
24025 * config.gcc (powerpc*-linux*-*): Likewise.
24026 * configure: Regenerate.
24027 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
24028 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
24029 used without modification.
24031 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24033 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
24034 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
24035 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
24036 MISC_BUILTIN_SPEC_BARRIER.
24037 (rs6000_init_builtins): Likewise.
24038 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
24040 (speculation_barrier): New define_insn.
24041 * doc/extend.texi: Document __builtin_speculation_barrier.
24043 2018-01-11 Jakub Jelinek <jakub@redhat.com>
24046 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
24047 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
24048 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
24050 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
24051 integral modes instead of "ss" and "sd".
24052 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
24053 vectors with 32-bit and 64-bit elements.
24054 (vecdupssescalarmodesuffix): New mode attribute.
24055 (vec_dup<mode>): Use it.
24057 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
24060 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
24061 frame if argument is passed on stack.
24063 2018-01-11 Jakub Jelinek <jakub@redhat.com>
24066 * ree.c (combine_reaching_defs): Optimize also
24067 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
24068 reg2=any_extend(exp); reg1=reg2;, formatting fix.
24070 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
24072 PR middle-end/83189
24073 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
24075 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
24077 PR middle-end/83718
24078 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
24079 after they are computed.
24081 2018-01-11 Bin Cheng <bin.cheng@arm.com>
24083 PR tree-optimization/83695
24084 * gimple-loop-linterchange.cc
24085 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
24086 reset cached scev information after interchange.
24087 (pass_linterchange::execute): Remove call to scev_reset_htab.
24089 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24091 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
24092 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
24093 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
24094 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
24095 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
24096 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
24097 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
24098 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
24099 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
24100 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
24101 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
24102 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
24103 (V_lane_reg): Likewise.
24104 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
24106 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
24107 (vfmal_lane_low<mode>_intrinsic,
24108 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
24109 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
24110 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
24111 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
24112 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
24113 vfmsl_lane_high<mode>_intrinsic): New define_insns.
24115 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24117 * config/arm/arm-cpus.in (fp16fml): New feature.
24118 (ALL_SIMD): Add fp16fml.
24119 (armv8.2-a): Add fp16fml as an option.
24120 (armv8.3-a): Likewise.
24121 (armv8.4-a): Add fp16fml as part of fp16.
24122 * config/arm/arm.h (TARGET_FP16FML): Define.
24123 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
24125 * config/arm/arm-modes.def (V2HF): Define.
24126 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
24127 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
24128 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
24129 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
24130 vfmsl_low, vfmsl_high): New set of builtins.
24131 * config/arm/iterators.md (PLUSMINUS): New code iterator.
24132 (vfml_op): New code attribute.
24133 (VFMLHALVES): New int iterator.
24134 (VFML, VFMLSEL): New mode attributes.
24135 (V_reg): Define mapping for V2HF.
24136 (V_hi, V_lo): New mode attributes.
24137 (VF_constraint): Likewise.
24138 (vfml_half, vfml_half_selector): New int attributes.
24139 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
24141 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
24142 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
24144 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
24145 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
24146 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
24147 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
24149 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
24150 Document new effective target and option set.
24152 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24154 * config/arm/arm-cpus.in (armv8_4): New feature.
24155 (ARMv8_4a): New fgroup.
24156 (armv8.4-a): New arch.
24157 * config/arm/arm-tables.opt: Regenerate.
24158 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
24159 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
24160 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
24161 Add matching rules for -march=armv8.4-a and extensions.
24162 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
24164 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
24167 * config/rx/rx.md (BW): New mode attribute.
24168 (sync_lock_test_and_setsi): Add mode suffix to insn output.
24170 2018-01-11 Richard Biener <rguenther@suse.de>
24172 PR tree-optimization/83435
24173 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
24174 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
24175 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
24177 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24178 Alan Hayward <alan.hayward@arm.com>
24179 David Sherwood <david.sherwood@arm.com>
24181 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
24183 (aarch64_classify_address): Initialize it. Track polynomial offsets.
24184 (aarch64_print_address_internal): Use it to check for a zero offset.
24186 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24187 Alan Hayward <alan.hayward@arm.com>
24188 David Sherwood <david.sherwood@arm.com>
24190 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
24191 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
24192 Return a poly_int64 rather than a HOST_WIDE_INT.
24193 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
24194 rather than a HOST_WIDE_INT.
24195 * config/aarch64/aarch64.h (aarch64_frame): Protect with
24196 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
24197 hard_fp_offset, frame_size, initial_adjust, callee_offset and
24198 final_offset from HOST_WIDE_INT to poly_int64.
24199 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
24200 to_constant when getting the number of units in an Advanced SIMD
24202 (aarch64_builtin_vectorized_function): Check for a constant number
24204 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
24206 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
24207 attribute instead of GET_MODE_NUNITS.
24208 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
24209 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
24210 GET_MODE_SIZE for fixed-size registers.
24211 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
24212 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
24213 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
24214 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
24215 (aarch64_print_operand, aarch64_print_address_internal)
24216 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
24217 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
24218 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
24219 Handle polynomial GET_MODE_SIZE.
24220 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
24221 wider than SImode without modification.
24222 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
24223 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
24224 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
24225 passing and returning SVE modes.
24226 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
24227 rather than GEN_INT.
24228 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
24229 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
24230 (aarch64_allocate_and_probe_stack_space): Likewise.
24231 (aarch64_layout_frame): Cope with polynomial offsets.
24232 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
24233 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
24234 polynomial offsets.
24235 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
24236 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
24237 poly_int64 rather than a HOST_WIDE_INT.
24238 (aarch64_get_separate_components, aarch64_process_components)
24239 (aarch64_expand_prologue, aarch64_expand_epilogue)
24240 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
24241 (aarch64_anchor_offset): New function, split out from...
24242 (aarch64_legitimize_address): ...here.
24243 (aarch64_builtin_vectorization_cost): Handle polynomial
24244 TYPE_VECTOR_SUBPARTS.
24245 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
24247 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
24248 number of elements from the PARALLEL rather than the mode.
24249 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
24250 rather than GET_MODE_BITSIZE.
24251 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
24252 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
24253 (aarch64_expand_vec_perm_const_1): Handle polynomial
24254 d->perm.length () and d->perm elements.
24255 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
24256 Apply to_constant to d->perm elements.
24257 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
24258 polynomial CONST_VECTOR_NUNITS.
24259 (aarch64_move_pointer): Take amount as a poly_int64 rather
24261 (aarch64_progress_pointer): Avoid temporary variable.
24262 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
24263 the mode attribute instead of GET_MODE.
24265 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24266 Alan Hayward <alan.hayward@arm.com>
24267 David Sherwood <david.sherwood@arm.com>
24269 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
24270 x exists before using it.
24271 (aarch64_add_constant_internal): Rename to...
24272 (aarch64_add_offset_1): ...this. Replace regnum with separate
24273 src and dest rtxes. Handle the case in which they're different,
24274 including when the offset is zero. Replace scratchreg with an rtx.
24275 Use 2 additions if there is no spare register into which we can
24276 move a 16-bit constant.
24277 (aarch64_add_constant): Delete.
24278 (aarch64_add_offset): Replace reg with separate src and dest
24279 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
24280 Use aarch64_add_offset_1.
24281 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
24282 an rtx rather than an int. Take the delta as a poly_int64
24283 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
24284 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
24285 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
24286 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
24287 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
24288 and aarch64_add_sp.
24289 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
24290 aarch64_add_constant.
24292 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24294 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
24295 Use scalar_float_mode.
24297 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24299 * config/aarch64/aarch64-simd.md
24300 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
24301 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
24302 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
24303 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
24304 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
24305 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
24306 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
24307 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
24308 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
24309 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
24311 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24314 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
24315 targ_options->x_arm_arch_string is non NULL.
24317 2018-01-11 Tamar Christina <tamar.christina@arm.com>
24319 * config/aarch64/aarch64.h
24320 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
24322 2018-01-11 Sudakshina Das <sudi.das@arm.com>
24325 * expmed.c (emit_store_flag_force): Swap if const op0
24326 and change VOIDmode to mode of op0.
24328 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24330 PR rtl-optimization/83761
24331 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
24332 than bytes to mode_for_size.
24334 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24336 PR middle-end/83189
24337 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
24338 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
24341 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24343 PR middle-end/83575
24344 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
24345 when in layout mode.
24346 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
24347 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
24350 2018-01-10 Michael Collison <michael.collison@arm.com>
24352 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
24353 * config/aarch64/aarch64-option-extension.def: Add
24354 AARCH64_OPT_EXTENSION of 'fp16fml'.
24355 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24356 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
24357 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
24358 * config/aarch64/constraints.md (Ui7): New constraint.
24359 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
24360 (VFMLA_SEL_W): Ditto.
24363 (VFMLA16_LOW): New int iterator.
24364 (VFMLA16_HIGH): Ditto.
24365 (UNSPEC_FMLAL): New unspec.
24366 (UNSPEC_FMLSL): Ditto.
24367 (UNSPEC_FMLAL2): Ditto.
24368 (UNSPEC_FMLSL2): Ditto.
24369 (f16mac): New code attribute.
24370 * config/aarch64/aarch64-simd-builtins.def
24371 (aarch64_fmlal_lowv2sf): Ditto.
24372 (aarch64_fmlsl_lowv2sf): Ditto.
24373 (aarch64_fmlalq_lowv4sf): Ditto.
24374 (aarch64_fmlslq_lowv4sf): Ditto.
24375 (aarch64_fmlal_highv2sf): Ditto.
24376 (aarch64_fmlsl_highv2sf): Ditto.
24377 (aarch64_fmlalq_highv4sf): Ditto.
24378 (aarch64_fmlslq_highv4sf): Ditto.
24379 (aarch64_fmlal_lane_lowv2sf): Ditto.
24380 (aarch64_fmlsl_lane_lowv2sf): Ditto.
24381 (aarch64_fmlal_laneq_lowv2sf): Ditto.
24382 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
24383 (aarch64_fmlalq_lane_lowv4sf): Ditto.
24384 (aarch64_fmlsl_lane_lowv4sf): Ditto.
24385 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
24386 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
24387 (aarch64_fmlal_lane_highv2sf): Ditto.
24388 (aarch64_fmlsl_lane_highv2sf): Ditto.
24389 (aarch64_fmlal_laneq_highv2sf): Ditto.
24390 (aarch64_fmlsl_laneq_highv2sf): Ditto.
24391 (aarch64_fmlalq_lane_highv4sf): Ditto.
24392 (aarch64_fmlsl_lane_highv4sf): Ditto.
24393 (aarch64_fmlalq_laneq_highv4sf): Ditto.
24394 (aarch64_fmlsl_laneq_highv4sf): Ditto.
24395 * config/aarch64/aarch64-simd.md:
24396 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
24397 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24398 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
24399 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24400 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
24401 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
24402 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
24403 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
24404 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
24405 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
24406 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
24407 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
24408 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
24409 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
24410 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
24411 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
24412 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
24413 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
24414 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
24415 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
24416 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
24417 (vfmlsl_low_u32): Ditto.
24418 (vfmlalq_low_u32): Ditto.
24419 (vfmlslq_low_u32): Ditto.
24420 (vfmlal_high_u32): Ditto.
24421 (vfmlsl_high_u32): Ditto.
24422 (vfmlalq_high_u32): Ditto.
24423 (vfmlslq_high_u32): Ditto.
24424 (vfmlal_lane_low_u32): Ditto.
24425 (vfmlsl_lane_low_u32): Ditto.
24426 (vfmlal_laneq_low_u32): Ditto.
24427 (vfmlsl_laneq_low_u32): Ditto.
24428 (vfmlalq_lane_low_u32): Ditto.
24429 (vfmlslq_lane_low_u32): Ditto.
24430 (vfmlalq_laneq_low_u32): Ditto.
24431 (vfmlslq_laneq_low_u32): Ditto.
24432 (vfmlal_lane_high_u32): Ditto.
24433 (vfmlsl_lane_high_u32): Ditto.
24434 (vfmlal_laneq_high_u32): Ditto.
24435 (vfmlsl_laneq_high_u32): Ditto.
24436 (vfmlalq_lane_high_u32): Ditto.
24437 (vfmlslq_lane_high_u32): Ditto.
24438 (vfmlalq_laneq_high_u32): Ditto.
24439 (vfmlslq_laneq_high_u32): Ditto.
24440 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
24441 (AARCH64_FL_FOR_ARCH8_4): New.
24442 (AARCH64_ISA_F16FML): New ISA flag.
24443 (TARGET_F16FML): New feature flag for fp16fml.
24444 (doc/invoke.texi): Document new fp16fml option.
24446 2018-01-10 Michael Collison <michael.collison@arm.com>
24448 * config/aarch64/aarch64-builtins.c:
24449 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
24450 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24451 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
24452 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
24453 (AARCH64_ISA_SHA3): New ISA flag.
24454 (TARGET_SHA3): New feature flag for sha3.
24455 * config/aarch64/iterators.md (sha512_op): New int attribute.
24456 (CRYPTO_SHA512): New int iterator.
24457 (UNSPEC_SHA512H): New unspec.
24458 (UNSPEC_SHA512H2): Ditto.
24459 (UNSPEC_SHA512SU0): Ditto.
24460 (UNSPEC_SHA512SU1): Ditto.
24461 * config/aarch64/aarch64-simd-builtins.def
24462 (aarch64_crypto_sha512hqv2di): New builtin.
24463 (aarch64_crypto_sha512h2qv2di): Ditto.
24464 (aarch64_crypto_sha512su0qv2di): Ditto.
24465 (aarch64_crypto_sha512su1qv2di): Ditto.
24466 (aarch64_eor3qv8hi): Ditto.
24467 (aarch64_rax1qv2di): Ditto.
24468 (aarch64_xarqv2di): Ditto.
24469 (aarch64_bcaxqv8hi): Ditto.
24470 * config/aarch64/aarch64-simd.md:
24471 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
24472 (aarch64_crypto_sha512su0qv2di): Ditto.
24473 (aarch64_crypto_sha512su1qv2di): Ditto.
24474 (aarch64_eor3qv8hi): Ditto.
24475 (aarch64_rax1qv2di): Ditto.
24476 (aarch64_xarqv2di): Ditto.
24477 (aarch64_bcaxqv8hi): Ditto.
24478 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
24479 (vsha512h2q_u64): Ditto.
24480 (vsha512su0q_u64): Ditto.
24481 (vsha512su1q_u64): Ditto.
24482 (veor3q_u16): Ditto.
24483 (vrax1q_u64): Ditto.
24484 (vxarq_u64): Ditto.
24485 (vbcaxq_u16): Ditto.
24486 * config/arm/types.md (crypto_sha512): New type attribute.
24487 (crypto_sha3): Ditto.
24488 (doc/invoke.texi): Document new sha3 option.
24490 2018-01-10 Michael Collison <michael.collison@arm.com>
24492 * config/aarch64/aarch64-builtins.c:
24493 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
24494 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24495 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
24496 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
24497 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
24498 (AARCH64_ISA_SM4): New ISA flag.
24499 (TARGET_SM4): New feature flag for sm4.
24500 * config/aarch64/aarch64-simd-builtins.def
24501 (aarch64_sm3ss1qv4si): Ditto.
24502 (aarch64_sm3tt1aq4si): Ditto.
24503 (aarch64_sm3tt1bq4si): Ditto.
24504 (aarch64_sm3tt2aq4si): Ditto.
24505 (aarch64_sm3tt2bq4si): Ditto.
24506 (aarch64_sm3partw1qv4si): Ditto.
24507 (aarch64_sm3partw2qv4si): Ditto.
24508 (aarch64_sm4eqv4si): Ditto.
24509 (aarch64_sm4ekeyqv4si): Ditto.
24510 * config/aarch64/aarch64-simd.md:
24511 (aarch64_sm3ss1qv4si): Ditto.
24512 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
24513 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
24514 (aarch64_sm4eqv4si): Ditto.
24515 (aarch64_sm4ekeyqv4si): Ditto.
24516 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
24517 (sm3part_op): Ditto.
24518 (CRYPTO_SM3TT): Ditto.
24519 (CRYPTO_SM3PART): Ditto.
24520 (UNSPEC_SM3SS1): New unspec.
24521 (UNSPEC_SM3TT1A): Ditto.
24522 (UNSPEC_SM3TT1B): Ditto.
24523 (UNSPEC_SM3TT2A): Ditto.
24524 (UNSPEC_SM3TT2B): Ditto.
24525 (UNSPEC_SM3PARTW1): Ditto.
24526 (UNSPEC_SM3PARTW2): Ditto.
24527 (UNSPEC_SM4E): Ditto.
24528 (UNSPEC_SM4EKEY): Ditto.
24529 * config/aarch64/constraints.md (Ui2): New constraint.
24530 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
24531 * config/arm/types.md (crypto_sm3): New type attribute.
24532 (crypto_sm4): Ditto.
24533 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
24534 (vsm3tt1aq_u32): Ditto.
24535 (vsm3tt1bq_u32): Ditto.
24536 (vsm3tt2aq_u32): Ditto.
24537 (vsm3tt2bq_u32): Ditto.
24538 (vsm3partw1q_u32): Ditto.
24539 (vsm3partw2q_u32): Ditto.
24540 (vsm4eq_u32): Ditto.
24541 (vsm4ekeyq_u32): Ditto.
24542 (doc/invoke.texi): Document new sm4 option.
24544 2018-01-10 Michael Collison <michael.collison@arm.com>
24546 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
24547 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
24548 (AARCH64_FL_FOR_ARCH8_4): New.
24549 (AARCH64_FL_V8_4): New flag.
24550 (doc/invoke.texi): Document new armv8.4-a option.
24552 2018-01-10 Michael Collison <michael.collison@arm.com>
24554 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24555 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
24556 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
24557 * config/aarch64/aarch64-option-extension.def: Add
24558 AARCH64_OPT_EXTENSION of 'sha2'.
24559 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
24560 (crypto): Disable sha2 and aes if crypto disabled.
24561 (crypto): Enable aes and sha2 if enabled.
24562 (simd): Disable sha2 and aes if simd disabled.
24563 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
24565 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
24566 (TARGET_SHA2): New feature flag for sha2.
24567 (TARGET_AES): New feature flag for aes.
24568 * config/aarch64/aarch64-simd.md:
24569 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
24570 conditional on TARGET_AES.
24571 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
24572 (aarch64_crypto_sha1hsi): Make pattern conditional
24574 (aarch64_crypto_sha1hv4si): Ditto.
24575 (aarch64_be_crypto_sha1hv4si): Ditto.
24576 (aarch64_crypto_sha1su1v4si): Ditto.
24577 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
24578 (aarch64_crypto_sha1su0v4si): Ditto.
24579 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
24580 (aarch64_crypto_sha256su0v4si): Ditto.
24581 (aarch64_crypto_sha256su1v4si): Ditto.
24582 (doc/invoke.texi): Document new aes and sha2 options.
24584 2018-01-10 Martin Sebor <msebor@redhat.com>
24586 PR tree-optimization/83781
24587 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
24590 2018-01-11 Martin Sebor <msebor@gmail.com>
24591 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24593 PR tree-optimization/83501
24594 PR tree-optimization/81703
24596 * tree-ssa-strlen.c (get_string_cst): Rename...
24597 (get_string_len): ...to this. Handle global constants.
24598 (handle_char_store): Adjust.
24600 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
24601 Jim Wilson <jimw@sifive.com>
24603 * config/riscv/riscv-protos.h (riscv_output_return): New.
24604 * config/riscv/riscv.c (struct machine_function): New naked_p field.
24605 (riscv_attribute_table, riscv_output_return),
24606 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
24607 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
24608 (riscv_compute_frame_info): Only compute frame->mask if not a naked
24610 (riscv_expand_prologue): Add early return for naked function.
24611 (riscv_expand_epilogue): Likewise.
24612 (riscv_function_ok_for_sibcall): Return false for naked function.
24613 (riscv_set_current_function): New.
24614 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
24615 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
24616 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
24617 * doc/extend.texi (RISC-V Function Attributes): New.
24619 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
24621 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
24622 check for 128-bit long double before checking TCmode.
24623 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
24624 128-bit long doubles before checking TFmode or TCmode.
24625 (FLOAT128_IBM_P): Likewise.
24627 2018-01-10 Martin Sebor <msebor@redhat.com>
24629 PR tree-optimization/83671
24630 * builtins.c (c_strlen): Unconditionally return zero for the empty
24632 Use -Warray-bounds for warnings.
24633 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
24634 for non-constant array indices with COMPONENT_REF, arrays of
24635 arrays, and pointers to arrays.
24636 (gimple_fold_builtin_strlen): Determine and set length range for
24637 non-constant character arrays.
24639 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
24641 PR middle-end/81897
24642 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
24645 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
24647 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
24649 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
24652 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
24653 VECTOR_MEM_ALTIVEC_OR_VSX_P.
24654 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
24655 indexed_or_indirect_operand predicate.
24656 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
24657 (*vsx_le_perm_load_v8hi): Likewise.
24658 (*vsx_le_perm_load_v16qi): Likewise.
24659 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
24660 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
24661 (*vsx_le_perm_store_v8hi): Likewise.
24662 (*vsx_le_perm_store_v16qi): Likewise.
24663 (eight unnamed splitters): Likewise.
24665 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
24667 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
24668 * config/rs6000/emmintrin.h: Likewise.
24669 * config/rs6000/mmintrin.h: Likewise.
24670 * config/rs6000/xmmintrin.h: Likewise.
24672 2018-01-10 David Malcolm <dmalcolm@redhat.com>
24675 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
24677 * tree.c (tree_nop_conversion): Return true for location wrapper
24679 (maybe_wrap_with_location): New function.
24680 (selftest::check_strip_nops): New function.
24681 (selftest::test_location_wrappers): New function.
24682 (selftest::tree_c_tests): Call it.
24683 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
24684 (maybe_wrap_with_location): New decl.
24685 (EXPR_LOCATION_WRAPPER_P): New macro.
24686 (location_wrapper_p): New inline function.
24687 (tree_strip_any_location_wrapper): New inline function.
24689 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
24692 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
24693 stack_realign_offset for the largest alignment of stack slot
24695 (ix86_find_max_used_stack_alignment): New function.
24696 (ix86_finalize_stack_frame_flags): Use it. Set
24697 max_used_stack_alignment if we don't realign stack.
24698 * config/i386/i386.h (machine_function): Add
24699 max_used_stack_alignment.
24701 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
24703 * config/arm/arm.opt (-mbranch-cost): New option.
24704 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
24707 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
24710 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
24711 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
24713 2018-01-10 Richard Biener <rguenther@suse.de>
24716 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
24717 early out so it also covers the case where we have a non-NULL
24720 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
24722 PR tree-optimization/83753
24723 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
24724 for non-strided grouped accesses if the number of elements is 1.
24726 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24729 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
24730 * i386.h (TARGET_USE_GATHER): Define.
24731 * x86-tune.def (X86_TUNE_USE_GATHER): New.
24733 2018-01-10 Martin Liska <mliska@suse.cz>
24736 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
24737 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
24739 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
24740 CLEANUP_NO_PARTITIONING is not set.
24742 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
24744 * doc/rtl.texi: Remove documentation of (const ...) wrappers
24745 for vectors, as a partial revert of r254296.
24746 * rtl.h (const_vec_p): Delete.
24747 (const_vec_duplicate_p): Don't test for vector CONSTs.
24748 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
24749 * expmed.c (make_tree): Likewise.
24752 * common.md (E, F): Use CONSTANT_P instead of checking for
24754 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
24755 checking for CONST_VECTOR.
24757 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24759 PR middle-end/83575
24760 * predict.c (force_edge_cold): Handle in more sane way edges
24761 with no prediction.
24763 2018-01-09 Carl Love <cel@us.ibm.com>
24765 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24767 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24768 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24769 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24770 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
24771 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24772 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
24773 * config/rs6000/rs6000-protos.h: Add extern defition for
24774 rs6000_generate_float2_double_code.
24775 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24777 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24778 (float2_v2df): Add define_expand.
24780 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
24783 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24784 op_mode in the force_to_mode call.
24786 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24788 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24789 instead of checking each element individually.
24790 (aarch64_evpc_uzp): Likewise.
24791 (aarch64_evpc_zip): Likewise.
24792 (aarch64_evpc_ext): Likewise.
24793 (aarch64_evpc_rev): Likewise.
24794 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24795 instead of checking each element individually. Return true without
24797 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
24798 whether all selected elements come from the same input, instead of
24799 checking each element individually. Remove calls to gen_rtx_REG,
24800 start_sequence and end_sequence and instead assert that no rtl is
24803 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24805 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
24806 order of HIGH and CONST checks.
24808 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
24810 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
24811 if the destination isn't an SSA_NAME.
24813 2018-01-09 Richard Biener <rguenther@suse.de>
24815 PR tree-optimization/83668
24816 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
24818 (canonicalize_loop_form): ... here, renamed from ...
24819 (canonicalize_loop_closed_ssa_form): ... this and amended to
24820 swap successor edges for loop exit blocks to make us use
24821 the RPO order we need for initial schedule generation.
24823 2018-01-09 Joseph Myers <joseph@codesourcery.com>
24825 PR tree-optimization/64811
24826 * match.pd: When optimizing comparisons with Inf, avoid
24827 introducing or losing exceptions from comparisons with NaN.
24829 2018-01-09 Martin Liska <mliska@suse.cz>
24832 * asan.c (shadow_mem_size): Add gcc_assert.
24834 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
24836 Don't save registers in main().
24839 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
24840 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
24841 * config/avr/avr.c (avr_set_current_function): Don't error if
24842 naked, OS_task or OS_main are specified at the same time.
24843 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
24845 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
24847 * common/config/avr/avr-common.c (avr_option_optimization_table):
24848 Switch on -mmain-is-OS_task for optimizing compilations.
24850 2018-01-09 Richard Biener <rguenther@suse.de>
24852 PR tree-optimization/83572
24853 * graphite.c: Include cfganal.h.
24854 (graphite_transform_loops): Connect infinite loops to exit
24855 and remove fake edges at the end.
24857 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24859 * ipa-inline.c (edge_badness): Revert accidental checkin.
24861 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
24864 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
24865 symbols; not inline clones.
24867 2018-01-09 Jakub Jelinek <jakub@redhat.com>
24870 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
24871 hard registers. Formatting fixes.
24873 PR preprocessor/83722
24874 * gcc.c (try_generate_repro): Pass
24875 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
24876 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
24879 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
24880 Kito Cheng <kito.cheng@gmail.com>
24882 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
24883 (riscv_leaf_function_p): Delete.
24884 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
24886 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24888 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
24890 (do_ifelse): New function.
24891 (do_isel): New function.
24892 (do_sub3): New function.
24893 (do_add3): New function.
24894 (do_load_mask_compare): New function.
24895 (do_overlap_load_compare): New function.
24896 (expand_compare_loop): New function.
24897 (expand_block_compare): Call expand_compare_loop() when appropriate.
24898 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
24899 option description.
24900 (-mblock-compare-inline-loop-limit): New option.
24902 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24905 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
24906 Reverse order of second and third operands in first alternative.
24907 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
24908 of first and second elements in UNSPEC_VPERMR vector.
24909 (altivec_expand_vec_perm_le): Likewise.
24911 2018-01-08 Jeff Law <law@redhat.com>
24913 PR rtl-optimizatin/81308
24914 * tree-switch-conversion.c (cfg_altered): New file scoped static.
24915 (process_switch): If group_case_labels makes a change, then set
24917 (pass_convert_switch::execute): If a switch is converted, then
24918 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
24920 PR rtl-optimization/81308
24921 * recog.c (split_all_insns): Conditionally cleanup the CFG after
24924 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
24926 PR target/83663 - Revert r255946
24927 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
24928 generation for cases where splatting a value is not useful.
24929 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
24930 across a vec_duplicate and a paradoxical subreg forming a vector
24931 mode to a vec_concat.
24933 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24935 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
24936 -march=armv8.3-a variants.
24937 * config/arm/t-multilib: Likewise.
24938 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
24940 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24942 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
24944 (cceq_ior_compare_complement): Give it a name so I can use it, and
24945 change boolean_or_operator predicate to boolean_operator so it can
24946 be used to generate a crand.
24947 (eqne): New code iterator.
24948 (bd/bd_neg): New code_attrs.
24949 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
24950 a single define_insn.
24951 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
24952 decrement (bdnzt/bdnzf/bdzt/bdzf).
24953 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
24954 with the new names of the branch decrement patterns, and added the
24955 names of the branch decrement conditional patterns.
24957 2018-01-08 Richard Biener <rguenther@suse.de>
24959 PR tree-optimization/83563
24960 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
24963 2018-01-08 Richard Biener <rguenther@suse.de>
24965 PR middle-end/83713
24966 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
24968 2018-01-08 Richard Biener <rguenther@suse.de>
24970 PR tree-optimization/83685
24971 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
24972 references to abnormals.
24974 2018-01-08 Richard Biener <rguenther@suse.de>
24977 * dwarf2out.c (output_indirect_strings): Handle empty
24978 skeleton_debug_str_hash.
24979 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
24981 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24983 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
24984 (emit_store_direct): Likewise.
24985 (arc_trampoline_adjust_address): Likewise.
24986 (arc_asm_trampoline_template): New function.
24987 (arc_initialize_trampoline): Use asm_trampoline_template.
24988 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
24989 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
24990 * config/arc/arc.md (flush_icache): Delete pattern.
24992 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
24994 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
24995 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
24998 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
25001 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
25002 by not USED_FOR_TARGET.
25003 (make_pass_resolve_sw_modes): Likewise.
25005 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
25007 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
25010 2018-01-08 Richard Biener <rguenther@suse.de>
25012 PR middle-end/83580
25013 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
25015 2018-01-08 Richard Biener <rguenther@suse.de>
25017 PR middle-end/83517
25018 * match.pd ((t * 2) / 2) -> t): Add missing :c.
25020 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
25022 PR middle-end/81897
25023 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
25024 basic blocks with a small number of successors.
25025 (convert_control_dep_chain_into_preds): Improve handling of
25027 (dump_predicates): Split apart into...
25028 (dump_pred_chain): ...here...
25029 (dump_pred_info): ...and here.
25030 (can_one_predicate_be_invalidated_p): Add debugging printfs.
25031 (can_chain_union_be_invalidated_p): Improve check for invalidation
25033 (uninit_uses_cannot_happen): Avoid unnecessary if
25034 convert_control_dep_chain_into_preds yielded nothing.
25036 2018-01-06 Martin Sebor <msebor@redhat.com>
25038 PR tree-optimization/83640
25039 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
25040 subtracting negative offset from size.
25041 (builtin_access::overlap): Adjust offset bounds of the access to fall
25042 within the size of the object if possible.
25044 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
25046 PR rtl-optimization/83699
25047 * expmed.c (extract_bit_field_1): Restrict the vector usage of
25048 extract_bit_field_as_subreg to cases in which the extracted
25049 value is also a vector.
25051 * lra-constraints.c (process_alt_operands): Test for the equivalence
25052 substitutions when detecting a possible reload cycle.
25054 2018-01-06 Jakub Jelinek <jakub@redhat.com>
25057 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
25058 by default if flag_selective_schedling{,2}. Formatting fixes.
25060 PR rtl-optimization/83682
25061 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
25062 if it has non-VECTOR_MODE element mode.
25063 (vec_duplicate_p): Likewise.
25065 PR middle-end/83694
25066 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
25067 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
25069 2018-01-05 Jakub Jelinek <jakub@redhat.com>
25072 * config/i386/i386-builtin.def
25073 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
25074 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
25075 Require also OPTION_MASK_ISA_AVX512F in addition to
25076 OPTION_MASK_ISA_GFNI.
25077 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
25078 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
25079 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
25080 to OPTION_MASK_ISA_GFNI.
25081 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
25082 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
25083 OPTION_MASK_ISA_AVX512BW.
25084 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
25085 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
25086 addition to OPTION_MASK_ISA_GFNI.
25087 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
25088 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
25089 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
25090 to OPTION_MASK_ISA_GFNI.
25091 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
25092 a requirement for all ISAs rather than any of them with a few
25094 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
25096 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
25097 bitmasks to be enabled with 3 exceptions, instead of requiring any
25098 enabled ISA with lots of exceptions.
25099 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
25100 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
25101 Change avx512bw in isa attribute to avx512f.
25102 * config/i386/sgxintrin.h: Add license boilerplate.
25103 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
25104 to __AVX512F__ and __AVX512VL to __AVX512VL__.
25105 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
25106 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
25108 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
25109 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
25110 temporarily sse2 rather than sse if not enabled already.
25113 * config/i386/sse.md (VI248_VLBW): Rename to ...
25114 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
25115 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
25116 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
25117 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
25118 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
25119 mode iterator instead of VI248_VLBW.
25121 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
25123 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
25124 (record_modified): Skip clobbers; add debug output.
25125 (param_change_prob): Use sreal frequencies.
25127 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
25129 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25130 punt for user-aligned variables.
25132 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
25134 * tree-chrec.c (chrec_contains_symbols): Return true for
25137 2018-01-05 Sudakshina Das <sudi.das@arm.com>
25140 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
25141 of (x|y) == x for BICS pattern.
25143 2018-01-05 Jakub Jelinek <jakub@redhat.com>
25145 PR tree-optimization/83605
25146 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
25147 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
25150 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
25152 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
25153 * config/epiphany/rtems.h: New file.
25155 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25156 Uros Bizjak <ubizjak@gmail.com>
25159 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
25160 QIreg_operand instead of register_operand predicate.
25161 * config/i386/i386.c (ix86_rop_should_change_byte_p,
25162 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
25163 comments instead of -fmitigate[-_]rop.
25165 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25168 * cgraphunit.c (symbol_table::compile): Switch to text_section
25169 before calling assembly_start debug hook.
25170 * run-rtl-passes.c (run_rtl_passes): Likewise.
25173 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25175 * tree-vrp.c (extract_range_from_binary_expr_1): Check
25176 range_int_cst_p rather than !symbolic_range_p before calling
25177 extract_range_from_multiplicative_op_1.
25179 2018-01-04 Jeff Law <law@redhat.com>
25181 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
25182 redundant test in assertion.
25184 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25186 * doc/rtl.texi: Document machine_mode wrapper classes.
25188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25190 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
25191 using tree_to_uhwi.
25193 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25195 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
25196 the VEC_PERM_EXPR fold to fail.
25198 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25201 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
25202 to switched_sections.
25204 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25207 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
25208 test for d.testing.
25210 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
25213 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
25214 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
25216 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25219 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
25220 is BLKmode and bitpos not zero or mode change is needed.
25222 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25225 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
25228 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
25231 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
25232 instead of MULT rtx. Update all corresponding splitters.
25233 (*saddl_se): Ditto.
25234 (*ssub<modesuffix>): Ditto.
25235 (*ssubl_se): Ditto.
25236 (*cmp_sadd_di): Update split patterns.
25237 (*cmp_sadd_si): Ditto.
25238 (*cmp_sadd_sidi): Ditto.
25239 (*cmp_ssub_di): Ditto.
25240 (*cmp_ssub_si): Ditto.
25241 (*cmp_ssub_sidi): Ditto.
25242 * config/alpha/predicates.md (const23_operand): New predicate.
25243 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
25244 Look for ASHIFT, not MULT inner operand.
25245 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
25247 2018-01-04 Martin Liska <mliska@suse.cz>
25249 PR gcov-profile/83669
25250 * gcov.c (output_intermediate_file): Add version to intermediate
25252 * doc/gcov.texi: Document new field 'version' in intermediate
25253 file format. Fix location of '-k' option of gcov command.
25255 2018-01-04 Martin Liska <mliska@suse.cz>
25258 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
25260 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25262 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
25264 2018-01-03 Martin Sebor <msebor@redhat.com>
25266 PR tree-optimization/83655
25267 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
25268 checking calls with invalid arguments.
25270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25272 * tree-vect-stmts.c (vect_get_store_rhs): New function.
25273 (vectorizable_mask_load_store): Delete.
25274 (vectorizable_call): Return false for masked loads and stores.
25275 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
25276 instead of gimple_assign_rhs1.
25277 (vectorizable_load): Handle IFN_MASK_LOAD.
25278 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
25280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25282 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
25284 (vectorizable_mask_load_store): ...here.
25285 (vectorizable_load): ...and here.
25287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25289 * tree-vect-stmts.c (vect_build_all_ones_mask)
25290 (vect_build_zero_merge_argument): New functions, split out from...
25291 (vectorizable_load): ...here.
25293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25295 * tree-vect-stmts.c (vect_check_store_rhs): New function,
25297 (vectorizable_mask_load_store): ...here.
25298 (vectorizable_store): ...and here.
25300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25302 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
25304 (vectorizable_mask_load_store): ...here.
25306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25308 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
25309 (vect_model_store_cost): Take a vec_load_store_type instead of a
25311 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
25312 (vect_model_store_cost): Take a vec_load_store_type instead of a
25314 (vectorizable_mask_load_store): Update accordingly.
25315 (vectorizable_store): Likewise.
25316 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
25318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25320 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
25321 IFN_MASK_LOAD calls here rather than...
25322 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
25324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25325 Alan Hayward <alan.hayward@arm.com>
25326 David Sherwood <david.sherwood@arm.com>
25328 * expmed.c (extract_bit_field_1): For vector extracts,
25329 fall back to extract_bit_field_as_subreg if vec_extract
25332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25333 Alan Hayward <alan.hayward@arm.com>
25334 David Sherwood <david.sherwood@arm.com>
25336 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
25337 they are variable or constant sized.
25338 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
25339 slots for constant-sized data.
25341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25342 Alan Hayward <alan.hayward@arm.com>
25343 David Sherwood <david.sherwood@arm.com>
25345 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
25346 handling COND_EXPRs with boolean comparisons, try to find a better
25347 basis for the mask type than the boolean itself.
25349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25351 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
25352 is calculated and how it can be overridden.
25353 * genmodes.c (max_bitsize_mode_any_mode): New variable.
25354 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
25356 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
25359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25360 Alan Hayward <alan.hayward@arm.com>
25361 David Sherwood <david.sherwood@arm.com>
25363 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
25364 Remove the mode argument.
25365 (aarch64_simd_valid_immediate): Remove the mode and inverse
25367 * config/aarch64/iterators.md (bitsize): New iterator.
25368 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
25369 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
25370 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
25371 aarch64_simd_valid_immediate.
25372 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
25373 (aarch64_reg_or_bic_imm): Likewise.
25374 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
25375 with an insn_type enum and msl with a modifier_type enum.
25376 Replace element_width with a scalar_mode. Change the shift
25377 to unsigned int. Add constructors for scalar_float_mode and
25378 scalar_int_mode elements.
25379 (aarch64_vect_float_const_representable_p): Delete.
25380 (aarch64_can_const_movi_rtx_p)
25381 (aarch64_simd_scalar_immediate_valid_for_move)
25382 (aarch64_simd_make_constant): Update call to
25383 aarch64_simd_valid_immediate.
25384 (aarch64_advsimd_valid_immediate_hs): New function.
25385 (aarch64_advsimd_valid_immediate): Likewise.
25386 (aarch64_simd_valid_immediate): Remove mode and inverse
25387 arguments. Rewrite to use the above. Use const_vec_duplicate_p
25388 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
25389 and aarch64_float_const_representable_p on the result.
25390 (aarch64_output_simd_mov_immediate): Remove mode argument.
25391 Update call to aarch64_simd_valid_immediate and use of
25392 simd_immediate_info.
25393 (aarch64_output_scalar_simd_mov_immediate): Update call
25396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25397 Alan Hayward <alan.hayward@arm.com>
25398 David Sherwood <david.sherwood@arm.com>
25400 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
25401 (mode_nunits): Likewise CONST_MODE_NUNITS.
25402 * machmode.def (ADJUST_NUNITS): Document.
25403 * genmodes.c (mode_data::need_nunits_adj): New field.
25404 (blank_mode): Update accordingly.
25405 (adj_nunits): New variable.
25406 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
25408 (emit_mode_size_inline): Set need_bytesize_adj for all modes
25409 listed in adj_nunits.
25410 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
25411 listed in adj_nunits. Don't emit case statements for such modes.
25412 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
25413 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
25414 nothing if adj_nunits is nonnull.
25415 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
25416 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
25417 (emit_mode_fbit): Update use of print_maybe_const_decl.
25418 (emit_move_size): Likewise. Treat the array as non-const
25420 (emit_mode_adjustments): Handle adj_nunits.
25422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25424 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
25425 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
25426 (VECTOR_MODES): Use it.
25427 (make_vector_modes): Take the prefix as an argument.
25429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25430 Alan Hayward <alan.hayward@arm.com>
25431 David Sherwood <david.sherwood@arm.com>
25433 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
25434 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
25435 for MODE_VECTOR_BOOL.
25436 * machmode.def (VECTOR_BOOL_MODE): Document.
25437 * genmodes.c (VECTOR_BOOL_MODE): New macro.
25438 (make_vector_bool_mode): New function.
25439 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
25441 * lto-streamer-in.c (lto_input_mode_table): Likewise.
25442 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
25444 * stor-layout.c (int_mode_for_mode): Likewise.
25445 * tree.c (build_vector_type_for_mode): Likewise.
25446 * varasm.c (output_constant_pool_2): Likewise.
25447 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
25448 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
25449 for MODE_VECTOR_BOOL.
25450 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
25451 of mode class checks.
25452 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
25453 instead of a list of mode class checks.
25454 (expand_vector_scalar_condition): Likewise.
25455 (type_for_widest_vector_mode): Handle BImode as an inner mode.
25457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25458 Alan Hayward <alan.hayward@arm.com>
25459 David Sherwood <david.sherwood@arm.com>
25461 * machmode.h (mode_size): Change from unsigned short to
25463 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
25464 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25465 or if measurement_type is not polynomial.
25466 (fixed_size_mode::includes_p): Check for constant-sized modes.
25467 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
25468 return a poly_uint16 rather than an unsigned short.
25469 (emit_mode_size): Change the type of mode_size from unsigned short
25470 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
25471 (emit_mode_adjustments): Cope with polynomial vector sizes.
25472 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25474 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25476 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
25477 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
25478 * caller-save.c (setup_save_areas): Likewise.
25479 (replace_reg_with_saved_mem): Likewise.
25480 * calls.c (emit_library_call_value_1): Likewise.
25481 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
25482 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
25483 (gen_lowpart_for_combine): Likewise.
25484 * convert.c (convert_to_integer_1): Likewise.
25485 * cse.c (equiv_constant, cse_insn): Likewise.
25486 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
25487 (cselib_subst_to_values): Likewise.
25488 * dce.c (word_dce_process_block): Likewise.
25489 * df-problems.c (df_word_lr_mark_ref): Likewise.
25490 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
25491 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
25492 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
25493 (rtl_for_decl_location): Likewise.
25494 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
25495 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
25496 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
25497 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
25498 (expand_expr_real_1): Likewise.
25499 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
25500 (pad_below): Likewise.
25501 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25502 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
25503 * ira.c (get_subreg_tracking_sizes): Likewise.
25504 * ira-build.c (ira_create_allocno_objects): Likewise.
25505 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
25506 (ira_sort_regnos_for_alter_reg): Likewise.
25507 * ira-costs.c (record_operand_costs): Likewise.
25508 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
25509 (resolve_simple_move): Likewise.
25510 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
25511 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
25512 (lra_constraints): Likewise.
25513 (CONST_POOL_OK_P): Reject variable-sized modes.
25514 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
25515 (add_pseudo_to_slot, lra_spill): Likewise.
25516 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25517 * optabs-query.c (get_best_extraction_insn): Likewise.
25518 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25519 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
25520 (expand_mult_highpart, valid_multiword_target_p): Likewise.
25521 * recog.c (offsettable_address_addr_space_p): Likewise.
25522 * regcprop.c (maybe_mode_change): Likewise.
25523 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
25524 * regrename.c (build_def_use): Likewise.
25525 * regstat.c (dump_reg_info): Likewise.
25526 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
25527 (find_reloads, find_reloads_subreg_address): Likewise.
25528 * reload1.c (eliminate_regs_1): Likewise.
25529 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
25530 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
25531 (simplify_binary_operation_1, simplify_subreg): Likewise.
25532 * targhooks.c (default_function_arg_padding): Likewise.
25533 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
25534 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
25535 (verify_gimple_assign_ternary): Likewise.
25536 * tree-inline.c (estimate_move_cost): Likewise.
25537 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25538 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
25539 (get_address_cost_ainc): Likewise.
25540 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
25541 (vect_supportable_dr_alignment): Likewise.
25542 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25543 (vectorizable_reduction): Likewise.
25544 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
25545 (vectorizable_operation, vectorizable_load): Likewise.
25546 * tree.c (build_same_sized_truth_vector_type): Likewise.
25547 * valtrack.c (cleanup_auto_inc_dec): Likewise.
25548 * var-tracking.c (emit_note_insn_var_location): Likewise.
25549 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
25550 (ADDR_VEC_ALIGN): Likewise.
25552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25553 Alan Hayward <alan.hayward@arm.com>
25554 David Sherwood <david.sherwood@arm.com>
25556 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
25558 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25559 or if measurement_type is polynomial.
25560 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
25561 * combine.c (make_extraction): Likewise.
25562 * dse.c (find_shift_sequence): Likewise.
25563 * dwarf2out.c (mem_loc_descriptor): Likewise.
25564 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
25565 (extract_bit_field, extract_low_bits): Likewise.
25566 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
25567 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
25568 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
25569 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
25570 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25571 * reload.c (find_reloads): Likewise.
25572 * reload1.c (alter_reg): Likewise.
25573 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
25574 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
25575 * tree-if-conv.c (predicate_mem_writes): Likewise.
25576 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
25577 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
25578 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
25579 * valtrack.c (dead_debug_insert_temp): Likewise.
25580 * varasm.c (mergeable_constant_section): Likewise.
25581 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
25583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25584 Alan Hayward <alan.hayward@arm.com>
25585 David Sherwood <david.sherwood@arm.com>
25587 * expr.c (expand_assignment): Cope with polynomial mode sizes
25588 when assigning to a CONCAT.
25590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25591 Alan Hayward <alan.hayward@arm.com>
25592 David Sherwood <david.sherwood@arm.com>
25594 * machmode.h (mode_precision): Change from unsigned short to
25596 (mode_to_precision): Return a poly_uint16 rather than an unsigned
25598 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
25599 or if measurement_type is not polynomial.
25600 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
25601 in which the mode is already known to be a scalar_int_mode.
25602 * genmodes.c (emit_mode_precision): Change the type of mode_precision
25603 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
25605 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25606 for GET_MODE_PRECISION.
25607 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25608 for GET_MODE_PRECISION.
25609 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
25611 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
25612 (expand_field_assignment, make_extraction): Likewise.
25613 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
25614 (get_last_value): Likewise.
25615 * convert.c (convert_to_integer_1): Likewise.
25616 * cse.c (cse_insn): Likewise.
25617 * expr.c (expand_expr_real_1): Likewise.
25618 * lra-constraints.c (simplify_operand_subreg): Likewise.
25619 * optabs-query.c (can_atomic_load_p): Likewise.
25620 * optabs.c (expand_atomic_load): Likewise.
25621 (expand_atomic_store): Likewise.
25622 * ree.c (combine_reaching_defs): Likewise.
25623 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
25624 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
25625 * tree.h (type_has_mode_precision_p): Likewise.
25626 * ubsan.c (instrument_si_overflow): Likewise.
25628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25629 Alan Hayward <alan.hayward@arm.com>
25630 David Sherwood <david.sherwood@arm.com>
25632 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
25633 polynomial numbers of units.
25634 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
25635 (valid_vector_subparts_p): New function.
25636 (build_vector_type): Remove temporary shim and take the number
25637 of units as a poly_uint64 rather than an int.
25638 (build_opaque_vector_type): Take the number of units as a
25639 poly_uint64 rather than an int.
25640 * tree.c (build_vector_from_ctor): Handle polynomial
25641 TYPE_VECTOR_SUBPARTS.
25642 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
25643 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
25644 (build_vector_from_val): If the number of units is variable,
25645 use build_vec_duplicate_cst for constant operands and
25646 VEC_DUPLICATE_EXPR otherwise.
25647 (make_vector_type): Remove temporary is_constant ().
25648 (build_vector_type, build_opaque_vector_type): Take the number of
25649 units as a poly_uint64 rather than an int.
25650 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
25652 * cfgexpand.c (expand_debug_expr): Likewise.
25653 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
25654 (store_constructor, expand_expr_real_1): Likewise.
25655 (const_scalar_mask_from_tree): Likewise.
25656 * fold-const-call.c (fold_const_reduction): Likewise.
25657 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
25658 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
25659 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
25660 (fold_relational_const): Likewise.
25661 (native_interpret_vector): Likewise. Change the size from an
25662 int to an unsigned int.
25663 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
25664 TYPE_VECTOR_SUBPARTS.
25665 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
25666 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
25667 duplicating a non-constant operand into a variable-length vector.
25668 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
25669 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
25670 * ipa-icf.c (sem_variable::equals): Likewise.
25671 * match.pd: Likewise.
25672 * omp-simd-clone.c (simd_clone_subparts): Likewise.
25673 * print-tree.c (print_node): Likewise.
25674 * stor-layout.c (layout_type): Likewise.
25675 * targhooks.c (default_builtin_vectorization_cost): Likewise.
25676 * tree-cfg.c (verify_gimple_comparison): Likewise.
25677 (verify_gimple_assign_binary): Likewise.
25678 (verify_gimple_assign_ternary): Likewise.
25679 (verify_gimple_assign_single): Likewise.
25680 * tree-pretty-print.c (dump_generic_node): Likewise.
25681 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25682 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
25683 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
25684 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
25685 (vect_shift_permute_load_chain): Likewise.
25686 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
25687 (expand_vector_condition, optimize_vector_constructor): Likewise.
25688 (lower_vec_perm, get_compute_type): Likewise.
25689 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25690 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
25691 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
25692 (vect_recog_mask_conversion_pattern): Likewise.
25693 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
25694 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
25695 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25696 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
25697 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
25698 (vectorizable_shift, vectorizable_operation, vectorizable_store)
25699 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
25700 (supportable_widening_operation): Likewise.
25701 (supportable_narrowing_operation): Likewise.
25702 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
25704 * varasm.c (output_constant): Likewise.
25706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25707 Alan Hayward <alan.hayward@arm.com>
25708 David Sherwood <david.sherwood@arm.com>
25710 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
25711 so that both the length == 3 and length != 3 cases set up their
25712 own permute vectors. Add comments explaining why we know the
25713 number of elements is constant.
25714 (vect_permute_load_chain): Likewise.
25716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25717 Alan Hayward <alan.hayward@arm.com>
25718 David Sherwood <david.sherwood@arm.com>
25720 * machmode.h (mode_nunits): Change from unsigned char to
25722 (ONLY_FIXED_SIZE_MODES): New macro.
25723 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
25724 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
25725 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
25727 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
25728 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
25729 or if measurement_type is not polynomial.
25730 * genmodes.c (ZERO_COEFFS): New macro.
25731 (emit_mode_nunits_inline): Make mode_nunits_inline return a
25733 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
25734 Use ZERO_COEFFS when emitting initializers.
25735 * data-streamer.h (bp_pack_poly_value): New function.
25736 (bp_unpack_poly_value): Likewise.
25737 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25738 for GET_MODE_NUNITS.
25739 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25740 for GET_MODE_NUNITS.
25741 * tree.c (make_vector_type): Remove temporary shim and make
25742 the real function take the number of units as a poly_uint64
25743 rather than an int.
25744 (build_vector_type_for_mode): Handle polynomial nunits.
25745 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
25746 * emit-rtl.c (const_vec_series_p_1): Likewise.
25747 (gen_rtx_CONST_VECTOR): Likewise.
25748 * fold-const.c (test_vec_duplicate_folding): Likewise.
25749 * genrecog.c (validate_pattern): Likewise.
25750 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
25751 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25752 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
25753 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
25754 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
25755 * rtlanal.c (subreg_get_info): Likewise.
25756 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25757 (vect_grouped_load_supported): Likewise.
25758 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
25759 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
25760 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
25761 (simplify_const_unary_operation, simplify_binary_operation_1)
25762 (simplify_const_binary_operation, simplify_ternary_operation)
25763 (test_vector_ops_duplicate, test_vector_ops): Likewise.
25764 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25765 instead of CONST_VECTOR_NUNITS.
25766 * varasm.c (output_constant_pool_2): Likewise.
25767 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25768 explicit-encoded elements in the XVEC for variable-length vectors.
25770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25772 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25775 Alan Hayward <alan.hayward@arm.com>
25776 David Sherwood <david.sherwood@arm.com>
25778 * coretypes.h (fixed_size_mode): Declare.
25779 (fixed_size_mode_pod): New typedef.
25780 * builtins.h (target_builtins::x_apply_args_mode)
25781 (target_builtins::x_apply_result_mode): Change type to
25782 fixed_size_mode_pod.
25783 * builtins.c (apply_args_size, apply_result_size, result_vector)
25784 (expand_builtin_apply_args_1, expand_builtin_apply)
25785 (expand_builtin_return): Update accordingly.
25787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25789 * cse.c (hash_rtx_cb): Hash only the encoded elements.
25790 * cselib.c (cselib_hash_rtx): Likewise.
25791 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25792 CONST_VECTOR encoding.
25794 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25795 Jeff Law <law@redhat.com>
25798 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
25799 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
25800 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
25801 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
25804 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
25805 explicitly probe *sp in a noreturn function if there were any callee
25806 register saves or frame pointer is needed.
25808 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25811 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
25812 BLKmode for ternary, binary or unary expressions.
25815 * var-tracking.c (delete_vta_debug_insn): New inline function.
25816 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
25817 insns from get_insns () to NULL instead of each bb separately.
25818 Use delete_vta_debug_insn. No longer static.
25819 (vt_debug_insns_local, variable_tracking_main_1): Adjust
25820 delete_vta_debug_insns callers.
25821 * rtl.h (delete_vta_debug_insns): Declare.
25822 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
25823 instead of variable_tracking_main.
25825 2018-01-03 Martin Sebor <msebor@redhat.com>
25827 PR tree-optimization/83603
25828 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
25829 arguments past the endof the argument list in functions declared
25830 without a prototype.
25831 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
25832 Avoid checking when arguments are null.
25834 2018-01-03 Martin Sebor <msebor@redhat.com>
25837 * doc/extend.texi (attribute const): Fix a typo.
25838 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
25839 issuing -Wsuggest-attribute for void functions.
25841 2018-01-03 Martin Sebor <msebor@redhat.com>
25843 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
25844 offset_int::from instead of wide_int::to_shwi.
25845 (maybe_diag_overlap): Remove assertion.
25846 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
25847 * gimple-ssa-sprintf.c (format_directive): Same.
25848 (parse_directive): Same.
25849 (sprintf_dom_walker::compute_format_length): Same.
25850 (try_substitute_return_value): Same.
25852 2018-01-03 Jeff Law <law@redhat.com>
25854 PR middle-end/83654
25855 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
25856 non-constant residual for zero at runtime and avoid probing in
25857 that case. Reorganize code for trailing problem to mirror handling
25860 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25862 PR tree-optimization/83501
25863 * tree-ssa-strlen.c (get_string_cst): New.
25864 (handle_char_store): Call get_string_cst.
25866 2018-01-03 Martin Liska <mliska@suse.cz>
25868 PR tree-optimization/83593
25869 * tree-ssa-strlen.c: Include tree-cfg.h.
25870 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
25871 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
25872 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
25874 (strlen_dom_walker::before_dom_children): Call
25875 gimple_purge_dead_eh_edges. Dump tranformation with details
25877 (strlen_dom_walker::before_dom_children): Update call by adding
25878 new argument cleanup_eh.
25879 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
25881 2018-01-03 Martin Liska <mliska@suse.cz>
25884 * cif-code.def (VARIADIC_THUNK): New enum value.
25885 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
25888 2018-01-03 Jan Beulich <jbeulich@suse.com>
25890 * sse.md (mov<mode>_internal): Tighten condition for when to use
25891 vmovdqu<ssescalarsize> for TI and OI modes.
25893 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25895 Update copyright years.
25897 2018-01-03 Martin Liska <mliska@suse.cz>
25900 * ipa-visibility.c (function_and_variable_visibility): Skip
25901 functions with noipa attribure.
25903 2018-01-03 Jakub Jelinek <jakub@redhat.com>
25905 * gcc.c (process_command): Update copyright notice dates.
25906 * gcov-dump.c (print_version): Ditto.
25907 * gcov.c (print_version): Ditto.
25908 * gcov-tool.c (print_version): Ditto.
25909 * gengtype.c (create_file): Ditto.
25910 * doc/cpp.texi: Bump @copying's copyright year.
25911 * doc/cppinternals.texi: Ditto.
25912 * doc/gcc.texi: Ditto.
25913 * doc/gccint.texi: Ditto.
25914 * doc/gcov.texi: Ditto.
25915 * doc/install.texi: Ditto.
25916 * doc/invoke.texi: Ditto.
25918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25920 * vector-builder.h (vector_builder::m_full_nelts): Change from
25921 unsigned int to poly_uint64.
25922 (vector_builder::full_nelts): Update prototype accordingly.
25923 (vector_builder::new_vector): Likewise.
25924 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
25925 (vector_builder::operator ==): Likewise.
25926 (vector_builder::finalize): Likewise.
25927 * int-vector-builder.h (int_vector_builder::int_vector_builder):
25928 Take the number of elements as a poly_uint64 rather than an
25930 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
25931 from unsigned int to poly_uint64.
25932 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
25933 (vec_perm_indices::new_vector): Likewise.
25934 (vec_perm_indices::length): Likewise.
25935 (vec_perm_indices::nelts_per_input): Likewise.
25936 (vec_perm_indices::input_nelts): Likewise.
25937 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
25938 number of elements per input as a poly_uint64 rather than an
25939 unsigned int. Use the original encoding for variable-length
25940 vectors, rather than clamping each individual element.
25941 For the second and subsequent elements in each pattern,
25942 clamp the step and base before clamping their sum.
25943 (vec_perm_indices::series_p): Handle polynomial element counts.
25944 (vec_perm_indices::all_in_range_p): Likewise.
25945 (vec_perm_indices_to_tree): Likewise.
25946 (vec_perm_indices_to_rtx): Likewise.
25947 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
25948 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
25949 (tree_vector_builder::new_binary_operation): Handle polynomial
25950 element counts. Return false if we need to know the number
25951 of elements at compile time.
25952 * fold-const.c (fold_vec_perm): Punt if the number of elements
25953 isn't known at compile time.
25955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25957 * vec-perm-indices.h (vec_perm_builder): Change element type
25958 from HOST_WIDE_INT to poly_int64.
25959 (vec_perm_indices::element_type): Update accordingly.
25960 (vec_perm_indices::clamp): Handle polynomial element_types.
25961 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25962 (vec_perm_indices::all_in_range_p): Likewise.
25963 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
25965 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
25966 polynomial vec_perm_indices element types.
25967 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
25968 * fold-const.c (fold_vec_perm): Likewise.
25969 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
25970 * tree-vect-generic.c (lower_vec_perm): Likewise.
25971 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25972 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
25973 element type to HOST_WIDE_INT.
25975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25976 Alan Hayward <alan.hayward@arm.com>
25977 David Sherwood <david.sherwood@arm.com>
25979 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
25980 rather than an int. Use plus_constant.
25981 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
25982 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
25984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25985 Alan Hayward <alan.hayward@arm.com>
25986 David Sherwood <david.sherwood@arm.com>
25988 * calls.c (emit_call_1, expand_call): Change struct_value_size from
25989 a HOST_WIDE_INT to a poly_int64.
25991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25992 Alan Hayward <alan.hayward@arm.com>
25993 David Sherwood <david.sherwood@arm.com>
25995 * calls.c (load_register_parameters): Cope with polynomial
25996 mode sizes. Require a constant size for BLKmode parameters
25997 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
25998 forces a parameter to be padded at the lsb end in order to
25999 fill a complete number of words, require the parameter size
26000 to be ordered wrt UNITS_PER_WORD.
26002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26003 Alan Hayward <alan.hayward@arm.com>
26004 David Sherwood <david.sherwood@arm.com>
26006 * reload1.c (spill_stack_slot_width): Change element type
26007 from unsigned int to poly_uint64_pod.
26008 (alter_reg): Treat mode sizes as polynomial.
26010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26011 Alan Hayward <alan.hayward@arm.com>
26012 David Sherwood <david.sherwood@arm.com>
26014 * reload.c (complex_word_subreg_p): New function.
26015 (reload_inner_reg_of_subreg, push_reload): Use it.
26017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26018 Alan Hayward <alan.hayward@arm.com>
26019 David Sherwood <david.sherwood@arm.com>
26021 * lra-constraints.c (process_alt_operands): Reject matched
26022 operands whose sizes aren't ordered.
26023 (match_reload): Refer to this check here.
26025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26026 Alan Hayward <alan.hayward@arm.com>
26027 David Sherwood <david.sherwood@arm.com>
26029 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
26030 that the mode size is in the set {1, 2, 4, 8, 16}.
26032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26033 Alan Hayward <alan.hayward@arm.com>
26034 David Sherwood <david.sherwood@arm.com>
26036 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
26037 Use plus_constant instead of gen_rtx_PLUS.
26039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26040 Alan Hayward <alan.hayward@arm.com>
26041 David Sherwood <david.sherwood@arm.com>
26043 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
26044 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
26045 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
26046 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
26047 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
26048 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
26049 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
26050 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
26051 * config/i386/i386.c (ix86_push_rounding): ...this new function.
26052 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
26054 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
26055 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
26056 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
26057 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
26058 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
26059 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
26060 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
26061 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
26062 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
26063 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
26065 * expr.c (emit_move_resolve_push): Treat the input and result
26066 of PUSH_ROUNDING as a poly_int64.
26067 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
26068 (emit_push_insn): Likewise.
26069 * lra-eliminations.c (mark_not_eliminable): Likewise.
26070 * recog.c (push_operand): Likewise.
26071 * reload1.c (elimination_effects): Likewise.
26072 * rtlanal.c (nonzero_bits1): Likewise.
26073 * calls.c (store_one_arg): Likewise. Require the padding to be
26074 known at compile time.
26076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26077 Alan Hayward <alan.hayward@arm.com>
26078 David Sherwood <david.sherwood@arm.com>
26080 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
26081 Use plus_constant instead of gen_rtx_PLUS.
26083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26084 Alan Hayward <alan.hayward@arm.com>
26085 David Sherwood <david.sherwood@arm.com>
26087 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
26088 rather than an int.
26090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26091 Alan Hayward <alan.hayward@arm.com>
26092 David Sherwood <david.sherwood@arm.com>
26094 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
26095 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
26096 via stack temporaries. Treat the mode size as polynomial too.
26098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26099 Alan Hayward <alan.hayward@arm.com>
26100 David Sherwood <david.sherwood@arm.com>
26102 * expr.c (expand_expr_real_2): When handling conversions involving
26103 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
26104 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
26105 as a poly_uint64 too.
26107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26108 Alan Hayward <alan.hayward@arm.com>
26109 David Sherwood <david.sherwood@arm.com>
26111 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
26113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26114 Alan Hayward <alan.hayward@arm.com>
26115 David Sherwood <david.sherwood@arm.com>
26117 * combine.c (can_change_dest_mode): Handle polynomial
26118 REGMODE_NATURAL_SIZE.
26119 * expmed.c (store_bit_field_1): Likewise.
26120 * expr.c (store_constructor): Likewise.
26121 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
26122 and polynomial REGMODE_NATURAL_SIZE.
26123 (gen_lowpart_common): Likewise.
26124 * reginfo.c (record_subregs_of_mode): Likewise.
26125 * rtlanal.c (read_modify_subreg_p): Likewise.
26127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26128 Alan Hayward <alan.hayward@arm.com>
26129 David Sherwood <david.sherwood@arm.com>
26131 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
26132 numbers of elements.
26134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26135 Alan Hayward <alan.hayward@arm.com>
26136 David Sherwood <david.sherwood@arm.com>
26138 * match.pd: Cope with polynomial numbers of vector elements.
26140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26141 Alan Hayward <alan.hayward@arm.com>
26142 David Sherwood <david.sherwood@arm.com>
26144 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
26145 in a POINTER_PLUS_EXPR.
26147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26148 Alan Hayward <alan.hayward@arm.com>
26149 David Sherwood <david.sherwood@arm.com>
26151 * omp-simd-clone.c (simd_clone_subparts): New function.
26152 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
26153 (ipa_simd_modify_function_body): Likewise.
26155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26156 Alan Hayward <alan.hayward@arm.com>
26157 David Sherwood <david.sherwood@arm.com>
26159 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
26160 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
26161 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
26162 (expand_vector_condition, vector_element): Likewise.
26163 (subparts_gt): New function.
26164 (get_compute_type): Use subparts_gt.
26165 (count_type_subparts): Delete.
26166 (expand_vector_operations_1): Use subparts_gt instead of
26167 count_type_subparts.
26169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26170 Alan Hayward <alan.hayward@arm.com>
26171 David Sherwood <david.sherwood@arm.com>
26173 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
26174 (vect_compile_time_alias): ...this new function. Do the calculation
26175 on poly_ints rather than trees.
26176 (vect_prune_runtime_alias_test_list): Update call accordingly.
26178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26179 Alan Hayward <alan.hayward@arm.com>
26180 David Sherwood <david.sherwood@arm.com>
26182 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
26184 (vect_schedule_slp_instance): Likewise.
26186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26187 Alan Hayward <alan.hayward@arm.com>
26188 David Sherwood <david.sherwood@arm.com>
26190 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
26191 constant and extern definitions for variable-length vectors.
26192 (vect_get_constant_vectors): Note that the number of units
26193 is known to be constant.
26195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26196 Alan Hayward <alan.hayward@arm.com>
26197 David Sherwood <david.sherwood@arm.com>
26199 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
26200 of units as polynomial. Choose between WIDE and NARROW based
26203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26204 Alan Hayward <alan.hayward@arm.com>
26205 David Sherwood <david.sherwood@arm.com>
26207 * tree-vect-stmts.c (simd_clone_subparts): New function.
26208 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
26210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26211 Alan Hayward <alan.hayward@arm.com>
26212 David Sherwood <david.sherwood@arm.com>
26214 * tree-vect-stmts.c (vectorizable_call): Treat the number of
26215 vectors as polynomial. Use build_index_vector for
26216 IFN_GOMP_SIMD_LANE.
26218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26219 Alan Hayward <alan.hayward@arm.com>
26220 David Sherwood <david.sherwood@arm.com>
26222 * tree-vect-stmts.c (get_load_store_type): Treat the number of
26223 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
26224 for variable-length vectors.
26225 (vectorizable_mask_load_store): Treat the number of units as
26226 polynomial, asserting that it is constant if the condition has
26227 already been enforced.
26228 (vectorizable_store, vectorizable_load): Likewise.
26230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26231 Alan Hayward <alan.hayward@arm.com>
26232 David Sherwood <david.sherwood@arm.com>
26234 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
26235 of units as polynomial. Punt if we can't tell at compile time
26236 which vector contains the final result.
26238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26239 Alan Hayward <alan.hayward@arm.com>
26240 David Sherwood <david.sherwood@arm.com>
26242 * tree-vect-loop.c (vectorizable_induction): Treat the number
26243 of units as polynomial. Punt on SLP inductions. Use an integer
26244 VEC_SERIES_EXPR for variable-length integer reductions. Use a
26245 cast of such a series for variable-length floating-point
26248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26249 Alan Hayward <alan.hayward@arm.com>
26250 David Sherwood <david.sherwood@arm.com>
26252 * tree.h (build_index_vector): Declare.
26253 * tree.c (build_index_vector): New function.
26254 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
26255 of units as polynomial, forcibly converting it to a constant if
26256 vectorizable_reduction has already enforced the condition.
26257 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
26258 to create a {1,2,3,...} vector.
26259 (vectorizable_reduction): Treat the number of units as polynomial.
26260 Choose vectype_in based on the largest scalar element size rather
26261 than the smallest number of units. Enforce the restrictions
26264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26265 Alan Hayward <alan.hayward@arm.com>
26266 David Sherwood <david.sherwood@arm.com>
26268 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
26269 number of units as polynomial.
26271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26272 Alan Hayward <alan.hayward@arm.com>
26273 David Sherwood <david.sherwood@arm.com>
26275 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
26276 * target.def (autovectorize_vector_sizes): Return the vector sizes
26277 by pointer, using vector_sizes rather than a bitmask.
26278 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
26279 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
26280 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
26282 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
26283 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
26284 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
26285 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
26286 * omp-general.c (omp_max_vf): Likewise.
26287 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26288 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
26289 * tree-vect-loop.c (vect_analyze_loop): Likewise.
26290 * tree-vect-slp.c (vect_slp_bb): Likewise.
26291 * doc/tm.texi: Regenerate.
26292 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
26294 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
26295 the vector size as a poly_uint64 rather than an unsigned int.
26296 (current_vector_size): Change from an unsigned int to a poly_uint64.
26297 (get_vectype_for_scalar_type): Update accordingly.
26298 * tree.h (build_truth_vector_type): Take the size and number of
26299 units as a poly_uint64 rather than an unsigned int.
26300 (build_vector_type): Add a temporary overload that takes
26301 the number of units as a poly_uint64 rather than an unsigned int.
26302 * tree.c (make_vector_type): Likewise.
26303 (build_truth_vector_type): Take the number of units as a poly_uint64
26304 rather than an unsigned int.
26306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26307 Alan Hayward <alan.hayward@arm.com>
26308 David Sherwood <david.sherwood@arm.com>
26310 * target.def (get_mask_mode): Take the number of units and length
26311 as poly_uint64s rather than unsigned ints.
26312 * targhooks.h (default_get_mask_mode): Update accordingly.
26313 * targhooks.c (default_get_mask_mode): Likewise.
26314 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
26315 * doc/tm.texi: Regenerate.
26317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26318 Alan Hayward <alan.hayward@arm.com>
26319 David Sherwood <david.sherwood@arm.com>
26321 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
26322 * omp-general.c (omp_max_vf): Likewise.
26323 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
26324 (expand_omp_simd): Handle polynomial safelen.
26325 * omp-low.c (omplow_simd_context): Add a default constructor.
26326 (omplow_simd_context::max_vf): Change from int to poly_uint64.
26327 (lower_rec_simd_input_clauses): Update accordingly.
26328 (lower_rec_input_clauses): Likewise.
26330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26331 Alan Hayward <alan.hayward@arm.com>
26332 David Sherwood <david.sherwood@arm.com>
26334 * tree-vectorizer.h (vect_nunits_for_cost): New function.
26335 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
26336 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
26337 (vect_analyze_slp_cost): Likewise.
26338 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
26339 (vect_model_load_cost): Likewise.
26341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26342 Alan Hayward <alan.hayward@arm.com>
26343 David Sherwood <david.sherwood@arm.com>
26345 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
26346 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
26347 from an unsigned int * to a poly_uint64_pod *.
26348 (calculate_unrolling_factor): New function.
26349 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
26351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26352 Alan Hayward <alan.hayward@arm.com>
26353 David Sherwood <david.sherwood@arm.com>
26355 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
26356 from an unsigned int to a poly_uint64.
26357 (_loop_vec_info::slp_unrolling_factor): Likewise.
26358 (_loop_vec_info::vectorization_factor): Change from an int
26360 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
26361 (vect_get_num_vectors): New function.
26362 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
26363 (vect_get_num_copies): Use vect_get_num_vectors.
26364 (vect_analyze_data_ref_dependences): Change max_vf from an int *
26365 to an unsigned int *.
26366 (vect_analyze_data_refs): Change min_vf from an int * to a
26368 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26369 than an unsigned HOST_WIDE_INT.
26370 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
26371 (vect_analyze_data_ref_dependence): Change max_vf from an int *
26372 to an unsigned int *.
26373 (vect_analyze_data_ref_dependences): Likewise.
26374 (vect_compute_data_ref_alignment): Handle polynomial vf.
26375 (vect_enhance_data_refs_alignment): Likewise.
26376 (vect_prune_runtime_alias_test_list): Likewise.
26377 (vect_shift_permute_load_chain): Likewise.
26378 (vect_supportable_dr_alignment): Likewise.
26379 (dependence_distance_ge_vf): Take the vectorization factor as a
26380 poly_uint64 rather than an unsigned HOST_WIDE_INT.
26381 (vect_analyze_data_refs): Change min_vf from an int * to a
26383 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
26384 vfm1 as a poly_uint64 rather than an int. Make the same change
26385 for the returned bound_scalar.
26386 (vect_gen_vector_loop_niters): Handle polynomial vf.
26387 (vect_do_peeling): Likewise. Update call to
26388 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
26389 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
26391 * tree-vect-loop.c (vect_determine_vectorization_factor)
26392 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
26393 (vect_get_known_peeling_cost): Likewise.
26394 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
26395 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
26396 (vect_transform_loop): Likewise. Use the lowest possible VF when
26397 updating the upper bounds of the loop.
26398 (vect_min_worthwhile_factor): Make static. Return an unsigned int
26399 rather than an int.
26400 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
26401 polynomial unroll factors.
26402 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
26403 (vect_make_slp_decision): Likewise.
26404 (vect_supported_load_permutation_p): Likewise, and polynomial
26406 (vect_analyze_slp_cost): Handle polynomial vf.
26407 (vect_slp_analyze_node_operations): Likewise.
26408 (vect_slp_analyze_bb_1): Likewise.
26409 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26410 than an unsigned HOST_WIDE_INT.
26411 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
26412 (vectorizable_load): Handle polynomial vf.
26413 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
26415 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
26417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26418 Alan Hayward <alan.hayward@arm.com>
26419 David Sherwood <david.sherwood@arm.com>
26421 * match.pd: Handle bit operations involving three constants
26422 and try to fold one pair.
26424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26426 * tree-vect-loop-manip.c: Include gimple-fold.h.
26427 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
26428 niters_maybe_zero parameters. Handle other cases besides a step of 1.
26429 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
26430 Add a path that uses a step of VF instead of 1, but disable it
26432 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
26433 and niters_no_overflow parameters. Update calls to
26434 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
26435 Create a new SSA name if the latter choses to use a ste other
26436 than zero, and return it via niters_vector_mult_vf_var.
26437 * tree-vect-loop.c (vect_transform_loop): Update calls to
26438 vect_do_peeling, vect_gen_vector_loop_niters and
26439 slpeel_make_loop_iterate_ntimes.
26440 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
26441 (vect_gen_vector_loop_niters): Update declarations after above changes.
26443 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
26445 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
26446 128-bit round to integer instructions.
26447 (ceil<mode>2): Likewise.
26448 (btrunc<mode>2): Likewise.
26449 (round<mode>2): Likewise.
26451 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26453 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
26454 unaligned VSX load/store on P8/P9.
26455 (expand_block_clear): Allow the use of unaligned VSX
26456 load/store on P8/P9.
26458 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26460 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
26462 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
26463 swap associated with both a load and a store.
26465 2018-01-02 Andrew Waterman <andrew@sifive.com>
26467 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
26468 * config/riscv/riscv.md (clear_cache): Use it.
26470 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
26472 * web.c: Remove out-of-date comment.
26474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26476 * expr.c (fixup_args_size_notes): Check that any existing
26477 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
26478 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
26479 (emit_single_push_insn): ...here.
26481 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26483 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
26484 (const_vector_encoded_nelts): New function.
26485 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
26486 (const_vector_int_elt, const_vector_elt): Declare.
26487 * emit-rtl.c (const_vector_int_elt_1): New function.
26488 (const_vector_elt): Likewise.
26489 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
26490 of CONST_VECTOR_ELT.
26492 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26494 * expr.c: Include rtx-vector-builder.h.
26495 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
26496 directly on the tree encoding.
26497 (const_vector_from_tree): Likewise.
26498 * optabs.c: Include rtx-vector-builder.h.
26499 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
26500 sequence of "u" values.
26501 * vec-perm-indices.c: Include rtx-vector-builder.h.
26502 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
26503 directly on the vec_perm_indices encoding.
26505 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26507 * doc/rtl.texi (const_vector): Describe new encoding scheme.
26508 * Makefile.in (OBJS): Add rtx-vector-builder.o.
26509 * rtx-vector-builder.h: New file.
26510 * rtx-vector-builder.c: Likewise.
26511 * rtl.h (rtx_def::u2): Add a const_vector field.
26512 (CONST_VECTOR_NPATTERNS): New macro.
26513 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
26514 (CONST_VECTOR_DUPLICATE_P): Likewise.
26515 (CONST_VECTOR_STEPPED_P): Likewise.
26516 (CONST_VECTOR_ENCODED_ELT): Likewise.
26517 (const_vec_duplicate_p): Check for a duplicated vector encoding.
26518 (unwrap_const_vec_duplicate): Likewise.
26519 (const_vec_series_p): Check for a non-duplicated vector encoding.
26520 Say that the function only returns true for integer vectors.
26521 * emit-rtl.c: Include rtx-vector-builder.h.
26522 (gen_const_vec_duplicate_1): Delete.
26523 (gen_const_vector): Call gen_const_vec_duplicate instead of
26524 gen_const_vec_duplicate_1.
26525 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
26526 (gen_const_vec_duplicate): Use rtx_vector_builder.
26527 (gen_const_vec_series): Likewise.
26528 (gen_rtx_CONST_VECTOR): Likewise.
26529 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
26530 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26531 Build a new vector rather than modifying a CONST_VECTOR in-place.
26532 (handle_special_swappables): Update call accordingly.
26533 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
26534 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26535 Build a new vector rather than modifying a CONST_VECTOR in-place.
26536 (handle_special_swappables): Update call accordingly.
26538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26540 * simplify-rtx.c (simplify_const_binary_operation): Use
26541 CONST_VECTOR_ELT instead of XVECEXP.
26543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26545 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
26546 the selector elements to be different from the data elements
26547 if the selector is a VECTOR_CST.
26548 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
26549 ssizetype for the selector.
26551 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26553 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
26554 before testing each element individually.
26555 * tree-vect-generic.c (lower_vec_perm): Likewise.
26557 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26559 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
26560 * selftest-run-tests.c (selftest::run_tests): Call it.
26561 * vector-builder.h (vector_builder::operator ==): New function.
26562 (vector_builder::operator !=): Likewise.
26563 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
26564 (vec_perm_indices::all_from_input_p): New function.
26565 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26566 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
26567 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
26568 instead of reading the VECTOR_CST directly. Detect whether both
26569 vector inputs are the same before constructing the vec_perm_indices,
26570 and update the number of inputs argument accordingly. Use the
26571 utility functions added above. Only construct sel2 if we need to.
26573 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26575 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
26576 the broadcast of the low byte.
26577 (expand_mult_highpart): Use an explicit encoding for the permutes.
26578 * optabs-query.c (can_mult_highpart_p): Likewise.
26579 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
26580 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26581 (vectorizable_bswap): Likewise.
26582 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
26583 explicit encoding for the power-of-2 permutes.
26584 (vect_permute_store_chain): Likewise.
26585 (vect_grouped_load_supported): Likewise.
26586 (vect_permute_load_chain): Likewise.
26588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26590 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
26591 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
26592 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
26593 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26594 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
26595 (vect_gen_perm_mask_any): Likewise.
26597 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26599 * int-vector-builder.h: New file.
26600 * vec-perm-indices.h: Include int-vector-builder.h.
26601 (vec_perm_indices): Redefine as an int_vector_builder.
26602 (auto_vec_perm_indices): Delete.
26603 (vec_perm_builder): Redefine as a stand-alone class.
26604 (vec_perm_indices::vec_perm_indices): New function.
26605 (vec_perm_indices::clamp): Likewise.
26606 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
26607 (vec_perm_indices::new_vector): New function.
26608 (vec_perm_indices::new_expanded_vector): Update for new
26609 vec_perm_indices class.
26610 (vec_perm_indices::rotate_inputs): New function.
26611 (vec_perm_indices::all_in_range_p): Operate directly on the
26612 encoded form, without computing elided elements.
26613 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
26614 encoding. Update for new vec_perm_indices class.
26615 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
26616 the given vec_perm_builder.
26617 (expand_vec_perm_var): Update vec_perm_builder constructor.
26618 (expand_mult_highpart): Use vec_perm_builder instead of
26619 auto_vec_perm_indices.
26620 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
26621 vec_perm_indices instead of auto_vec_perm_indices. Use a single
26622 or double series encoding as appropriate.
26623 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
26624 vec_perm_indices instead of auto_vec_perm_indices.
26625 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26626 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26627 (vect_permute_store_chain): Likewise.
26628 (vect_grouped_load_supported): Likewise.
26629 (vect_permute_load_chain): Likewise.
26630 (vect_shift_permute_load_chain): Likewise.
26631 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26632 (vect_transform_slp_perm_load): Likewise.
26633 (vect_schedule_slp_instance): Likewise.
26634 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26635 (vectorizable_mask_load_store): Likewise.
26636 (vectorizable_bswap): Likewise.
26637 (vectorizable_store): Likewise.
26638 (vectorizable_load): Likewise.
26639 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
26640 vec_perm_indices instead of auto_vec_perm_indices. Use
26641 tree_to_vec_perm_builder to read the vector from a tree.
26642 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
26643 vec_perm_builder instead of a vec_perm_indices.
26644 (have_whole_vector_shift): Use vec_perm_builder and
26645 vec_perm_indices instead of auto_vec_perm_indices. Leave the
26646 truncation to calc_vec_perm_mask_for_shift.
26647 (vect_create_epilog_for_reduction): Likewise.
26648 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
26649 from auto_vec_perm_indices to vec_perm_indices.
26650 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26651 instead of changing individual elements.
26652 (aarch64_vectorize_vec_perm_const): Use new_vector to install
26653 the vector in d.perm.
26654 * config/arm/arm.c (expand_vec_perm_d::perm): Change
26655 from auto_vec_perm_indices to vec_perm_indices.
26656 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26657 instead of changing individual elements.
26658 (arm_vectorize_vec_perm_const): Use new_vector to install
26659 the vector in d.perm.
26660 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
26661 Update vec_perm_builder constructor.
26662 (rs6000_expand_interleave): Likewise.
26663 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
26664 (rs6000_expand_interleave): Likewise.
26666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26668 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
26669 to qimode could truncate the indices.
26670 * optabs.c (expand_vec_perm_var): Likewise.
26672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26674 * Makefile.in (OBJS): Add vec-perm-indices.o.
26675 * vec-perm-indices.h: New file.
26676 * vec-perm-indices.c: Likewise.
26677 * target.h (vec_perm_indices): Replace with a forward class
26679 (auto_vec_perm_indices): Move to vec-perm-indices.h.
26680 * optabs.h: Include vec-perm-indices.h.
26681 (expand_vec_perm): Delete.
26682 (selector_fits_mode_p, expand_vec_perm_var): Declare.
26683 (expand_vec_perm_const): Declare.
26684 * target.def (vec_perm_const_ok): Replace with...
26685 (vec_perm_const): ...this new hook.
26686 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
26687 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
26688 * doc/tm.texi: Regenerate.
26689 * optabs.def (vec_perm_const): Delete.
26690 * doc/md.texi (vec_perm_const): Likewise.
26691 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
26692 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
26693 expand_vec_perm for constant permutation vectors. Assert that
26694 the mode of variable permutation vectors is the integer equivalent
26695 of the mode that is being permuted.
26696 * optabs-query.h (selector_fits_mode_p): Declare.
26697 * optabs-query.c: Include vec-perm-indices.h.
26698 (selector_fits_mode_p): New function.
26699 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
26700 is defined, instead of checking whether the vec_perm_const_optab
26701 exists. Use targetm.vectorize.vec_perm_const instead of
26702 targetm.vectorize.vec_perm_const_ok. Check whether the indices
26703 fit in the vector mode before using a variable permute.
26704 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
26705 vec_perm_indices instead of an rtx.
26706 (expand_vec_perm): Replace with...
26707 (expand_vec_perm_const): ...this new function. Take the selector
26708 as a vec_perm_indices rather than an rtx. Also take the mode of
26709 the selector. Update call to shift_amt_for_vec_perm_mask.
26710 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
26711 Use vec_perm_indices::new_expanded_vector to expand the original
26712 selector into bytes. Check whether the indices fit in the vector
26713 mode before using a variable permute.
26714 (expand_vec_perm_var): Make global.
26715 (expand_mult_highpart): Use expand_vec_perm_const.
26716 * fold-const.c: Includes vec-perm-indices.h.
26717 * tree-ssa-forwprop.c: Likewise.
26718 * tree-vect-data-refs.c: Likewise.
26719 * tree-vect-generic.c: Likewise.
26720 * tree-vect-loop.c: Likewise.
26721 * tree-vect-slp.c: Likewise.
26722 * tree-vect-stmts.c: Likewise.
26723 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
26725 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
26726 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
26727 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
26728 (aarch64_vectorize_vec_perm_const): ...this new function.
26729 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26730 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26731 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
26732 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
26733 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26734 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26735 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
26737 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
26738 check for NEON modes.
26739 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
26740 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
26741 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
26742 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
26744 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
26745 the old VEC_PERM_CONST conditions.
26746 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
26747 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
26748 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
26749 (ia64_vectorize_vec_perm_const_ok): Merge into...
26750 (ia64_vectorize_vec_perm_const): ...this new function.
26751 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
26752 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
26753 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
26754 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
26755 * config/mips/mips.c (mips_expand_vec_perm_const)
26756 (mips_vectorize_vec_perm_const_ok): Merge into...
26757 (mips_vectorize_vec_perm_const): ...this new function.
26758 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
26759 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
26760 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
26761 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
26762 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
26763 (rs6000_expand_vec_perm_const): Delete.
26764 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26766 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26767 (altivec_expand_vec_perm_const_le): Take each operand individually.
26768 Operate on constant selectors rather than rtxes.
26769 (altivec_expand_vec_perm_const): Likewise. Update call to
26770 altivec_expand_vec_perm_const_le.
26771 (rs6000_expand_vec_perm_const): Delete.
26772 (rs6000_vectorize_vec_perm_const_ok): Delete.
26773 (rs6000_vectorize_vec_perm_const): New function.
26774 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26775 an element count and rtx array.
26776 (rs6000_expand_extract_even): Update call accordingly.
26777 (rs6000_expand_interleave): Likewise.
26778 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26779 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26780 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26781 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26782 (rs6000_expand_vec_perm_const): Delete.
26783 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26784 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26785 (altivec_expand_vec_perm_const_le): Take each operand individually.
26786 Operate on constant selectors rather than rtxes.
26787 (altivec_expand_vec_perm_const): Likewise. Update call to
26788 altivec_expand_vec_perm_const_le.
26789 (rs6000_expand_vec_perm_const): Delete.
26790 (rs6000_vectorize_vec_perm_const_ok): Delete.
26791 (rs6000_vectorize_vec_perm_const): New function. Remove stray
26792 reference to the SPE evmerge intructions.
26793 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26794 an element count and rtx array.
26795 (rs6000_expand_extract_even): Update call accordingly.
26796 (rs6000_expand_interleave): Likewise.
26797 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
26798 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
26800 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26802 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26804 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
26805 vector mode and that that mode matches the mode of the data
26807 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
26808 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
26809 directly using expand_vec_perm_1 when forcing selectors into
26811 (expand_vec_perm_var): New function, split out from expand_vec_perm.
26813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26815 * optabs-query.h (can_vec_perm_p): Delete.
26816 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
26817 * optabs-query.c (can_vec_perm_p): Split into...
26818 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
26819 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
26820 particular selector is valid.
26821 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26822 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26823 (vect_grouped_load_supported): Likewise.
26824 (vect_shift_permute_load_chain): Likewise.
26825 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26826 (vect_transform_slp_perm_load): Likewise.
26827 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26828 (vectorizable_bswap): Likewise.
26829 (vect_gen_perm_mask_checked): Likewise.
26830 * fold-const.c (fold_ternary_loc): Likewise. Don't take
26831 implementations of variable permutation vectors into account
26832 when deciding which selector to use.
26833 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
26834 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
26835 with a false third argument.
26836 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
26837 to test whether the constant selector is valid and can_vec_perm_var_p
26838 to test whether a variable selector is valid.
26840 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26842 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
26843 * optabs-query.c (can_vec_perm_p): Likewise.
26844 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
26845 instead of vec_perm_indices.
26846 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
26847 (vect_gen_perm_mask_checked): Likewise,
26848 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
26849 (vect_gen_perm_mask_checked): Likewise,
26851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26853 * optabs-query.h (qimode_for_vec_perm): Declare.
26854 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
26855 (qimode_for_vec_perm): ...this new function.
26856 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
26858 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26860 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
26861 does not have a conditional at the top.
26863 2018-01-02 Richard Biener <rguenther@suse.de>
26865 * ipa-inline.c (big_speedup_p): Fix expression.
26867 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26870 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
26873 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
26877 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
26878 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
26879 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
26880 cond_taken_branch_cost 3->4.
26882 2018-01-01 Jakub Jelinek <jakub@redhat.com>
26884 PR tree-optimization/83581
26885 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
26886 TODO_cleanup_cfg if any changes have been made.
26888 PR middle-end/83608
26889 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
26890 convert_modes if target mode has the right side, but different mode
26893 PR middle-end/83609
26894 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
26895 last argument when extracting from CONCAT. If either from_real or
26896 from_imag is NULL, use expansion through memory. If result is not
26897 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
26898 the parts directly to inner mode, if even that fails, use expansion
26901 PR middle-end/83623
26902 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
26903 check for bswap in mode rather than HImode and use that in expand_unop
26906 Copyright (C) 2018 Free Software Foundation, Inc.
26908 Copying and distribution of this file, with or without modification,
26909 are permitted in any medium without royalty provided the copyright
26910 notice and this notice are preserved.