1 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3 * doc/invoke.texi (-flinker-output): Document
5 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7 * passes.c (ipa_write_summaries): Only modify statements if body
9 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10 incrementally linking.
11 (ipa_passes): Likewise.
12 * lto-cgraph.c (lto_output_node): When incrementally linking do not
13 pass down resolution info.
14 * common.opt (flag_incremental_link): Update info.
15 * gcc.c (plugin specs): Turn flinker-output=* to
16 -plugin-opt=-linker-output-known
17 * toplev.c (compile_file): Also cut compilation when doing incremental
19 * flag-types. (enum lto_partition_model): Add
20 LTO_LINKER_OUTPUT_NOLTOREL.
21 (invoke.texi): Add -flinker-output docs.
22 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
23 link same way as WPA; do not stream in dead initializers.
25 * dwarf2out.c (dwarf2out_die_ref_for_decl,
26 darf2out_register_external_decl): Support incremental link.
28 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
30 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
32 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
34 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
35 it down to simple_object_copy_lto_debug_sections.
36 (run_gcc): Determine incremental LTO link time and configure
37 lto1 into non-wpa mode, disable renaming of debug sections.
39 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
41 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
42 descriptions of various incorrectly documented functions.
44 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
47 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
48 address check not strict.
50 2018-05-30 Richard Biener <rguenther@suse.de>
52 PR tree-optimization/85964
53 * tracer.c (better_p): Drop initialized count check, we only
54 call the function with initialized counts now.
55 (find_best_successor): Do find a best edge if one
56 has uninitialized count.
57 (find_best_predecessor): Likewise. Do BB frequency check only
58 if count is initialized.
60 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
62 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
63 (aarch64_ldrstr_offset_compare): New.
64 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
66 (aarch64_gen_adjusted_ldpstp): Likewise.
68 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
70 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
71 Check for subset of GENERAL_REGS and FP_REGS.
72 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
75 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
77 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
78 and wi::to_poly_offset. Add the current offset and then check
79 whether the sum fits, rather than using an unchecked addition of
80 a checked term. Check for a shwi rather than a uhwi.
81 * expr.c (get_bit_range): Use tree_to_poly_uint64.
82 (store_constructor): Use poly_int_tree_p.
83 (expand_expr_real_1): Likewise.
84 * function.c (assign_temp): Likewise.
85 * fold-const.c (const_binop): Use poly_int_tree_p and
87 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
89 * ipa-icf-gimple.c (func_checker::compare_operand): Use
90 to_poly_offset for MEM offsets.
91 * ipa-icf.c (sem_variable::equals): Likewise.
92 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
93 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
94 wi::to_poly_offset for BIT_FIELD_REF offsets.
95 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
97 * var-tracking.c (emit_note_insn_var_location): Use
100 2018-05-29 Jim Wilson <jimw@sifive.com>
102 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
104 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
107 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
108 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
110 (sse4_1_round<mode>2): Use nonimmediate_operand
111 for operand 1 predicate.
113 2018-05-29 Martin Sebor <msebor@redhat.com>
114 Richard Biener <rguenther@suse.de>
117 * calls.c (get_size_range): Call determine_value_range instead
119 * tree-vrp.h (determine_value_range): Declared new function.
120 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
122 2018-05-29 Richard Biener <rguenther@suse.de>
124 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
125 sure to use non-pattern stmts for get_earlier_stmt arguments.
126 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
127 called on pattern stmts.
128 (get_later_stmt): Likewise.
130 2018-05-29 Martin Liska <mliska@suse.cz>
132 PR gcov-profile/85759
133 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
136 2018-05-29 Jakub Jelinek <jakub@redhat.com>
138 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
140 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
144 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
145 VEC_PACK_FLOAT_EXPR): New tree codes.
146 * tree-pretty-print.c (op_code_prio): Handle
147 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
148 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
149 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
150 * tree-inline.c (estimate_operator_cost): Likewise.
151 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
152 * fold-const.c (const_binop): Likewise.
153 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
154 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
155 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
156 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
157 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
158 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
159 * expr.c (expand_expr_real_2): Likewise.
160 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
161 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
162 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
164 * optabs.c (expand_widen_pattern_expr): For
165 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
166 sign from result type rather than operand's type.
167 (expand_binop_directly): For vec_packu_float_optab and
168 vec_packs_float_optab allow result type to be different from operand's
170 * optabs-tree.c (optab_for_tree_code): Handle
171 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
172 VEC_PACK_FLOAT_EXPR. Formatting fixes.
173 * tree-vect-generic.c (expand_vector_operations_1): Handle
174 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
176 * tree-vect-stmts.c (supportable_widening_operation): Handle
178 (supportable_narrowing_operation): Handle FLOAT_EXPR.
179 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
180 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
181 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
182 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
184 (vec_pack<floatprefix>_float_<mode>): New expander.
185 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
187 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
188 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
189 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
190 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
191 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
193 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
194 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
195 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
196 VEC_PACK_FLOAT_EXPR): Document.
198 2018-05-29 Richard Biener <rguenther@suse.de>
200 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
202 (stmt_vec_info_vec): Make pointer.
203 (init_stmt_vec_info_vec): Remove.
204 (free_stmt_vec_info_vec): Likewise.
205 (set_stmt_vec_info_vec): New function.
206 (free_stmt_vec_infos): Likewise.
207 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
208 (set_vinfo_for_stmt): Likewise.
209 (get_earlier_stmt): Likewise.
210 (get_later_stmt): Likewise.
211 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
212 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
213 (vec_info::~vec_info): Free stmt_vec_infos.
214 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
215 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
216 (pass_slp_vectorize::execute): Likewise.
217 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
218 (free_stmt_vec_info_vec): Likewise.
219 (set_stmt_vec_info_vec): New function.
220 (free_stmt_vec_infos): Likewise.
221 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
222 the global stmt_vec_info_vec.
223 * tree-parloops.c (gather_scalar_reductions): Use
224 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
227 2018-05-29 Richard Biener <rguenther@suse.de>
229 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
231 2018-05-29 Martin Liska <mliska@suse.cz>
232 David Malcolm <dmalcolm@redhat.com>
234 * vec.c (test_reverse): New.
235 (vec_c_tests): Add new test.
236 * vec.h (vl_ptr>::reverse): New function.
238 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
240 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
242 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
245 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
247 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
249 2018-05-28 Richard Biener <rguenther@suse.de>
251 PR tree-optimization/85933
252 * tree-vect-data-refs.c (vect_record_base_alignments): Only
253 look at stmts marked as vectorizable.
255 2018-05-28 Richard Biener <rguenther@suse.de>
257 PR tree-optimization/85934
258 * tree-vect-generic.c (expand_vector_operations_1): Hoist
259 vector boolean check before scalar optimization.
261 2018-05-28 Jakub Jelinek <jakub@redhat.com>
263 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
266 2018-05-28 Mark Wielaard <mark@klomp.org>
268 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
269 if it is an expression containing a minus sign.
271 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
273 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
275 2018-05-27 Paul Koning <ni1d@arrl.net>
277 * config/pdp11/pdp11.md (truncsihi2): Remove.
279 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
280 Chung-Ju Wu <jasonwucj@gmail.com>
282 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
284 (unaligned_store_dw): Ditto.
285 * config/nds32/nds32-memory-manipulation.c
286 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
287 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
288 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
289 (emit_setmem_word_loop): Rename to ...
290 (emit_setmem_doubleword_loop): ... this.
291 (nds32_gen_dup_4_byte_to_word_value): New function.
292 (nds32_gen_dup_8_byte_to_double_word_value): New function.
293 (nds32_expand_setmem_loop): Refine implementation.
294 (nds32_expand_setmem_loop_v3m): Ditto.
295 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
298 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
300 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
302 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
304 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
305 (nds32_init_machine_status): Initialize machine->attr_naked_p and
306 machine->attr_no_prologue_p.
307 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
308 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
309 (nds32_expand_epilogue): Consider attr_naked_p.
310 (nds32_expand_epilogue_v3pop): Likewise.
311 (nds32_can_use_return_insn): Likewise.
312 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
313 attr_no_prologue_p fields.
314 * config/nds32/nds32.opt (mret-in-naked-func): New option.
316 2018-05-27 Jakub Jelinek <jakub@redhat.com>
319 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
322 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
324 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
326 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
328 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
330 (*<floatsuffix>floatv2div2sf2): Rename to ...
331 (*float<floatunssuffix>v2div2sf2): ... this.
332 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
333 (float<floatunssuffix>v2div2sf2_mask): ... this.
334 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
335 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
336 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
338 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
340 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
342 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
344 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
346 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
348 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
349 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
350 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
351 gen_ufix_truncv8dfv8si2.
352 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
353 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
354 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
355 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
356 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
357 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
358 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
359 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
361 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
365 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
367 2018-05-25 Jim Wilson <jimw@sifive.com>
369 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
370 * config/riscv/riscv.c (struct machine_function): Add
371 interrupt_handler_p and attribute_checked_p fields.
372 (riscv_attribute_table): Add interrupt.
373 (riscv_interrupt_type_p): New.
374 (riscv_save_reg_p): Save extra regs for interrupt handler.
375 (riscv_use_save_libcall): Return false for interrupt handler.
376 (riscv_first_stack_step): Add forward declaration.
377 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
378 for interrupt handler with large frame. Use it for saved reg list.
379 (riscv_expand_prologue): Move flag_stack_usage_info support to
380 eliminate duplication.
381 (riscv_expand_epilogue): Generate mret for interrupt handler.
382 (riscv_epilogue_uses): New.
383 (riscv_can_use_return_insn): Return false for interrupt handler.
384 (riscv_function_ok_for_sibcall): Likewise.
385 (riscv_set_current_function): Add interrupt handler support.
386 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
387 * config/riscv/riscv.md (UNSPECV_MRET): New.
389 (riscv_frflags, riscv_fsflags): Use tab after opcode.
391 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
393 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
395 PR tree-optimization/85712
396 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
397 this candidate has already been replaced in-situ by a copy.
399 2018-05-25 Jason Merrill <jason@redhat.com>
401 PR c++/80485 - inline function non-zero address.
402 * symtab.c (nonzero_address): Check DECL_COMDAT.
404 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
407 * config/alpha/alpha.md (ashlsi3): New insn pattern.
408 (*ashlsi_se): Rename from *ashldi_se. Define as sign
409 extension of SImode operation. Use const123_operand predicate.
411 (*saddl_se_1): Ditto.
413 (*ssubl_se_1): Ditto.
414 * config/alpha/predicates.md (const123_operand): New predicate.
415 * config/alpha/constraints.md (P): Use IN_RANGE.
417 2018-05-25 Richard Biener <rguenther@suse.de>
419 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
421 (ref_maybe_used_by_stmt_p): Likewise.
422 (stmt_may_clobber_ref_p): Likewise.
423 (stmt_may_clobber_ref_p_1): Likewise.
424 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
426 (ref_maybe_used_by_stmt_p): Likewise.
427 (stmt_may_clobber_ref_p): Likewise.
428 (stmt_may_clobber_ref_p_1): Likewise.
429 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
430 the alias oracle to disambiguate DRs with stmts DR analysis
432 (vect_analyze_data_refs): Do not give up on not analyzable
433 DRs for BB vectorization. Remove code truncating the dataref
436 2018-05-25 Jakub Jelinek <jakub@redhat.com>
439 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
440 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
441 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
443 2018-05-25 Richard Biener <rguenther@suse.de>
445 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
446 function, combining stmt data ref gathering and fatal analysis
448 (vect_analyze_data_refs): Remove now redudnant code and simplify.
449 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
450 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
451 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
452 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
454 2018-05-25 Bin Cheng <bin.cheng@arm.com>
456 PR tree-optimization/85720
457 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
458 SCC if all partitions are builtins.
459 (version_loop_by_alias_check): New parameter. Generate cancelable
460 runtime alias check if all partitions are builtins.
461 (distribute_loop): Update call to above function.
463 2018-05-25 Bin Cheng <bin.cheng@arm.com>
465 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
466 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
467 (parm_default_def_partition_arg): Ditto.
468 (set_parm_default_def_partition): Ditto.
469 (get_parm_default_def_partitions): Ditto and make it static.
470 (get_undefined_value_partitions): Ditto and make it static.
471 (remove_ssa_form): Refactor call to init_var_map here.
472 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
473 computation for loop region.
474 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
475 (register_default_def): Delete.
476 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
477 (parm_default_def_partition_arg): Ditto.
478 (set_parm_default_def_partition): Ditto.
479 (get_parm_default_def_partitions): Ditto and make it static.
480 (get_undefined_value_partitions): Ditto and make it static.
481 (coalesce_with_default, coalesce_with_default): Update comment.
482 (create_coalesce_list_for_region): New func factored out from
483 create_outofssa_var_map.
484 (populate_coalesce_list_for_outofssa): New func factored out from
485 create_outofssa_var_map and coalesce_ssa_name.
486 (create_outofssa_var_map): Delete.
487 (coalesce_ssa_name): Refactor to support live range computation.
488 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
489 (get_parm_default_def_partitions): Delete.
490 (get_undefined_value_partitions): Ditto.
491 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
492 computation for loop region.
493 (new_tree_live_info, loe_visit_block): Ditto.
494 (live_worklist, set_var_live_on_entry): Ditto.
495 (calculate_live_on_exit, verify_live_on_entry): Ditto.
496 * tree-ssa-live.h (struct _var_map): New fields.
497 (init_var_map): Change decl.
498 (region_contains_p): New.
500 2018-05-25 Bin Cheng <bin.cheng@arm.com>
502 * tree-ssa-live.h (live_merge_and_clear): Delete.
504 2018-05-25 Richard Biener <rguenther@suse.de>
507 * tree-dump.c (dequeue_and_dump): Remove access to removed
508 operand 2 of a SWITCH_EXPR.
510 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
512 * doc/sourcebuild.texi (vect_double_cond_arith): Include
513 multiplication and division.
514 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
515 (cond_udiv@var{m}, cond_umod@var{m}): Document.
516 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
517 (cond_udiv_optab, cond_umod_optab): New optabs.
518 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
519 (IFN_COND_RDIV): New internal functions.
520 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
521 TRUNC_MOD_EXPR and RDIV_EXPR.
522 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
523 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
525 (SVE_INT_BINARY): Include mult.
526 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
527 (optab, sve_int_op): Handle mult.
528 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
530 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
531 for SVE_INT_BINARY_SD.
533 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
535 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
536 (optab, sve_int_op): Handle div and udiv.
537 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
538 for SVE_INT_BINARY_SD.
539 (*<optab><mode>3): New insn for the same.
541 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
543 * tree-vect-patterns.c: Include predict.h.
544 (vect_recog_divmod_pattern): Restrict check for division support
545 to when optimizing for size.
547 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
549 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
550 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
551 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
552 (gimple_match_op::set_op): Likewise.
553 (gimple_resimplify4): Declare.
554 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
555 (expr::gen_transform): Likewise.
556 (decision_tree::gen): Generate a simplification routine for 4 operands.
557 * gimple-match-head.c (gimple_simplify): Add an overload for
558 4 operands. In the top-level function, handle up to 4 call
559 arguments and call gimple_resimplify4.
560 (gimple_resimplify4): New function.
561 (build_call_internal): Pass a fourth operand.
562 (maybe_push_to_seq): Likewise.
563 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
564 Fold VEC_COND_EXPRs of an operation and a default value into
565 an IFN_COND_* function if possible.
566 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
568 (SVE_COND_FP_BINARY): Include them.
569 (optab, sve_fp_op): Handle them.
570 (SVE_INT_BINARY_REV): New code iterator.
571 (SVE_COND_FP_BINARY_REV): New int iterator.
572 (commutative): New int attribute.
573 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
575 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
577 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
578 (*cond_<optab><mode>): New patterns for reversed operands.
580 2018-05-25 Richard Biener <rguenther@suse.de>
582 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
583 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
584 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
585 (STMT_VINFO_GROUPED_ACCESS): Adjust.
586 * tree-vect-data-refs.c (everywhere): Adjust users.
587 * tree-vect-loop.c (everywhere): Likewise.
588 * tree-vect-slp.c (everywhere): Likewise.
589 * tree-vect-stmts.c (everywhere): Likewise.
590 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
592 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
594 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
596 (gcc_cv_as_section_exclude): ... this.
597 Try Solaris as #exclude syntax.
598 * configure: Regenerate.
599 * config.in: Regenerate.
600 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
602 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
603 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
605 * varasm.c (default_elf_asm_named_section): Don't check if
606 HAVE_GAS_SECTION_EXCLUDE is defined.
608 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
610 * doc/md.texi: Update the documentation of the cond_* optabs
611 to mention the new final operand. Fix GET_MODE_NUNITS call.
612 Describe the scalar case too.
613 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
614 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
616 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
617 (get_conditional_internal_fn): Update comment.
618 * tree-vect-loop.c (vectorizable_reduction): Pass the original
619 accumulator value as a final argument to conditional functions.
620 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
621 a define_expand and add an "else" operand. Assert for now that
622 the else operand is equal to operand 2. Use SVE_INT_BINARY and
623 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
624 (*cond_<optab><mode>): New patterns.
625 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
626 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
627 (UNSPEC_COND_EOR): Delete.
628 (optab): Remove associated mappings.
629 (SVE_INT_BINARY): New code iterator.
630 (sve_int_op): Remove int attribute and add "minus" to the code
632 (SVE_COND_INT_OP): Delete.
633 (SVE_COND_FP_OP): Rename to...
634 (SVE_COND_FP_BINARY): ...this.
636 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
638 * optabs.c (can_reuse_operands_p): New function.
639 (maybe_legitimize_operands): Try to reuse the results for
642 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
644 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
645 Add {q} suffix to insn mnemonic.
647 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
649 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
650 (msp430_warn_func_return): New.
652 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
654 * fold-const.c (tree_nonzero_bits): New function.
655 * fold-const.h (tree_nonzero_bits): Likewise.
656 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
657 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
659 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
663 * varasm.c (assemble_alias): Check ifunc_resolver only on
666 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
669 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
670 when memory input operand is handled.
672 2018-05-24 Luis Machado <luis.machado@linaro.org>
674 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
676 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
678 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
680 * match.pd: Delay FMA folds until after vectorization.
682 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
685 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
686 address check not strict.
688 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
690 * gimple-match.h (gimple_match_op): New class.
691 (mprts_hook): Replace parameters with a gimple_match_op *.
692 (maybe_build_generic_op): Likewise.
693 (gimple_simplified_result_is_gimple_val): Replace parameters with
694 a const gimple_match_op *.
695 (gimple_simplify): Replace code_helper * and tree * parameters with
696 a gimple_match_op * parameter.
697 (gimple_resimplify1): Replace code_helper *, tree and tree *
698 parameters with a gimple_match_op * parameter.
699 (gimple_resimplify2): Likewise.
700 (gimple_resimplify3): Likewise.
701 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
702 parameters with a gimple_match_op * parameter.
703 * gimple-match-head.c (gimple_simplify): Change prototypes of
704 auto-generated functions to take a gimple_match_op * instead of
705 separate code_helper * and tree * parameters. Make the same
706 change in the top-level overload and update calls to the
707 gimple_resimplify routines. Update calls to the auto-generated
708 functions and to maybe_push_res_to_seq in the publicly-facing
709 operation-specific gimple_simplify overloads.
710 (gimple_match_op::MAX_NUM_OPS): Define.
711 (gimple_resimplify1): Replace rcode and ops with a single res_op
712 parameter. Update call to gimple_simplify.
713 (gimple_resimplify2): Likewise.
714 (gimple_resimplify3): Likewise.
715 (mprts_hook): Replace parameters with a gimple_match_op *.
716 (maybe_build_generic_op): Likewise.
717 (build_call_internal): Replace type, nargs and ops with
719 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
720 with a single gimple_match_op *. Update calls to mprts_hook,
721 build_call_internal and gimple_simplified_result_is_gimple_val.
722 Factor out code that is common to the tree_code and combined_fn cases.
723 * genmatch.c (expr::gen_transform): Replace tem_code and
724 tem_ops with a gimple_match_op called tem_op. Update calls
725 to the gimple_resimplify functions and maybe_push_res_to_seq.
726 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
727 res_ops. Update call to the gimple_resimplify functions.
728 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
729 (decision_tree::gen): Make the functions take a gimple_match_op *
730 called res_op instead of separate res_code and res_ops parameters.
731 Update call accordingly.
732 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
733 and ops with a single res_op parameter. Update calls to
734 maybe_build_generic_op and maybe_push_res_to_seq.
735 (fold_stmt_1): Update calls to gimple_simplify and
736 replace_stmt_with_simplification.
737 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
738 and gimple_simplified_result_is_gimple_val.
739 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
741 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
742 with a gimple_match_op *.
743 (vn_nary_build_or_lookup): Likewise. Update call to
744 vn_nary_build_or_lookup_1.
745 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
746 gimple_match_op *. Update calls to the gimple_resimplify routines
747 and to gimple_simplified_result_is_gimple_val.
748 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
749 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
750 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
751 (visit_nary_op): Likewise.
752 (visit_reference_op_load): Likewise.
754 2018-05-23 Luis Machado <luis.machado@linaro.org>
756 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
757 modifier for printing the step amount.
759 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
762 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
765 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
767 * doc/sourcebuild.texi (Endianness): New subsubsection.
769 2018-05-23 Luis Machado <luis.machado@linaro.org>
771 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
772 <prefetch_dynamic_strides>: New const bool field.
773 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
774 prefetch_dynamic_strides.
775 (exynosm1_prefetch_tune): Likewise.
776 (thunderxt88_prefetch_tune): Likewise.
777 (thunderx_prefetch_tune): Likewise.
778 (thunderx2t99_prefetch_tune): Likewise.
779 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
781 (aarch64_override_options_internal): Update to set
782 PARAM_PREFETCH_DYNAMIC_STRIDES.
783 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
784 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
785 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
786 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
787 prefetch-dynamic-strides setting.
789 2018-05-23 Luis Machado <luis.machado@linaro.org>
791 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
792 <minimum_stride>: New const int field.
793 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
794 minimum_stride field defaulting to -1.
795 (exynosm1_prefetch_tune): Likewise.
796 (thunderxt88_prefetch_tune): Likewise.
797 (thunderx_prefetch_tune): Likewise.
798 (thunderx2t99_prefetch_tune): Likewise.
799 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
800 <default_opt_level>: Set to 3.
801 (aarch64_override_options_internal): Update to set
802 PARAM_PREFETCH_MINIMUM_STRIDE.
803 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
804 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
805 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
806 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
807 stride is constant and is below the minimum stride threshold.
809 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
811 * config/arm/arm-cpus.in (mode26): Delete.
812 (armv4): Delete mode26 reference.
813 * config/arm/arm.c (arm_configure_build_target): Delete use of
816 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
818 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
820 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
821 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
823 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
824 (floatunsdidf2): Ditto.
826 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
828 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
829 (fixuns_trunc<mode>si2_avx512f): Ditto.
830 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
831 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
832 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
834 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
836 PR rtl-optimization/79985
837 * df-scan.c (df_insn_refs_collect): Remove special case for
838 global registers and asm statements.
840 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
842 * extend.texi (Global Register Variables): Rewrite the bullet list.
843 Note that the register is available for allocation. Note that access
844 via inline asm must use constraints. Add note about async-signal
845 handlers. Remove paragraph about automagic register selection.
847 2018-05-23 Richard Biener <rguenther@suse.de>
849 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
850 of fixed offset from memset VN.
852 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
854 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
856 (alloc_cand_and_find_basis): Initialize first_interp field.
857 (slsr_process_mul): Modify first_interp field.
858 (slsr_process_add): Likewise.
859 (slsr_process_cast): Modify first_interp field for each new
861 (slsr_process_copy): Likewise.
862 (dump_candidate): Dump first_interp field.
863 (replace_mult_candidate): Process all interpretations, not just
865 (replace_rhs_if_not_dup): Likewise.
866 (replace_one_candidate): Likewise.
868 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
870 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
872 (aarch64_needs_frame_chain): New function.
873 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
875 2018-05-23 Sudakshina Das <sudi.das@arm.com>
878 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
879 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
880 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
881 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
882 as true for strict-align.
883 (aarch64_can_inline_p): Perform checks even when callee has no
884 attributes to check for strict alignment.
885 * doc/extend.texi (AArch64 Function Attributes): Document
887 * doc/invoke.texi: (AArch64 Options): Likewise.
889 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
891 PR tree-optimization/85853
892 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
893 the handling of the root of the node to...
894 (vect_slp_analyze_node_operations_1): ...this new function,
895 and run the whole thing with the child nodes' def types
896 set according to their SLP node's def type.
898 2018-05-23 Richard Biener <rguenther@suse.de>
901 * tree-data-ref.c (create_runtime_alias_checks): Defer
902 and ignore overflow warnings.
904 2018-05-23 Yury Gribov <tetra2005@gmail.com>
906 PR tree-optimization/85822
907 * tree-vrp.c (is_masked_range_test): Fix handling of negative
910 2018-05-23 Richard Biener <rguenther@suse.de>
912 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
913 memset constants via native_interpret_expr.
915 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
918 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
920 (cgraph_node::create_alias): Likewise.
921 (cgraph_node::get_availability): Check ifunc_resolver instead
922 of looking up ifunc attribute.
923 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
924 * varasm.c (do_assemble_alias): Likewise.
925 (assemble_alias): Likewise.
926 (default_binds_local_p_3): Likewise.
927 * cgraph.h (cgraph_node): Add ifunc_resolver.
928 (cgraph_node::only_called_directly_or_aliased_p): Return false
930 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
932 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
933 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
934 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
935 instead of looking up ifunc attribute.
937 2018-05-22 Luis Machado <luis.machado@linaro.org>
939 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
941 2018-05-22 Martin Sebor <msebor@redhat.com>
944 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
945 only when expasion succeeds.
946 (expand_builtin_strcmp): Same.
947 (expand_builtin_strncmp): Same.
949 2018-05-22 Martin Sebor <msebor@redhat.com>
951 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
953 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
954 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
956 * config/aarch64/aarch64-ldpstp.md: Replace uses of
957 aarch64_mem_pair_operand with memory_operand and delete operand swapping
959 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
960 Add check for legitimate_address.
961 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
962 (aarch64_swap_ldrstr_operands): New.
963 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
966 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
967 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
969 * config/aarch64/aarch64.md: New patterns to generate stp
971 (store_pair_sw, store_pair_dw): New patterns to generate stp for
972 single words and double words.
973 (load_pair_sw, load_pair_dw): Likewise.
974 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
976 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
978 * config/aarch64/aarch64-ldpstp.md: Modify peephole
979 for different mode ldpstp and add peephole for merged zero stores.
981 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
983 (aarch64_gen_store_pair): Rename calls to match new patterns.
984 (aarch64_gen_load_pair): Rename calls to match new patterns.
985 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
986 (load_pair<DREG:mode><DREG2:mode>): ... This.
987 (store_pair<mode>): Rename to...
988 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
989 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
991 (V_INT_EQUIV): Handle SImode.
992 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
995 2018-05-22 Martin Sebor <msebor@redhat.com>
998 * calls.c (maybe_warn_nonstring_arg): Use string length to set
999 or ajust the presumed bound on an operation to avoid unnecessary
1002 2018-05-22 Martin Sebor <msebor@redhat.com>
1004 PR tree-optimization/85826
1005 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1006 assuming that a DECL necesarily has a constant size.
1008 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1011 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1013 2018-05-22 Richard Biener <rguenther@suse.de>
1015 PR tree-optimization/85834
1016 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1017 non-constant and non-zero memset arguments.
1019 2018-05-22 Martin Liska <mliska@suse.cz>
1022 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1024 2018-05-22 Richard Biener <rguenther@suse.de>
1026 PR tree-optimization/85863
1027 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1028 comparisons when vectype is specified.
1029 (vectorizable_condition): Do not specify vectype for
1030 vect_is_simple_cond when SLP vectorizing.
1032 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1035 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1036 define __ibm128 as long double.
1037 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1038 as a distinct type when IEEE 128-bit support is enabled.
1039 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1040 128-bit types to use the correct functions.
1041 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1042 convert between 128-bit floating point types that have different
1043 modes but the same representation, instead of using gen_lowpart to
1045 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1047 (IFKF_reg): New attributes to give the register constraints for
1049 (extend<mode>tf2_internal): New insns to mark an explicit
1050 conversion between 128-bit floating point types that have a
1051 different mode but share the same representation.
1053 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1055 PR tree-optimization/85814
1056 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1057 a null return from get_strinfo when unsharing the next
1058 strinfo in the chain.
1060 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1063 * varasm.c (weak_finish): Clean up weak_decls.
1065 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1067 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1068 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1069 UNSPEC_UADALP values.
1070 * config/aarch64/iterators.md (ABAL): New int iterator.
1073 (sur): Add mappings for the above.
1074 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1076 (aarch64_<sur>abal<mode>_4): Likewise.
1077 (aarch64_<sur>adalp<mode>_3): Likewise.
1078 (<sur>sadv16qi): New define_expand.
1080 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1082 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1083 (*movdf_internal): Ditto.
1084 (*rcpsf2_sse): Ditto.
1085 (*rsqrtsf2_sse): Ditto.
1086 (*sqrt<mode>2_sse): Ditto.
1088 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1090 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1092 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1093 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1094 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1095 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1097 * config/aarch64/arm_neon.h: Likewise.
1098 * config/aarch64/iterators.md (VQ_I): New.
1100 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1102 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1104 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1107 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1109 * config/nds32/constraints.md (S): New constraint.
1110 * config/nds32/nds32.md (call_internal): Use constraint S.
1111 (call_value_internal): Likewise.
1112 (sibcall_internal): Likewise.
1113 (sibcall_value_internal): Likewise.
1115 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1116 Chung-Ju Wu <jasonwucj@gmail.com>
1118 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1121 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1122 Chung-Ju Wu <jasonwucj@gmail.com>
1124 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1125 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1126 (nds32_rtx_costs_impl): Simplify.
1127 (nds32_address_cost_impl): Simplify.
1128 (nds32_init_rtx_costs): New function.
1129 (nds32_rtx_costs_speed_prefer): Likewise.
1130 (nds32_rtx_costs_size_prefer): Likewise.
1131 (nds32_address_cost_speed_prefer): Likewise.
1132 (nds32_address_cost_speed_fwprop): Likewise.
1133 (nds32_address_cost_size_prefer): Likewise.
1134 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1135 * config/nds32/nds32.c (nds32_option_override): Use
1136 nds32_init_rtx_costs function.
1138 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1140 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1141 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1142 (TARGET_PIPELINE_N8): Likewise.
1143 (TARGET_PIPELINE_N10): Likewise.
1144 (TARGET_PIPELINE_N13): Likewise.
1145 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1147 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1149 * config/nds32/nds32-fpu.md: Update copyright year.
1151 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1153 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1155 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1157 * config/nds32/nds32.c
1158 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1159 * config/nds32/nds32.opt (minline-asm-r15): New option.
1161 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1163 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1165 * config/nds32/nds32.md (abssi2): New pattern.
1167 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1169 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1170 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1171 Merge insn pattern from sse_cvtsi2ss<round_name> and
1172 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1173 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1174 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1175 using SWI48 mode iterator.
1176 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1177 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1178 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1179 pattern from sse_cvttss2si<round_saeonly_name>
1180 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1181 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1182 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1183 using SWI48 mode iterator.
1184 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1185 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1186 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1187 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1188 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1189 using SWI48 mode iterator.
1190 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1191 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1192 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1193 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1194 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1195 SWI48 mode iterator.
1196 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1197 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1198 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1199 pattern from sse_cvttsd2si<round_saeonly_name>
1200 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1202 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1204 * config/nds32/nds32-md-auxiliary.c
1205 (nds32_valid_smw_lwm_base_p): Refine.
1206 (nds32_output_smw_single_word): Refine.
1207 (nds32_output_smw_double_word): New.
1208 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1210 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1212 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1213 (nds32_output_stack_pop): Refine.
1214 (nds32_expand_unaligned_load): Refine.
1215 (nds32_expand_unaligned_store): Refine.
1217 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1218 Chung-Ju Wu <jasonwucj@gmail.com>
1220 * config/nds32/constants.md: Add TP_REGNUM constant.
1221 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1222 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1224 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1225 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1226 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1227 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1228 and PIC code generation.
1229 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1231 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1233 * config/nds32/nds32.md: Support TLS and PIC.
1234 * config/nds32/nds32.c: Support TLS and PIC.
1235 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1236 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1239 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1241 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1242 mode with E_ prefix.
1244 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1245 Chung-Ju Wu <jasonwucj@gmail.com>
1247 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1248 * config/nds32/nds32-md-auxiliary.c
1249 (symbolic_reference_mentioned_p): New.
1250 (nds32_legitimize_ict_address): New.
1251 (nds32_expand_ict_move): New.
1252 (nds32_indirect_call_referenced_p): New.
1253 (nds32_symbol_binds_local_p): Delete.
1254 (nds32_long_call_p): Modify.
1255 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1256 * config/nds32/nds32-protos.h
1257 (symbolic_reference_mentioned_p): Declare.
1258 (nds32_legitimize_ict_address): Declare.
1259 (nds32_expand_ict_move): Declare.
1260 (nds32_indirect_call_referenced_p): Declare.
1261 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1262 (nds32_relax_group): Use nds32_ict_const_p as condition.
1263 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1264 (nds32_asm_file_start): Output ict_model directive in asm code.
1265 (nds32_legitimate_address_p): Consider indirect call.
1266 (nds32_print_operand): Consider indirect call.
1267 (nds32_print_operand_address): Consider indirect call.
1268 (nds32_insert_attributes): Handle "indirect_call" attribute.
1269 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1270 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1271 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1272 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1273 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1274 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1275 (TARGET_ICT_MODEL_SMALL): Define.
1276 (TARGET_ICT_MODEL_LARGE): Define.
1277 * config/nds32/nds32.md (movsi): Consider ict model.
1278 (call, call_value): Consider ict model.
1279 (sibcall, sibcall_value): Consider ict model.
1280 * config/nds32/nds32.opt (mict-model): New option.
1281 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1284 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1285 Monk Chiang <sh.chiang04@gmail.com>
1286 Jim Wilson <jimw@sifive.com>
1288 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1289 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1290 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1291 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1292 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1293 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1294 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1295 compute save_libcall_adjustment properly.
1296 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1297 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1298 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1299 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1300 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1301 (ABI_SPEC): Handle mabi=ilp32e.
1302 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1303 (RVE): Add RVE mask.
1304 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1305 <-march>: Add rv32e as an example.
1307 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1310 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1311 (intra_create_variable_infos): Handle C++ constructors.
1313 2018-05-18 Martin Liska <mliska@suse.cz>
1315 * passes.def: Remove a redundant pass.
1317 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1320 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1322 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1324 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1326 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1328 (armv2, armv2a, armv3, armv3m): Delete architectures.
1329 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1330 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1331 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1333 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1334 (*mulsidi3adddi): Likewise.
1335 (mulsidi3): Likewise.
1336 (*mulsidi3_nov6): Likewise.
1337 (umulsidi3): Likewise.
1338 (umulsidi3_nov6): Likewise.
1339 (umaddsidi4): Likewise.
1340 (*umulsidi3adddi): Likewise.
1341 (smulsi3_highpart): Likewise.
1342 (*smulsi3_highpart_nov6): Likewise.
1343 (umulsi3_highpart): Likewise.
1344 (*umulsi3_highpart_nov6): Likewise.
1345 * config/arm/arm.h (arm_arch3m): Delete.
1346 * config/arm/arm.c (arm_arch3m): Delete.
1347 (arm_option_override_internal): Update armv3-related comment.
1348 (arm_configure_build_target): Delete use of isa_bit_mode32.
1349 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1350 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1351 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1353 (mulusa3): Likewise.
1354 * config/arm/arm-protos.h (arm_arch3m): Delete.
1355 * config/arm/arm-tables.opt: Regenerate.
1356 * config/arm/arm-tune.md: Likewise.
1357 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1358 deleted architectures.
1360 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1362 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1363 (armv5t, armv5te): New features.
1364 (ARMv5, ARMv5e): Delete fgroups.
1365 (ARMv5t, ARMv5te): Adjust for above changes.
1367 (armv5, armv5e): Delete arches.
1368 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1370 (*call_reg_arm): Likewise.
1371 (*call_value_reg_armv5): Likewise.
1372 (*call_value_reg_arm): Likewise.
1373 (*call_symbol): Likewise.
1374 (*call_value_symbol): Likewise.
1375 (*sibcall_insn): Likewise.
1376 (*sibcall_value_insn): Likewise.
1378 (prefetch): Likewise.
1379 (define_split and define_peephole2 dependent on arm_arch5):
1381 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1383 (TARGET_ARM_QBIT): Likewise.
1384 (TARGET_DSP_MULTIPLY): Likewise.
1385 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1386 (arm_arch5, arm_arch5e): Delete.
1387 (arm_arch5t, arm_arch5te): Declare.
1388 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1389 (arm_arch5t): Declare.
1390 (arm_option_reconfigure_globals): Update for the above.
1391 (arm_options_perform_arch_sanity_checks): Update comment, replace
1392 use of arm_arch5 with arm_arch5t.
1393 (use_return_insn): Likewise.
1394 (arm_emit_call_insn): Likewise.
1395 (output_return_instruction): Likewise.
1396 (arm_final_prescan_insn): Likewise.
1397 (arm_coproc_builtin_available): Likewise.
1398 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1399 arm_arch5e with arm_arch5t and arm_arch5te.
1400 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1401 (arm_arch5t, arm_arch5te): Declare.
1402 * config/arm/arm-tables.opt: Regenerate.
1403 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1404 * config/arm/t-multilib: Likewise.
1405 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1406 instead of arm_arch5.
1407 (*call_reg_thumb1): Likewise.
1408 (*call_value_reg_thumb1_v5): Likewise.
1409 (*call_value_reg_thumb1): Likewise.
1410 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1412 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1414 2018-05-18 Martin Liska <mliska@suse.cz>
1416 PR gcov-profile/84846
1417 * doc/gcov.texi: Document -t option of gcov tool.
1419 2018-05-18 Martin Liska <mliska@suse.cz>
1421 PR gcov-profile/84846
1422 * gcov.c (print_usage): Add new -t option.
1423 (process_args): Handle the option.
1424 (generate_results): Use stdout as output when requested by
1427 2018-05-18 Martin Liska <mliska@suse.cz>
1429 PR gcov-profile/84846
1430 * coverage.c (coverage_init): Write PWD to .gcno file.
1431 * doc/gcov.texi: Document how working directory is printed.
1432 * gcov-dump.c (dump_gcov_file): Print PWD.
1433 * gcov.c (output_intermediate_file): Likewise.
1434 (read_graph_file): Read PWD string.
1435 (output_lines): Print PWD.
1437 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1440 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1441 for retval and return false if all args to phi are zero.
1443 2018-05-18 Richard Biener <rguenther@suse.de>
1445 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1447 (evrp_dom_walker::before_dom_children): Call it.
1449 2018-05-18 Richard Biener <rguenther@suse.de>
1451 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1452 results when processing array refs with variable index.
1454 2018-05-18 Toon Moene <toon@moene.org>
1456 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1457 directly after that of -floop-interchange. Indicate that both
1458 options are enabled by default when specifying -O3.
1460 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1462 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1463 iterator. Delete separate integer-mode vec_set<mode> expander.
1464 (aarch64_simd_vec_setv2di): Delete.
1465 (vec_setv2di): Delete.
1466 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1467 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1468 the "w, r" alternative.
1470 2018-05-18 Martin Liska <mliska@suse.cz>
1472 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1473 * tree-pass.h (make_pass_lower_switch_O0): New function.
1474 * tree-switch-conversion.c (node_has_low_bound): Remove.
1475 (node_has_high_bound): Likewise.
1476 (node_is_bounded): Likewise.
1477 (class pass_lower_switch): Make it a template type and create
1479 (pass_lower_switch::execute): Add template argument.
1480 (make_pass_lower_switch): New function.
1481 (make_pass_lower_switch_O0): New function.
1482 (do_jump_if_equal): Remove.
1483 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1484 all the hard work to tree optimization passes.
1486 2018-05-18 Martin Liska <mliska@suse.cz>
1488 * dbgcnt.c (limit_low): Renamed from limit.
1489 (limit_high): New variable.
1490 (dbg_cnt_is_enabled): Check for upper limit.
1491 (dbg_cnt): Adjust dumping.
1492 (dbg_cnt_set_limit_by_index): Add new argument for high
1494 (dbg_cnt_set_limit_by_name): Likewise.
1495 (dbg_cnt_process_single_pair): Parse new format.
1496 (dbg_cnt_process_opt): Use strtok.
1497 (dbg_cnt_list_all_counters): Remove 'value' and add
1499 * doc/invoke.texi: Document changes.
1501 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1503 * doc/sourcebuild.texi (scalar_all_fma): Document.
1504 * tree.def (FMA_EXPR): Delete.
1505 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1506 * internal-fn.c (ternary_direct): New macro.
1507 (expand_ternary_optab_fn): Likewise.
1508 (direct_ternary_optab_supported_p): Likewise.
1509 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1510 * builtins.c (fold_builtin_fma): Delete.
1511 (fold_builtin_3): Don't call it.
1512 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1513 * expr.c (expand_expr_real_2): Likewise.
1514 * fold-const.c (operand_equal_p): Likewise.
1515 (fold_ternary_loc): Likewise.
1516 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1517 * gimple.c (DEFTREECODE): Likewise.
1518 * gimplify.c (gimplify_expr): Likewise.
1519 * optabs-tree.c (optab_for_tree_code): Likewise.
1520 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1521 * tree-eh.c (operation_could_trap_p): Likewise.
1522 (stmt_could_throw_1_p): Likewise.
1523 * tree-inline.c (estimate_operator_cost): Likewise.
1524 * tree-pretty-print.c (dump_generic_node): Likewise.
1525 (op_code_prio): Likewise.
1526 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1527 * tree-ssa-operands.c (get_expr_operands): Likewise.
1528 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1529 * fold-const-call.h (fold_fma): Delete.
1530 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1531 CFN_FNMA and CFN_FNMS.
1533 * genmatch.c (combined_fn): New enum.
1534 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1535 (commutative_op): New function.
1536 (commutate): Use it. Handle more than 2 operands.
1537 (dt_operand::gen_gimple_expr): Use commutative_op.
1538 (parser::parse_expr): Allow :c to be used with non-binary
1539 operators if the commutative operand is known.
1540 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1541 CFN_FMS, CFN_FNMA and CFN_FNMS.
1542 (backprop::process_assign_use): Remove FMA_EXPR handling.
1543 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1544 (gen_hsa_fma): New function.
1545 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1546 IFN_FNMA and IFN_FNMS.
1547 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1548 * gimple-fold.h (follow_all_ssa_edges): Declare.
1549 * gimple-fold.c (follow_all_ssa_edges): New function.
1550 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1551 gimple_build interface and use follow_all_ssa_edges to fold the result.
1552 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1553 instead of checking for optabs directly.
1554 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1555 rather than FMA_EXPRs.
1556 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1557 call to IFN_FMA instead of an FMA_EXPR.
1559 2018-05-17 Jim Wilson <jimw@sifive.com>
1561 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1562 sign extended promoted subreg, and the range does not have the sign bit
1563 set, then do a sign extend.
1565 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1566 test, check for sign extended subreg and/or constant operands, and
1567 do a sign extend in that case.
1569 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1571 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1572 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1574 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1575 Change logics_shift_reg to logics_shift_imm.
1576 (thunderx2t99_fp_loadpair_basic): Delete.
1577 (thunderx2t99_fp_storepair_basic): Delete.
1578 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1579 (thunderx2t99_asimd_polynomial): Delete.
1580 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1581 and neon_fp_mul_d_scalar_q.
1582 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1583 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1584 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1585 (thunderx2t99_asimd_lut): Add missing tbl types.
1586 (thunderx2t99_asimd_ext): Delete.
1587 (thunderx2t99_asimd_load1_1_mult): Delete.
1588 (thunderx2t99_asimd_load1_2_mult): Delete.
1589 (thunderx2t99_asimd_load1_ldp): New.
1590 (thunderx2t99_asimd_load1): New.
1591 (thunderx2t99_asimd_load2): Add missing *load2* types.
1592 (thunderx2t99_asimd_load3): New.
1593 (thunderx2t99_asimd_load4): New.
1594 (thunderx2t99_asimd_store1_1_mult): Delete.
1595 (thunderx2t99_asimd_store1_2_mult): Delete.
1596 (thunderx2t99_asimd_store2_mult): Delete.
1597 (thunderx2t99_asimd_store2_onelane): Delete.
1598 (thunderx2t99_asimd_store_stp): New.
1599 (thunderx2t99_asimd_store1): New.
1600 (thunderx2t99_asimd_store2): New.
1601 (thunderx2t99_asimd_store3): New.
1602 (thunderx2t99_asimd_store4): New.
1604 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1606 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1607 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1609 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1610 Segher Boessenkool <segher@kernel.crashing.org>
1613 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1616 2018-05-17 Richard Biener <rguenther@suse.de>
1618 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1619 for pruning loop and prune defs feeding only already visited PHIs.
1621 2018-05-17 Richard Biener <rguenther@suse.de>
1623 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1625 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1626 Richard Biener <rguenther@suse.de>
1628 PR tree-optimization/85793
1629 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1630 for VMAT_ELEMENTWISE.
1632 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1634 * internal-fn.h (lookup_internal_fn): Declare
1635 * internal-fn.c (lookup_internal_fn): New function.
1636 * gimple.c (gimple_build_call_from_tree): Handle calls to
1638 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1639 internal function names.
1640 * tree-pretty-print.c (dump_generic_node): Likewise.
1641 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1643 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1645 * gimple-fold.h (gimple_build): Make the function forms take
1646 combined_fn rather than built_in_function.
1647 (gimple_simplify): Likewise.
1648 * gimple-match-head.c (gimple_simplify): Likewise.
1649 * gimple-fold.c (gimple_build): Likewise.
1650 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1651 rather than gimple_build_call_internal.
1652 (get_initial_defs_for_reduction): Likewise.
1653 (vect_create_epilog_for_reduction): Likewise.
1654 (vectorizable_live_operation): Likewise.
1656 2018-05-17 Martin Liska <mliska@suse.cz>
1658 * gimple-ssa-sprintf.c (format_directive): Do not use
1659 space in between 'G_' and '('.
1661 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1664 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1665 even if the mask is not all ones.
1668 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1670 (ix86_gimple_fold_builtin): Likewise.
1673 * config/i386/i386.c: Include tree-vector-builder.h.
1674 (ix86_vector_shift_count): New function.
1675 (ix86_fold_builtin): Fold shift builtins by scalar count.
1676 (ix86_gimple_fold_builtin): Likewise.
1678 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1679 _mm512_setzero): New intrinsics.
1681 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1682 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1684 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1685 code generation for cases where splatting a value is not useful.
1686 * simplify-rtx.c (simplify_ternary_operation): Simplify
1687 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1688 mode to a vec_concat.
1690 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1692 * config.gcc: Support "goldmont-plus".
1693 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1695 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1696 PROCESSOR_GOLDMONT_PLUS.
1697 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1698 (processor_target_table): Add "goldmont-plus".
1699 (PTA_GOLDMONT_PLUS): Define.
1700 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1701 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1702 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1703 (fold_builtin_cpu): Add "goldmont-plus".
1704 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1705 (ix86_option_override_internal): Add "goldmont-plus".
1706 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1707 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1708 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1709 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1711 2018-05-17 Richard Biener <rguenther@suse.de>
1713 PR tree-optimization/85757
1714 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1715 remove defs that only feed that PHI from further processing.
1717 2018-05-16 Jim Wilson <jimw@sifive.com>
1719 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1721 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1723 2018-05-16 Mark Wielaard <mark@klomp.org>
1725 * dwarf2out.c (count_index_strings): New function.
1726 (output_indirect_strings): Call count_index_strings and generate
1727 header for dwarf_version >= 5.
1729 2018-05-16 Mark Wielaard <mark@klomp.org>
1731 * dwarf2out.c (dwarf_FORM): New function.
1732 (set_indirect_string): Use dwarf_FORM.
1733 (reset_indirect_string): Likewise.
1734 (size_of_die): Likewise.
1735 (value_format): Likewise.
1736 (output_die): Likewise.
1737 (add_skeleton_AT_string): Likewise.
1738 (output_macinfo_op): Likewise.
1739 (index_string): Likewise.
1740 (output_index_string_offset): Likewise.
1741 (output_index_string): Likewise.
1742 (count_index_strings): Likewise.
1744 2018-05-16 Carl Love <cel@us.ibm.com>
1746 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1747 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1749 2018-05-16 Martin Jambor <mjambor@suse.cz>
1751 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1752 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1754 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1756 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1757 (fnma<mode>4): Likewise.
1758 (fms<mode>4): Likewise.
1759 (fnms<mode>4): Likewise.
1760 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1761 (aarch64_fnma<mode>4): Likewise.
1762 (aarch64_fms<mode>4): Likewise.
1763 (aarch64_fnms<mode>4): Likewise.
1764 (aarch64_fnmadd<mode>4): Likewise.
1766 2018-05-16 Jason Merrill <jason@redhat.com>
1768 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1770 2018-05-16 Richard Biener <rguenther@suse.de>
1772 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1773 (dump_stmt_cost): Declare.
1774 (add_stmt_cost): Dump cost we add.
1775 (add_stmt_costs): New function.
1776 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1778 (vect_analyze_stmt): Adjust prototype.
1779 (vectorizable_condition): Likewise.
1780 (vectorizable_live_operation): Likewise.
1781 (vectorizable_reduction): Likewise.
1782 (vectorizable_induction): Likewise.
1783 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1784 cost vector to pass to vectorizable_ and record afterwards.
1785 (vect_model_reduction_cost): Take cost vector argument and adjust.
1786 (vect_model_induction_cost): Likewise.
1787 (vectorizable_reduction): Likewise.
1788 (vectorizable_induction): Likewise.
1789 (vectorizable_live_operation): Likewise.
1790 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1791 SLP_TREE_NUMBER_OF_VEC_STMTS.
1792 (vect_analyze_slp_cost_1): Remove.
1793 (vect_analyze_slp_cost): Likewise.
1794 (vect_slp_analyze_node_operations): Take visited args and
1795 a target cost vector. Avoid processing already visited stmt sets.
1796 (vect_slp_analyze_operations): Use a local cost vector to gather
1797 costs and register those of non-discarded instances.
1798 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1799 (vect_schedule_slp_instance): Remove copying of
1800 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1802 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1803 adding cost. Record cost entry location.
1804 (vect_prologue_cost_for_slp_op): Function to compute cost of
1805 a constant or invariant generated for SLP vect in the prologue,
1806 split out from vect_analyze_slp_cost_1.
1807 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1808 (vect_model_promotion_demotion_cost): Likewise.
1809 (vect_model_store_cost): Likewise, make static.
1810 (vect_model_load_cost): Likewise.
1811 (vectorizable_bswap): Add cost vector arg and adjust.
1812 (vectorizable_call): Likewise.
1813 (vectorizable_simd_clone_call): Likewise.
1814 (vectorizable_conversion): Likewise.
1815 (vectorizable_assignment): Likewise.
1816 (vectorizable_shift): Likewise.
1817 (vectorizable_operation): Likewise.
1818 (vectorizable_store): Likewise.
1819 (vectorizable_load): Likewise.
1820 (vectorizable_condition): Likewise.
1821 (vectorizable_comparison): Likewise.
1822 (can_vectorize_live_stmts): Likewise.
1823 (vect_analyze_stmt): Likewise.
1824 (vect_transform_stmt): Adjust calls to vectorizable_*.
1825 * tree-vectorizer.c: Include gimple-pretty-print.h.
1826 (dump_stmt_cost): New function.
1828 2018-05-16 Richard Biener <rguenther@suse.de>
1830 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1831 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1832 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1833 (check_name): New callback.
1834 (dse_classify_store): Track cycles via a visited bitmap of PHI
1835 defs and simplify handling of in-loop and across loop dead stores
1836 and properly fail for loop-variant refs. Handle byte-tracking with
1837 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1840 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1842 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1843 (vect_get_mask_type_for_stmt): Likewise.
1844 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1846 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1847 to determine the statement's vector type and the vector type that
1848 should be used for calculating nunits. Deal with cases in which
1849 the type has to be deferred.
1850 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1851 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1852 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1853 (vect_determine_vf_for_stmt): New functions, split out from...
1854 (vect_determine_vectorization_factor): ...here.
1855 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1856 (vect_get_mask_type_for_stmt): New functions, split out from
1857 vect_determine_vectorization_factor.
1859 2018-05-16 Richard Biener <rguenther@suse.de>
1861 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1862 verify the [VEC_]COND_EXPR embedded comparison.
1864 2018-05-15 Martin Sebor <msebor@redhat.com>
1866 PR tree-optimization/85753
1867 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1868 RECORD_TYPE in addition to ARRAY_TYPE.
1870 2018-05-15 Martin Sebor <msebor@redhat.com>
1873 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1875 2018-05-15 Richard Biener <rguenther@suse.de>
1877 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1878 add by_clobber_p one. Change algorithm to collect all defs
1879 representing uses we need to walk and try reducing them to
1880 a single one before failing.
1881 (dse_dom_walker::dse_optimize_stmt): Adjust.
1883 2018-05-13 Mark Wielaard <mark@klomp.org>
1885 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1886 (size_of_loc_descr): Likewise.
1887 (output_loc_operands): Likewise.
1888 (output_loc_operands_raw): Likewise.
1889 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1890 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1891 (hash_loc_operands): Likewise.
1892 (compare_loc_operands): Likewise.
1894 2018-05-14 Mark Wielaard <mark@klomp.org>
1896 * dwarf2out.c (count_index_addrs): New function.
1897 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1899 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1901 PR tree-optimization/83648
1902 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1903 return value as malloc candidate.
1905 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1908 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1909 as true in call to suggest_attribute.
1911 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1913 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1914 -mreadonly-in-sdata.
1916 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1918 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1920 (aarch64_crypto_aesd_fused): Likewise.
1922 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1924 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1925 (movsi_aarch64): Likewise.
1926 (load_pairsi): Likewise.
1927 (load_pairdi): Likewise.
1928 (store_pairsi): Likewise.
1929 (store_pairdi): Likewise.
1930 (load_pairsf): Likewise.
1931 (load_pairdf): Likewise.
1932 (store_pairsf): Likewise.
1933 (store_pairdf): Likewise.
1934 (zero_extend): Likewise.
1935 (trunc): Swap alternatives.
1936 (fcvt_target): Add '?' to prefer w over r.
1938 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1941 * config/i386/i386.md: Disallow non-commutative arithmetics in
1942 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1943 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1944 in the peephole2 before it.
1946 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1948 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1949 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1950 (ix86_handle_option): Handle -mcldemote.
1951 * config.gcc: New header.
1952 * config/i386/cldemoteintrin.h: New file.
1953 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1954 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1956 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1957 OPTION_MASK_ISA_CLDEMOTE.
1958 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1959 (ix86_valid_target_attribute_inner_p): Ditto.
1960 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1961 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1962 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1963 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1964 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1966 * config/i386/i386.opt: Add -mcldemote.
1967 * config/i386/x86intrin.h: New header.
1968 * doc/invoke.texi: Add -mcldemote.
1970 2018-05-14 Richard Biener <rguenther@suse.de>
1972 * doc/match-and-simplify.texi: Adjust :s documentation.
1974 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1976 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1977 intended memcpy size.
1978 (REORDER_45): Likewise.
1980 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1982 * sort.cc: New file.
1983 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1984 * vec.c (qsort_chk): Use gcc_qsort.
1985 * Makefile.in (OBJS-libcommon): Add sort.o.
1986 (build/sort.o): New target. Use it...
1987 (BUILD_RTL): ... here, and...
1988 (build/gencfn-macros): ... here, and...
1989 (build/genmatch): ... here.
1991 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1992 Chung-Ju Wu <jasonwucj@gmail.com>
1994 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1995 * config/nds32/nds32-graywolf.md: New file.
1996 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1997 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1999 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2000 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2001 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2002 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2003 * config/nds32/pipelines.md: Include n15 settings.
2005 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2006 Chung-Ju Wu <jasonwucj@gmail.com>
2008 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2009 * config/nds32/nds32-n13.md: New file.
2010 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2011 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2013 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2014 * config/nds32/nds32.md (pipeline_model): Add n13.
2015 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2016 * config/nds32/pipelines.md: Include n13 settings.
2018 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2019 Chung-Ju Wu <jasonwucj@gmail.com>
2021 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2022 * config/nds32/nds32-n10.md: New file.
2023 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2024 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2026 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2027 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2028 * config/nds32/nds32.md (pipeline_model): Add n10.
2029 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2030 * config/nds32/pipelines.md: Include n10 settings.
2032 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2033 Kito Cheng <kito.cheng@gmail.com>
2034 Chung-Ju Wu <jasonwucj@gmail.com>
2036 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2037 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2038 Add enum values for DSP extension instructions.
2039 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2041 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2042 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2044 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2045 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2046 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2047 * config/nds32/nds32-intrinsic.md: Likewise.
2048 * config/nds32/nds32_intrinsic.h: Likewise.
2049 * config/nds32/nds32-md-auxiliary.c: Likewise.
2050 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2051 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2052 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2053 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2054 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2055 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2056 TYPE_DMAC in switch statement.
2057 * config/nds32/nds32.c: New checking and implementation for DSP
2058 extension instructions.
2059 * config/nds32/nds32.h: Likewise.
2060 * config/nds32/nds32.md: Likewise.
2061 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2062 * config/nds32/predicates.md: Implement new predicates for DSP
2065 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2067 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2068 Reformat alternatives and attributes so it is easier to identify
2069 which constraints/attributes go with which instruction.
2070 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2071 (mov<mode>_softfloat32, FMOVE64): Likewise.
2072 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2073 (mov<mode>_softfloat64, FMOVE64): Likewise.
2075 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2077 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2079 (Basic PowerPC Built-in Functions): The new name of the
2080 subsection previously known as "PowerPC Built-in Functions".
2081 (Basic PowerPC Built-in Functions Available on all Configurations):
2083 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2084 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2085 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2086 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2088 2018-05-11 Martin Jambor <mjambor@suse.cz>
2091 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2094 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2097 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2099 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2101 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2102 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2103 (ix86_handle_option): Handle -mwaitpkg.
2104 * config.gcc: New header.
2105 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2106 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2107 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2109 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2110 OPTION_MASK_ISA_WAITPKG.
2111 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2112 (ix86_option_override_internal): Add PTA_WAITPKG.
2113 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2114 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2115 IX86_BUILTIN_TPAUSE.
2116 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2117 __builtin_ia32_umwait and __builtin_ia32_tpause.
2118 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2119 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2120 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2121 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2122 UNSPECV_TPAUSE): New.
2123 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2124 * config/i386/i386.opt: Add -mwaitpkg.
2125 * config/i386/waitpkgintrin.h: New file.
2126 * config/i386/x86intrin.h: New header.
2127 * doc/invoke.texi: Add -mwaitpkg.
2129 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2132 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2134 (cortex-m0): Use armv6s-m isa.
2135 (cortex-m0plus): Likewise.
2136 (cortex-m1): Likewise.
2137 (cortex-m0.small-multiply): Likewise.
2138 (cortex-m0plus.small-multiply): Likewise.
2139 (cortex-m1.small-multiply): Likewise.
2141 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2142 Jakub Jelinek <jakub@redhat.com>
2144 PR tree-optimization/85692
2145 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2146 source permute as well.
2148 2018-05-11 Martin Liska <mliska@suse.cz>
2151 * doc/extend.texi: Document LLVM style format for no_sanitize
2154 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2156 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2157 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2158 (mode_supports_vsx_dform_quad): Likewise.
2159 (mode_supports_vmx_dform): Move these functions to be next to the
2160 other mode_supports functions.
2161 (mode_supports_dq_form): Likewise.
2162 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2163 mode_supports_dq_form.
2164 (reg_offset_addressing_ok_p): Likewise.
2165 (offsettable_ok_by_alignment): Likewise.
2166 (rs6000_legitimate_offset_address_p): Likewise.
2167 (legitimate_lo_sum_address_p): Likewise.
2168 (rs6000_legitimize_address): Likewise.
2169 (rs6000_legitimize_reload_address): Likewise.
2170 (rs6000_secondary_reload_inner): Likewise.
2171 (rs6000_preferred_reload_class): Likewise.
2172 (rs6000_output_move_128bit): Likewise.
2174 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2176 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2177 Generate SImode target register for null target.
2178 <case IX86_BUILTIN_XGETBV>: Ditto.
2179 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2180 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2182 2018-05-10 Carl Love <cel@us.ibm.com>
2184 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2185 dcbtt and dcbtstt if operands[2] is 0.
2187 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2190 * config/i386/sse.md (usadv64qi): New expander.
2192 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2194 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2195 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2196 -maltivec=be support.
2197 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2198 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2199 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2200 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2201 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2202 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2203 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2204 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2205 altivec_vsumsws): Adjust.
2206 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2207 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2208 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2210 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2211 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2212 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2213 (altivec_lve<VI_char>x): Delete expand.
2214 (*altivec_lve<VI_char>x_internal): Rename to...
2215 (altivec_lve<VI_char>x): ... this.
2216 (altivec_lvxl_<mode>): Delete expand.
2217 (*altivec_lvxl_<mode>_internal): Rename to ...
2218 (altivec_lvxl_<mode>): ... this.
2219 (altivec_stvxl_<mode>): Delete expand.
2220 (*altivec_stvxl_<mode>_internal): Rename to ...
2221 (altivec_stvxl_<mode>): ... this.
2222 (altivec_stve<VI_char>x): Delete expand.
2223 (*altivec_stve<VI_char>x_internal): Rename to ...
2224 (altivec_stve<VI_char>x): ... this.
2225 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2226 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2227 reduc_plus_scal_<mode>): Adjust.
2228 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2230 (rs6000_cpu_cpp_builtins): Adjust.
2231 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2232 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2233 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2234 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2235 -maltivec=be support.
2236 (rs6000_split_vec_extract_var): Adjust.
2237 (rs6000_split_v4si_init): Adjust.
2238 (swap_selector_for_mode): Delete.
2239 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2240 altivec_expand_stvex_be): Delete.
2241 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2242 -maltivec=be support.
2243 (rs6000_gimple_fold_builtin): Ditto.
2244 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2246 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2247 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2248 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2249 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2250 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2251 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2252 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2253 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2254 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2255 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2256 anonymous split): Adjust.
2257 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2258 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2260 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2262 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2263 when --with-gxx-include-dir is also specified.
2264 * configure: Regenerate.
2266 2018-05-09 Jim Wilson <jimw@sifive.com>
2269 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2270 * config/riscv/t-withmultilib: New.
2271 * config/riscv/withmultilib.h: New.
2272 * doc/install.texi: Document RISC-V --with-multilib-list support.
2274 2018-05-09 Richard Biener <rguenther@suse.de>
2276 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2278 (vect_bb_vectorization_profitable_p): Adjust. Compute
2279 actual scalar cost using the cost vector and the add_stmt_cost
2282 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2284 PR rtl-optimization/85645
2285 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2286 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2288 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2290 PR rtl-optimization/85645
2291 * shrink-wrap.c (spread_components): Return a boolean saying if
2292 anything was changed.
2293 (try_shrink_wrapping_separate): Iterate spread_components until
2294 nothing changes anymore.
2296 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2298 PR rtl-optimization/85645
2299 * regrename.c (build_def_use): Also kill the chains that include the
2300 destination of a REG_CFA_REGISTER note.
2302 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2304 PR rtl-optimization/85645
2305 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2306 insn that has a REG_CFA_REGISTER note.
2308 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2310 * cfgexpand.c (expand_clobber): New function.
2311 (expand_gimple_stmt_1): Use it.
2312 * tree-vect-stmts.c (vect_clobber_variable): New function,
2314 (vectorizable_simd_clone_call): ...here.
2315 (vectorizable_store): Emit a clobber either side of an
2316 IFN_STORE_LANES sequence.
2317 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2319 2018-05-09 Tom de Vries <tom@codesourcery.com>
2322 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2323 (define_insn "trap_if_false"): Add exit after trap.
2325 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2327 PR rtl-optimization/85638
2328 * bb-reorder.c: Include common/common-target.h.
2329 (create_forwarder_block): New function extracted from...
2330 (fix_up_crossing_landing_pad): ...here. Rename into...
2331 (dw2_fix_up_crossing_landing_pad): ...this.
2332 (sjlj_fix_up_crossing_landing_pad): New function.
2333 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2334 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2335 from both partitions and exit the loop after one iteration.
2337 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2340 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2342 (Basic PowerPC Built-in Functions): The new name of the
2343 subsection previously known as "PowerPC Built-in Functions".
2344 (Basic PowerPC Built-in Functions Available on all Configurations):
2346 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2348 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2350 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2352 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2355 2018-05-08 Jim Wilson <jimw@sifive.com>
2357 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2358 (LD_EMUL_SUFFIX): New.
2359 (LINK_SPEC): Use it.
2361 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2363 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2365 (Basic PowerPC Built-in Functions): The new name of the
2366 subsection previously known as "PowerPC Built-in Functions".
2367 (Basic PowerPC Built-in Functions Available on all Configurations):
2369 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2371 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2373 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2375 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2378 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2381 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2382 after cmpelim optimization.
2384 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2386 * config.gcc: Support "goldmont".
2387 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2388 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2390 * config/i386/i386.c (m_GOLDMONT): Define.
2391 (processor_target_table): Add "goldmont".
2392 (PTA_GOLDMONT): Define.
2393 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2394 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2395 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2396 (fold_builtin_cpu): Add "goldmont".
2397 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2398 (ix86_option_override_internal): Add "goldmont".
2399 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2400 (processor_type): Add PROCESSOR_GOLDMONT.
2401 * config/i386/i386.md: Add CPU "glm".
2402 * config/i386/glm.md: New file.
2403 * config/i386/x86-tune.def: Add m_GOLDMONT.
2404 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2406 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2409 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2411 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2412 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2413 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2416 * config/i386/i386.c (ix86_fold_builtin): Handle
2417 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2420 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2421 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2423 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2426 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2427 (check_arch): Likewise.
2428 (check_fpu): Return the result rather than printing it.
2429 (end arch): Fix operator precedence.
2430 (end cpu): Likewise.
2431 (END): Print the result from check_fpu.
2433 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2434 Alan Hayward <alan.hayward@arm.com>
2435 David Sherwood <david.sherwood@arm.com>
2437 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2438 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2439 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2440 (*fcmuo<mode>_and): New patterns.
2442 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2444 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2445 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2446 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2447 (cmp_op, sve_imm_con): New code attributes.
2448 (SVE_COND_INT_CMP, imm_con): Delete.
2449 (cmp_op): Remove above unspecs from int attribute.
2450 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2452 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2453 comparison-specific unspecs.
2454 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2455 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2456 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2457 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2458 (*vec_fcm<cmp_op><mode>): Rename to...
2459 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2460 (*vec_fcmuo<mode>): Rename to...
2461 (*fcmuo<mode>): ...this and adjust likewise.
2462 (*pred_fcm<cmp_op><mode>): New pattern.
2463 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2464 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2466 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2468 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2469 (aarch64_emit_sve_predicated_cond): New function.
2470 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2471 (aarch64_emit_unspec_cond_or): Replace with...
2472 (aarch64_emit_sve_or_conds): ...this new function. Use
2473 aarch64_emit_sve_ptrue_op for the individual comparisons and
2474 aarch64_emit_binop to OR them together.
2475 (aarch64_emit_inverted_unspec_cond): Replace with...
2476 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2477 aarch64_emit_sve_ptrue_op for the comparison and
2478 aarch64_emit_unop to invert the result.
2479 (aarch64_expand_sve_vec_cmp_float): Update after the above
2480 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2482 2018-05-07 Nathan Sidwell <nathan@acm.org>
2484 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2485 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2486 (Backwards Compatibility): Likewise.
2488 2018-05-07 Luis Machado <luis.machado@linaro.org>
2492 2018-05-07 Luis Machado <luis.machado@linaro.org>
2494 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2495 <prefetch_dynamic_strides>: New const bool field.
2496 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2497 prefetch_dynamic_strides.
2498 (exynosm1_prefetch_tune): Likewise.
2499 (thunderxt88_prefetch_tune): Likewise.
2500 (thunderx_prefetch_tune): Likewise.
2501 (thunderx2t99_prefetch_tune): Likewise.
2502 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2503 (aarch64_override_options_internal): Update to set
2504 PARAM_PREFETCH_DYNAMIC_STRIDES.
2505 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2506 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2507 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2508 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2509 prefetch-dynamic-strides setting.
2511 2018-05-07 Luis Machado <luis.machado@linaro.org>
2513 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2514 <minimum_stride>: New const int field.
2515 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2516 minimum_stride field.
2517 (exynosm1_prefetch_tune): Likewise.
2518 (thunderxt88_prefetch_tune): Likewise.
2519 (thunderx_prefetch_tune): Likewise.
2520 (thunderx2t99_prefetch_tune): Likewise.
2521 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2522 (aarch64_override_options_internal): Update to set
2523 PARAM_PREFETCH_MINIMUM_STRIDE.
2524 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2525 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2526 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2527 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2528 stride is constant and is below the minimum stride threshold.
2530 2018-05-07 Luis Machado <luis.machado@linaro.org>
2532 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2535 2018-05-07 Luis Machado <luis.machado@linaro.org>
2537 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2538 <prefetch_dynamic_strides>: New const bool field.
2539 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2540 prefetch_dynamic_strides.
2541 (exynosm1_prefetch_tune): Likewise.
2542 (thunderxt88_prefetch_tune): Likewise.
2543 (thunderx_prefetch_tune): Likewise.
2544 (thunderx2t99_prefetch_tune): Likewise.
2545 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2546 (aarch64_override_options_internal): Update to set
2547 PARAM_PREFETCH_DYNAMIC_STRIDES.
2548 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2549 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2550 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2551 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2552 prefetch-dynamic-strides setting.
2554 2018-05-07 Luis Machado <luis.machado@linaro.org>
2556 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2557 <minimum_stride>: New const int field.
2558 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2559 minimum_stride field.
2560 (exynosm1_prefetch_tune): Likewise.
2561 (thunderxt88_prefetch_tune): Likewise.
2562 (thunderx_prefetch_tune): Likewise.
2563 (thunderx2t99_prefetch_tune): Likewise.
2564 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2565 (aarch64_override_options_internal): Update to set
2566 PARAM_PREFETCH_MINIMUM_STRIDE.
2567 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2568 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2569 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2570 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2571 stride is constant and is below the minimum stride threshold.
2573 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2576 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2577 the type is addressable. Don't force op into register if it has
2580 2018-05-05 Roland McGrath <mcgrathr@google.com>
2583 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2584 any section for which we don't know a specific type it should have,
2585 regardless of name. Previously this was done only for the exact
2586 names ".init_array", ".fini_array", and ".preinit_array".
2587 (default_elf_asm_named_section): Add comment about
2588 relationship with default_section_type_flags and SECTION_NOTYPE.
2589 (get_section): Don't consider it a type conflict if one side has
2590 SECTION_NOTYPE and the other doesn't, as long as neither has the
2591 SECTION_BSS et al used in the default_section_type_flags logic.
2593 2018-05-05 Tom de Vries <tom@codesourcery.com>
2596 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2597 (workaround_barsyncs): New function.
2598 (nvptx_reorg): Use workaround_barsyncs.
2599 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2600 (define_expand "nvptx_membar_cta"): New define_expand.
2601 (define_insn "*nvptx_membar_cta"): New insn.
2603 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2605 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2606 To improve optimization opportunities.
2607 * builtin-types.def: The new needed builtin types for the above.
2609 2018-05-04 Richard Biener <rguenther@suse.de>
2611 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2612 * gimple-ssa-store-merging.c
2613 (imm_store_chain_info::output_merged_store): Remove redundant create,
2614 release split_store vector contents on failure.
2615 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2616 scalar stmt vector on cache hit.
2618 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2620 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2622 * config.gcc (powerpc-xilinx-eabi*): Remove.
2623 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2625 (fusion_addis_mem_combo_load): Ditto.
2626 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2628 (rs6000_cpu_cpp_builtins): Ditto.
2629 * config/rs6000/rs6000-linux.c
2630 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2631 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2632 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2634 (rs6000_setup_reg_addr_masks): Ditto.
2635 (rs6000_init_hard_regno_mode_ok): Ditto.
2636 (rs6000_option_override_internal): Ditto.
2637 (legitimate_lo_sum_address_p): Ditto.
2638 (rs6000_legitimize_address): Ditto.
2639 (rs6000_legitimize_reload_address): Ditto.
2640 (rs6000_legitimate_address_p): Ditto.
2641 (abi_v4_pass_in_fpr): Ditto.
2642 (setup_incoming_varargs): Ditto.
2643 (rs6000_gimplify_va_arg): Ditto.
2644 (rs6000_split_multireg_move): Ditto.
2645 (rs6000_savres_strategy): Ditto.
2646 (rs6000_emit_prologue_components): Ditto.
2647 (rs6000_emit_epilogue_components): Ditto.
2648 (rs6000_emit_prologue): Ditto.
2649 (rs6000_emit_epilogue): Ditto.
2650 (rs6000_elf_file_end): Ditto.
2651 (rs6000_function_value): Ditto.
2652 (rs6000_libcall_value): Ditto.
2653 * config/rs6000/rs6000.h: Ditto.
2654 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2655 (TARGET_MINMAX): ... this. New.
2656 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2657 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2658 (*movsi_internal1_single): Delete.
2659 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2660 mfpu=, mxilinx-fpu): Delete.
2661 * config/rs6000/singlefp.h: Delete.
2662 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2663 * config/rs6000/t-rs6000: Ditto.
2664 * config/rs6000/t-xilinx: Delete.
2665 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2666 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2667 (VStype_simple): Delete.
2668 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2669 * config/rs6000/xfpu.h: Delete.
2670 * config/rs6000/xfpu.md: Delete.
2671 * config/rs6000/xilinx.h: Delete.
2672 * config/rs6000/xilinx.opt: Delete.
2673 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2674 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2676 2018-05-04 Tom de Vries <tom@codesourcery.com>
2679 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2682 2018-05-04 Richard Biener <rguenther@suse.de>
2685 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2686 (expand_complex_div_wide): Likewise.
2687 (expand_complex_operations_1): Likewise.
2688 (expand_complex_libcall): Preserve EH info of the original stmt.
2689 (tree_lower_complex): Handle removed blocks.
2690 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2691 on complex multiplication and division libcall builtins.
2693 2018-05-04 Richard Biener <rguenther@suse.de>
2696 * fold-const.c (negate_expr_p): Restrict negation of operand
2697 zero of a division to when we know that can happen without
2699 (fold_negate_expr_1): Likewise.
2701 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2704 * real.h (real_nextafter): Declare.
2705 * real.c (real_nextafter): New function.
2706 * fold-const-call.c (fold_const_nextafter): New function.
2707 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2708 CASE_CFN_NEXTTOWARD.
2709 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2710 even when arg1_mode is different from arg0_mode.
2712 2018-05-03 Nathan Sidwell <nathan@acm.org>
2714 * doc/extend.texi (Deprecated Features): Remove
2716 (Backwards Compatibility): Likewise.
2717 * doc/invoke.texi (C++ Language Options): Likewise.
2718 (C++ Dialect Options): Likewise.
2720 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2723 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2724 _mm512_mask_mullox_epi64): New intrinsics.
2726 2018-05-03 Tom de Vries <tom@codesourcery.com>
2729 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2730 dump files): Add offload-tree.
2732 2018-05-03 Richard Biener <rguenther@suse.de>
2734 PR tree-optimization/85615
2735 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2736 to loops not nested in BBs loop father to avoid creating multi-entry
2739 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2741 PR tree-optimization/70291
2742 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2743 arguments. Change return type to tree. Emit libcall as a new
2744 statement rather than replacing existing one when inplace_p is true.
2745 (expand_complex_multiplication_components): New function.
2746 (expand_complex_multiplication): Expand floating-point complex
2747 multiplication using the above.
2748 (expand_complex_division): Rename inner_type parameter to type.
2749 Update expand_complex_libcall call-site.
2750 (expand_complex_operations_1): Update expand_complex_multiplication
2751 and expand_complex_division call-sites.
2753 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2756 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2757 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2758 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2759 the highest significant bit of the shift count mask is clear. In
2760 check whether and[sq]i3 is needed verify that all significant bits
2761 of the shift count other than the highest are set.
2763 2018-05-02 Tom de Vries <tom@codesourcery.com>
2766 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2767 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2768 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2769 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2770 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2771 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2772 __builtin_goacc_parlevel_size.
2774 2018-05-02 Richard Biener <rguenther@suse.de>
2776 PR tree-optimization/85597
2777 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2778 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2781 2018-05-02 Tom de Vries <tom@codesourcery.com>
2784 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2785 dump files): Add ltrans-tree.
2787 2018-05-02 Tom de Vries <tom@codesourcery.com>
2790 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2791 dump files): Add wpa-ipa.
2793 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2795 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2796 powerpc*-*-linux*paired* target.
2797 * config/rs6000/750cl.h: Delete.
2798 * config/rs6000/paired.h: Delete.
2799 * config/rs6000/paired.md: Delete.
2800 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2802 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2803 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2804 comment. Remove paired float support.
2805 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2806 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2808 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2809 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2811 * config/rs6000/rs6000.c: Remove paired float support.
2812 (paired_expand_vector_init, paired_expand_vector_move,
2813 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2814 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2815 paired_expand_builtin, paired_expand_predicate_builtin,
2816 paired_init_builtins): Delete.
2817 * config/rs6000/rs6000.h: Remove paired float support.
2818 * config/rs6000/rs6000.md: Remove paired float support.
2819 (move_from_CR_ov_bit): Delete.
2820 * config/rs6000/rs6000.opt (mpaired): Delete.
2821 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2822 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2824 2018-05-02 Richard Biener <rguenther@suse.de>
2827 * gimplify.c (gimplify_save_expr): When in SSA form allow
2828 SAVE_EXPRs to compute to SSA vars.
2830 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2833 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2834 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2835 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2836 clobber operands[2], instead use a new pseudo. Formatting fixes.
2838 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2840 PR tree-optimization/85586
2841 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2842 exit early for statements in the same group if the accesses are
2845 2018-05-02 Tom de Vries <tom@codesourcery.com>
2848 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2851 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2853 PR tree-optimization/85143
2854 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2856 2018-05-01 Tom de Vries <tom@codesourcery.com>
2859 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2860 not found" error message.
2862 2018-05-01 Tom de Vries <tom@codesourcery.com>
2865 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2866 * vec.c (test_ordered_remove_if): New function.
2867 (vec_c_tests): Call test_ordered_remove_if.
2868 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2869 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2870 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2871 VEC_ORDERED_REMOVE_IF.
2873 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2875 PR tree-optimization/82665
2876 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2877 pointer subtraction where arguments come from a memchr call.
2879 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2881 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2882 --push-state --as-needed and --pop-state instead of --as-needed and
2883 --no-as-needed if ld supports it.
2884 * configure: Regenerated.
2887 * doc/install.texi2html: Replace _002d with - and _002a with * in
2888 generated html files using sed.
2890 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2893 * gcc-rich-location.c (blank_line_before_p): New function.
2894 (use_new_line): New function.
2895 (gcc_rich_location::add_fixit_insert_formatted): New function.
2896 * gcc-rich-location.h
2897 (gcc_rich_location::add_fixit_insert_formatted): New function.
2899 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2901 * selftest.c (assert_streq): Rename "expected" and "actual" to
2902 "val1" and "val2". Extend NULL-handling to cover both inputs
2903 symmetrically, while still requiring both to be non-NULL for a pass.
2904 * selftest.h (assert_streq): Rename "expected" and "actual" to
2906 (ASSERT_EQ): Likewise.
2907 (ASSERT_EQ_AT): Likewise.
2908 (ASSERT_KNOWN_EQ): Likewise.
2909 (ASSERT_KNOWN_EQ_AT): Likewise.
2910 (ASSERT_NE): Likewise.
2911 (ASSERT_MAYBE_NE): Likewise.
2912 (ASSERT_MAYBE_NE_AT): Likewise.
2913 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2914 the assertion to pass.
2915 (ASSERT_STREQ_AT): Likewise.
2917 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2919 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2920 interaction with -pie.
2922 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2924 * selftest.h: Fix alphabetization of per-source-file selftest
2927 2018-04-30 Jason Merrill <jason@redhat.com>
2929 PR c++/61982 - dead stores to destroyed objects.
2930 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2933 2018-04-30 Jason Merrill <jason@redhat.com>
2935 * tree.c (build_clobber): New.
2936 * tree.h: Declare it.
2937 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2939 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2941 * diagnostic-show-locus.c (layout::layout): Update for
2942 location_get_source_line returning a char_span.
2943 (struct char_span): Move to input.h.
2944 (struct correction): Update for fields in char_span becoming
2946 (struct source_line): Update for location_get_source_line
2947 returning a char_span.
2948 (layout::print_line): Likewise.
2949 * edit-context.c (edited_file::print_content): Likewise.
2950 (edited_file::print_diff_hunk): Likewise.
2951 (edited_file::print_run_of_changed_lines): Likewise.
2952 (edited_file::get_num_lines): Likewise.
2953 (edited_line::edited_line): Likewise.
2954 * final.c (asm_show_source): Likewise.
2955 * input.c (location_get_source_line): Convert return type
2956 from const char * to char_span, losing the final "line_len"
2958 (dump_location_info): Update for the above.
2959 (get_substring_ranges_for_loc): Likewise. Use a char_span
2960 when handling the literal within the line.
2961 (test_reading_source_line): Update for location_get_source_line
2962 returning a char_span.
2963 * input.h (class char_span): Move here from
2964 diagnostic-show-locus.c, converting from a struct to a class.
2965 Make data members private.
2966 (char_span::operator bool): New.
2967 (char_span::length): New.
2968 (char_span::get_buffer): New.
2969 (char_span::operator[]): New.
2970 (char_span::subspan): Make const.
2971 (char_span::xstrdup): New.
2972 (location_get_source_line): Convert return type from const char *
2973 to char_span, losing the final "line_size" param.
2975 2018-04-30 Jan Hubicka <jh@suse.cz>
2977 * lto-wrapper.c (ltrans_priorities): New static var.
2978 (cmp_priority): New.
2979 (run_gcc): Read priorities and if doing parallel build order
2980 the Makefile by them.
2982 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2984 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2986 2018-04-30 Richard Biener <rguenther@suse.de>
2988 * tree-cfg.c (verify_address): Remove base argument, add
2989 flag whether to check TREE_ADDRESSABLE and do that.
2990 (verify_expr): Remove.
2991 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2992 (verify_gimple_assign_single): Likewise.
2993 (verify_gimple_switch): Likewise.
2994 (verify_expr_location_1): Dereference tp once. Add (disabled)
2995 piece from verify_expr.
2996 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2998 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3000 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3002 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3004 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3005 (small_data_pattern): Likewise.
3006 (arc_rewrite_small_data): Likewise.
3007 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3008 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3009 (get_symbol_alignment): New function.
3010 (legitimate_small_data_address_p): Likewise.
3011 (legitimate_scaled_address): Update, call
3012 legitimate_small_data_address_p.
3013 (output_sdata): New static variable.
3014 (arc_print_operand): Update how we handle small data operands.
3015 (arc_print_operand_address): Likewise.
3016 (arc_legitimate_address_p): Update, use
3017 legitimate_small_data_address_p.
3018 (arc_rewrite_small_data_p): Remove.
3019 (arc_rewrite_small_data_1): Likewise.
3020 (arc_rewrite_small_data): Likewise.
3021 (small_data_pattern): Likewise.
3022 (compact_sda_memory_operand): Update to use
3023 legitimate_small_data_address_p and get_symbol_alignment.
3024 (prepare_move_operands): Don't rewite sdata pattern.
3025 (prepare_extend_operands): Remove.
3026 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3028 (zero_extendqisi2): Likewise.
3029 (zero_extendhisi2): Likewise.
3030 (extendqihi2): Likewise.
3031 (extendqisi2): Likewise.
3032 (extendhisi2): Likewise.
3036 * config/arc/constraints.md (Usd): Change it to memory constraint.
3038 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3040 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3041 as source of std instructions.
3042 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3043 allow 6-bit constants as source for store instructions.
3044 (movdi_insn): Update instruction pattern to allow 6-bit constants
3045 as source for store instructions.
3047 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3049 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3051 2018-04-30 Nathan Sidwell <nathan@acm.org>
3052 Sandra Loosemore <sandra@codesourcery.com>
3054 * dumpfile.c (dump_open): Allow '-' for stdout.
3055 * doc/invoke.texi (Developer Options): Document dump filename
3056 determination early. Document stdin/stdout selection.
3058 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3060 Microblaze Target: PIC data text relative
3062 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3063 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3065 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3066 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3067 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3068 New addressing mode for data-text relative position indepenedent code.
3069 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3070 'ADDRESS_SYMBOLIC_TXT_REL'.
3071 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3072 (microblaze_legitimate_pic_operand): Exclude function calls from
3073 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3074 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3076 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3077 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3078 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3079 for 'address + offset'.
3080 (microblaze_expand_prologue): Add new function prologue call for
3082 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3083 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3084 table in case of TARGET_PIC_DATA_TEXT_REL.
3085 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3086 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3087 Add new macros 'UNSPEC_TEXT',
3088 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3089 + exclude function calls from 'UNSPEC_PLT' in case of data text
3091 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3092 new target hook for generating address diff vector tables in case of
3094 * doc/tm.texi : Regenerate.
3095 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3096 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3097 of addr diff vector generation.
3098 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3099 target hook definition.
3100 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3101 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3102 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3103 option for data text relative.
3105 2018-04-30 Richard Biener <rguenther@suse.de>
3107 * tree-chrec.h (evolution_function_is_constant_p): Remove
3109 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3111 2018-04-30 Richard Biener <rguenther@suse.de>
3114 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3116 2018-04-30 Richard Biener <rguenther@suse.de>
3118 PR tree-optimization/28364
3119 PR tree-optimization/85275
3120 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3121 copying first exit test.
3123 2018-04-28 Mark Wielaard <mark@klomp.org>
3125 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3127 (dwarf_AT): Handle DW_AT_addr_base.
3128 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3130 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3133 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3134 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3135 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3136 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3138 2018-04-28 Richard Biener <rguenther@suse.de>
3140 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3141 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3142 to reflect use. Only add interesting stmts.
3144 2018-04-27 Martin Jambor <mjambor@suse.cz>
3147 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3148 the jump function allows for passing through aggregate values.
3150 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3152 * input.h (in_system_header_at): Convert from macro to inline
3154 (from_macro_expansion_at): Likewise.
3155 (from_macro_definition_at): Likewise.
3157 2018-04-27 Jeff Law <law@redhat.com>
3159 * config.gcc: Mark tile* targets as deprecated/obsolete.
3161 2018-04-27 Richard Biener <rguenther@suse.de>
3163 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3166 2018-04-27 Richard Biener <rguenther@suse.de>
3168 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3170 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3172 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3173 with Yd constraint. Set "preferred_for_speed" attribute from
3174 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3176 (*movdi_internal): Ditto.
3177 (movti_interunit splitters): Remove
3178 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3179 (movdi_interunit splitters): Ditto.
3180 * config/i386/constraints.md (Ye): Remove.
3181 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3183 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3186 * config/aarch64/constraints.md (Usg): Limit to 31.
3189 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3191 PR tree-optimization/85529
3192 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3193 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3194 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3195 zero extension or masking of the MSB bit.
3196 (optimize_range_tests): Add FIRST_BB argument, pass it through
3197 to optimize_range_tests_var_bound.
3198 (maybe_optimize_range_tests, reassociate_bb): Adjust
3199 optimize_range_tests callers.
3201 2018-04-26 Richard Biener <rguenther@suse.de>
3202 Jakub Jelinek <jakub@redhat.com>
3204 * cgraph.h (symbol_table): Just declare debug method here.
3205 * symtab.c (symbol_table::debug): Define.
3207 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3209 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3211 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3213 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3214 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3215 (*movdi_internal): Substitute Yi and Yj constraint with x
3216 and Ym and Yn constraint with y constraint. Update "isa"
3217 attribute and set "preferred_for_speed" attribute from
3218 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3219 (*movsi_internal): Ditto.
3220 (*movdf_internal): Ditto.
3221 (*movsf_internal): Ditto.
3222 (*zero_extendsidi2): Ditto.
3223 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3224 (sse2_loadld): Ditto.
3225 (*vec_extract<ssevecmodelower>_0): Ditto.
3226 (*vec_extractv4si_0_zext_sse4): Ditto.
3227 (vec_concatv2di): Ditto.
3228 (*vec_dup<mode>): Ditto.
3229 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3230 * config/i386/constraints.md (Yi): Remove.
3235 2018-04-26 Nathan Sidwell <nathan@acm.org>
3237 * dumpfile.c (dump_open): New.
3238 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3239 (dump_finish): Detect stdio/stderr by value not name.
3241 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3243 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3245 2018-04-26 Tom de Vries <tom@codesourcery.com>
3248 * config/nvptx/nvptx.c (verify_neutering_jumps)
3249 (verify_neutering_labels): New function
3250 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3252 2018-04-26 Tom de Vries <tom@codesourcery.com>
3255 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3256 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3259 2018-04-26 Richard Biener <rguenther@suse.de>
3260 Tom de Vries <tom@codesourcery.com>
3263 * lto-streamer-out.c (output_function): Fixup loops if required to match
3264 discovery done in the reader.
3266 2018-04-26 Richard Biener <rguenther@suse.de>
3268 PR tree-optimization/85116
3269 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3270 have a loop exit from the single latch predecessor. Remove
3271 case of header with just condition.
3272 (ch_base::copy_headers): Exclude infinite loops from any
3274 (pass_ch::execute): Record exits.
3276 2018-04-26 Richard Biener <rguenther@suse.de>
3278 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3279 prologue cost vector and pass it to vect_get_load_cost.
3280 (vect_get_peeling_costs_all_drs): Likewise.
3281 (vect_peeling_hash_get_lowest_cost): Likewise.
3282 (vect_enhance_data_refs_alignment): Likewise.
3284 2018-04-26 Richard Biener <rguenther@suse.de>
3287 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3288 checking of integer<->pointer conversions.
3289 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3290 sign-/zero-extending pointer types.
3291 (expand_omp_for_static_chunk): Likewise.
3293 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3294 Jean Lee <xiaoyur347@gmail.com>
3296 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3297 (TARGET_ASAN_SHADOW_OFFSET): Define.
3298 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3299 true for -fsanitize=address.
3301 2018-04-25 Mark Wielaard <mark@klomp.org>
3303 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3306 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3308 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3309 than "alu", remove explicit "memory" and "imm_disp" attributes.
3310 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3313 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3314 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3315 gen_lowpart_no_emit.
3317 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3320 * config/i386/i386.c (ix86_expand_builtin): Change memory
3321 operand to XI, extend p0 to Pmode.
3322 * config/i386/i386.md: Change unspec volatile and operand
3323 1 mode to XI, change operand 0 mode to P.
3325 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3327 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3328 GET_MODE_MASK before any checking.
3329 (nds32_can_use_bset_p): Likewise.
3330 (nds32_can_use_btgl_p): Likewise.
3332 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3334 * config/nds32/nds32-doubleword.md: New define_split pattern for
3335 illegal register number.
3337 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3339 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3341 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3343 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3345 2018-04-25 Richard Biener <rguenther@suse.de>
3347 * lto-streamer.h (LTO_major_version): Bump to 8.
3349 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3351 * BASE-VER: Set to 9.0.0.
3353 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3355 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3356 in __abskf2 and __powikf2.
3358 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3361 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3362 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3363 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3364 Use the above on operand 2. Reindent.
3365 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3367 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3370 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3372 * config/i386/i386.opt (mcet): Removed.
3373 * doc/install.texi: Remove -mcet documentation.
3374 * doc/invoke.texi: Likewise.
3376 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3379 * doc/install.texi: Remove -mcet from bootstrap-cet.
3381 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3384 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3385 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3389 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3390 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3391 containing a CONST_VECTOR.
3393 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3395 * doc/install.texi: Update newlib dependency for nvptx.
3397 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3400 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3401 instead of INTVAL when shifting x left.
3403 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3405 PR tree-optimization/85478
3406 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3407 vect_grouped_store_supported for single element vectors.
3409 2018-04-24 Richard Biener <rguenther@suse.de>
3412 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3413 load cost increase to the case of non-constant step.
3415 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3418 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3419 destination if any_malformed_asm.
3421 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3424 * expr.c (store_field): In the bitfield case, if the value comes from
3425 a function call and is returned in registers by means of a PARALLEL,
3426 do not change the mode of the temporary unless BLKmode and VOIDmode.
3428 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3430 PR rtl-optimization/85423
3431 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3432 dependencies to debug insns when the previous insn is non-debug.
3434 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3436 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3437 enums into a single definition.
3438 (fls): Fix predicates and printing.
3441 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3443 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3444 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3445 and short u6 immediate.
3446 (check_if_valid_sleep_operand): Remove.
3447 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3449 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3451 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3452 flag_always_save_lp condition.
3453 * config/nds32/nds32.opt (malways-save-lp): New option.
3455 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3457 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3458 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3459 * config/nds32/nds32.h
3460 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3461 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3463 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3465 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3466 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3468 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3469 Chung-Ju Wu <jasonwucj@gmail.com>
3471 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3473 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3474 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3476 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3478 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3480 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3482 * config/nds32/nds32-protos.h (nds32_data_alignment,
3483 nds32_local_alignment): Declare.
3484 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3485 nds32_local_alignment): New functions.
3486 (TARGET_CONSTANT_ALIGNMENT): Define.
3487 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3489 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3491 * config/nds32/nds32.c
3492 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3493 (TARGET_MODES_TIEABLE_P): Likewise.
3495 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3497 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3500 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3501 Chung-Ju Wu <jasonwucj@gmail.com>
3503 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3504 for unaligned access.
3505 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3507 * config/nds32/nds32-intrinsic.md: Likewise.
3508 * config/nds32/nds32_intrinsic.h: Likewise.
3509 * config/nds32/nds32.h (nds32_builtins): Likewise.
3510 * config/nds32/nds32.opt (munaligned-access): New option.
3511 * config/nds32/nds32.c (nds32_asm_file_start): Display
3512 flag_unaligned_access status.
3514 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3516 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3517 -mno-relax is present.
3518 * config/riscv/linux.h (LINK_SPEC): Ditto.
3520 2018-04-20 Martin Sebor <msebor@redhat.com>
3523 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3525 (gimple_fold_builtin_stxcpy_chk): Same.
3526 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3528 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3531 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3532 __powikf2 when long double is IEEE 128-bit.
3534 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3536 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3537 step to make sure stack always aligned.
3539 2018-04-20 Carl Love <cel@us.ibm.com>
3542 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3543 size check for arg0.
3545 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3546 Tom de Vries <tom@codesourcery.com>
3549 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3550 Emit insns for calls too.
3551 (nvptx_find_par): Always look for worker-level predecessor insn.
3552 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3554 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3555 (nvptx_process_pars): Propagate frames for calls.
3557 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3560 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3562 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3563 (ix86_handle_option): Don't handle OPT_mibt.
3564 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3566 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3568 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3569 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3570 (ix86_target_macros): Define __CET__ with flag_cf_protection
3571 for -fcf-protection.
3572 * config/i386/i386.c (isa2_opts): Remove -mibt.
3573 * config/i386/i386.h (TARGET_IBT): Removed.
3574 (TARGET_IBT_P): Likewise.
3575 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3576 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3577 * config/i386/i386.opt (mcet): Update help message.
3580 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3581 -mcet as an alias for -mshstk.
3583 2018-04-20 Richard Biener <rguenther@suse.de>
3586 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3587 complexity by forcing a single use of the multiply operand.
3589 2018-04-20 Martin Jambor <mjambor@suse.cz>
3592 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3593 recursion dependency to only apply to non-clones.
3595 2018-04-20 Martin Jambor <mjambor@suse.cz>
3598 * ipa-cp.c (create_specialized_node): Check that clones of
3599 self-recursive edges exist during IPA-CP.
3601 2018-04-19 Toon Moene <toon@moene.org>
3603 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3606 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3608 PR tree-optimization/85467
3609 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3610 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3611 VECTOR_CST element to type.
3613 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3616 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3617 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3618 (builtin_longjmp): Likewise.
3619 (save_stack_nonlocal): New pattern.
3620 (restore_stack_nonlocal): Likewise.
3622 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3625 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3626 Replace ASM_OUTPUT_LABEL with fprintf.
3628 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3631 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3632 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3633 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3634 define __IBT__ and __SHSTK__ for -fcf-protection.
3635 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3637 (ix86_trampoline_init): Likewise.
3638 (x86_output_mi_thunk): Likewise.
3639 (ix86_notrack_prefixed_insn_p): Likewise.
3640 (ix86_option_override_internal): Don't disallow -fcf-protection.
3641 * config/i386/i386.md (rdssp<mode>): Also enable for
3643 (incssp<mode>): Likewise.
3644 (nop_endbr): Likewise.
3645 * config/i386/i386.opt (mcet): Change help message to built-in
3649 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3650 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3651 enable CET built-in functions.
3653 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3655 * common/config/i386/i386-common.c
3656 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3657 OPTION_MASK_ISA_MOVDIRI_UNSET,
3658 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3659 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3660 * config.gcc (movdirintrin.h): New header.
3661 * config/i386/cpuid.h (bit_MOVDIRI,
3662 bit_MOVDIR64B): New bits.
3663 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3665 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3666 (VOID, PVOID, PCVOID)): New function types.
3667 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3668 __builtin_ia32_directstoreu_u64,
3669 __builtin_ia32_movdir64b): New builtins.
3670 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3671 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3673 (ix86_valid_target_attribute_inner_p): Ditto.
3674 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3675 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3676 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3677 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3678 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3679 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3680 (movdiri<mode>, movdir64b_<mode>): New.
3681 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3682 * config/i386/immintrin.h: Include movdirintrin.h.
3683 * config/i386/movdirintrin.h: New file.
3684 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3686 2018-04-19 Richard Biener <rguenther@suse.de>
3689 * cfg.c (clear_bb_flags): When loop state says we have
3690 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3692 2018-04-19 Richard Biener <rguenther@suse.de>
3694 PR tree-optimization/84737
3695 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3696 copying restrict info.
3697 (vect_setup_realignment): Use it.
3698 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3699 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3700 the first DR to all generated stores.
3701 (vectorizable_load): Likewise for loads.
3703 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3705 PR tree-optimization/85446
3706 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3707 the integral and pointer types to have the same precision.
3709 * doc/install.texi: Document --disable-cet being the default and
3712 2018-04-18 Martin Liska <mliska@suse.cz>
3714 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3717 2018-04-18 Martin Liska <mliska@suse.cz>
3720 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3723 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3724 arguments if they are comparable.
3726 2018-04-18 Martin Liska <mliska@suse.cz>
3729 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3732 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3735 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3738 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3739 ENDBR after calling __morestack.
3741 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3744 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3745 by using gcc_base_ver to generate a gcc_driver_version, and use
3746 it when generating GCC_DRIVER_NAME.
3747 * configure: Regenerate.
3749 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3752 * config.gcc: Obsolete powerpc*-*-*spe*.
3754 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3757 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3758 (stabstr_D): Change type of unum from unsigned int to
3759 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3762 2018-04-17 Jim Wilson <jimw@sifive.com>
3765 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3766 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3767 Set arg_pointer_offset after using pretend_args_size.
3769 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3771 PR rtl-optimization/85431
3772 * dse.c (record_store): Ignore zero width stores.
3775 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3776 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3777 __builtin_stack_restore rather than after it.
3778 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3779 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3780 argument instead of virtual_dynamic_stack_rtx.
3782 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3784 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3786 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3787 Add note to error message to explain internal mapping of overloaded
3788 built-in function name to non-overloaded built-in function name.
3789 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3792 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3795 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3796 where the inputs overlap with the output.
3798 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3800 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3801 (=v, v) alternative and explicit "memory" attribute.
3802 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3803 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3805 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3806 "sselog1" type instead of "sselog".
3807 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3808 "sselog". Remove explicit "memory" attribute.
3809 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3810 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3812 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3813 "sselog1" type instead of "sselog". Remove explicit "memory"
3815 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3816 use "sselog1" type instead of "sselog". Remove explicit "memory"
3818 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3819 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3821 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3822 "sselog1" type instead of "sselog". Remove explicit "memory"
3824 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3825 use "sselog1" type instead of "sselog". Remove explicit "memory"
3829 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3832 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3835 2018-04-17 Martin Jambor <mjambor@suse.cz>
3838 * ipa-cp.c (create_specialized_node): Call
3839 expand_all_artificial_thunks if necessary.
3841 2018-04-17 Martin Liska <mliska@suse.cz>
3844 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3845 in message, remote space in between '_G' and '('.
3847 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3850 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3851 avx512f_vmcmp<mode>3<round_saeonly_name>,
3852 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3853 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3854 avx512f_rndscale<mode><round_saeonly_name>,
3855 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3856 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3857 Use %<iptr>2 instead of %2 for -masm=intel.
3858 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3859 avx512f_vcvttss2usi<round_saeonly_name>,
3860 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3862 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3863 avx512f_vcvttsd2usi<round_saeonly_name>,
3864 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3865 Use %q1 instead of %1 for -masm=intel.
3866 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3867 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3868 of %3 for -masm=intel.
3869 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3871 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3873 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3875 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3876 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3878 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3880 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3881 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3882 %g1 and one with %0 and %1.
3883 (avx512er_vmrcp28<mode><round_saeonly_name>,
3884 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3886 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3887 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3888 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3889 of %0 and %{%4%} for -masm=intel.
3890 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3891 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3892 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3893 order of %0 and %{%5%}%{z%} for -masm=intel.
3895 2018-04-17 Jan Hubicka <jh@suse.cz>
3898 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3900 2018-04-17 Martin Liska <mliska@suse.cz>
3903 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3904 for target_clone error message. Make default implementation
3905 clone to be a local declaration.
3906 (separate_attrs): Add new argument and check for an empty
3908 (expand_target_clones): Handle it.
3909 (ipa_target_clone): Make redirection just for target_clones
3912 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3913 Tom de Vries <tom@codesourcery.com>
3916 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3917 tiled basic blocks without omp continue statements.
3919 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3922 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3923 vec_extract expression as having side effects to make sure it gets
3926 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3929 * config/i386/i386.c (get_builtin_code_for_version): Check
3932 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3935 * gcc/config.gcc: Support "skylake".
3936 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3938 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3939 (processor_target_table): Add "skylake".
3940 (ix86_option_override_internal): Add "skylake".
3941 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3942 PROCESSOR_CANNONLAKE.
3943 (get_builtin_code_for_version): Fix priority for
3944 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3945 PROCESSOR_SKYLAKE-AVX512.
3946 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3947 (processor_type): Add PROCESSOR_SKYLAKE.
3949 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3950 Jason Merrill <jason@redhat.com>
3953 * convert.c (convert_to_integer_1): Use direct recursion for
3954 enumeral types and types with a precision less than the number
3955 of bits in their mode.
3957 2018-04-16 Julia Koval <julia.koval@intel.com>
3960 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3961 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3963 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3966 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3967 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3968 and -mno-direct-move.
3970 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3973 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3974 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3975 Ensure negative shifts result in {0}.
3977 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3979 PR rtl-optimization/79916
3980 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3981 regs (if any) to define how to gnerate SD moves when LRA is in
3984 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3986 PR rtl-optimization/85393
3987 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3988 * except.c (expand_dw2_landing_pad_for_region): Make static.
3989 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3990 a label and unconditional jump to old_bb, rather than
3991 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3994 PR rtl-optimization/85376
3995 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3996 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3997 instead of a specific value.
3999 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4000 Bin Cheng <bin.cheng@arm.com>
4002 PR tree-optimization/82965
4003 PR tree-optimization/83991
4004 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4005 by_profile_only parameter.
4006 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4007 information if the loop was predicted to iterate too many times.
4008 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4010 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4013 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4016 2018-04-13 Martin Liska <mliska@suse.cz>
4017 Jakub Jelinek <jakub@redhat.com>
4020 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4021 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4022 * builtins.c (expand_builtin_memory_copy_args): Use
4023 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4024 handle dest_addr == pc_rtx.
4026 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4029 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4030 asked to not generate direct moves.
4031 (fix_trunc<mode>si2_stfiwx): Similar.
4032 (fix_trunc<mode>si2_internal): Similar.
4034 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4037 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4038 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4039 lookup if dest in some wider mode is known to be const0_rtx and
4040 if so, record permanent equivalence for it to be ZERO_EXTEND of
4041 the narrower mode destination.
4043 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4045 * lto-streamer-out.c (output_function): Revert 259346.
4046 * omp-expand.c (expand_oacc_for): Likewise.
4048 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4050 PR rtl-optimization/85354
4051 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4052 * sel-sched.c (sel_global_init): ... here.
4054 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4057 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4058 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4059 mode for PE-COFF targets.
4060 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4061 (i386_pe_asm_lto_end): Likewise.
4062 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4063 (TARGET_ASM_LTO_END): Likewise.
4064 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4065 (i386_pe_asm_lto_start): New function.
4066 (i386_pe_asm_lto_end): Likewise.
4068 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4069 Richard Biener <rguenther@suse.de>
4072 * lto-streamer-out.c (output_function): Fix CFG loop state before
4074 * omp-expand.c (expand_oacc_for): Handle calls to internal
4075 functions like regular functions.
4077 2018-04-12 Richard Biener <rguenther@suse.de>
4080 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4081 for the early LTO debug to properly generate references to it
4082 during DIE emission. Do not re-use that for the skeleton for
4084 (dwarf2out_early_finish): Likewise.
4086 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4089 * config/i386/sse.md
4090 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4091 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4092 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4093 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4094 and output is a reg, avoid creating invalid lowpart subreg, but
4095 instead split into a 512-bit move. Don't split if not AVX512VL,
4096 input is xmm16+ reg and output is a mem.
4097 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4098 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4099 xmm16+ reg and output is a mem.
4101 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4103 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4104 also for flag_dwarf2_cfi_asm.
4106 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4108 PR rtl-optimization/85342
4109 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4110 a bool scalar var inside of the loop instead. Don't try to update
4111 recog_data.operand after failed apply_change_group.
4113 2018-04-12 Tom de Vries <tom@codesourcery.com>
4116 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4117 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4118 array with flexible array member as array without given dimension.
4119 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4120 argument for undefined param to true.
4122 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4125 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4126 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4127 from PowerPC section.
4128 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4129 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4130 help text that is too long.
4131 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4132 help text that is too long.
4133 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4134 help text that is too long.
4136 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4138 * config/alpha/alpha.md (stack_probe_internal): Rename
4139 from "probe_stack". Update all callers.
4141 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4143 PR rtl-optimization/84566
4144 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4145 sched_macro_fuse_insns.
4147 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4150 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4151 (compute_block_dependences): ... from here.
4153 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4155 PR tree-optimization/85331
4156 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4157 from int to HOST_WIDE_INT.
4159 2018-04-11 Martin Jambor <mjambor@suse.cz>
4162 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4163 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4164 not the same as the source val.
4165 (cgraph_edge_brings_value_p): New parameter.
4166 (gather_edges_for_value): Pass destination value to
4167 cgraph_edge_brings_value_p.
4168 (perhaps_add_new_callers): Likewise.
4169 (get_info_about_necessary_edges): Likewise and exclude values brought
4170 only by self-recursive edges.
4171 (create_specialized_node): Redirect only clones of self-calling edges.
4172 (+self_recursive_pass_through_p): New function.
4173 (find_more_scalar_values_for_callers_subset): Use it.
4174 (find_aggregate_values_for_callers_subset): Likewise.
4175 (known_aggs_to_agg_replacement_list): Removed.
4176 (decide_whether_version_node): Re-calculate known constants for all
4177 remaining context clones.
4179 2018-04-11 Richard Biener <rguenther@suse.de>
4182 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4183 from early DWARF output.
4184 (dwarf2out_early_finish): Output line info unconditionally into
4185 early DWARF and add reference to it.
4187 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4190 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4191 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4192 other than V2DFmode using iptr mode attribute.
4193 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4195 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4197 PR rtl-optimization/84659
4198 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4200 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4203 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4205 (output_loc_list): Pass address of a dummy size variable even in the
4206 locview handling loop.
4207 (index_location_lists): Add comment on why skip_loc_list_entry can't
4210 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4213 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4216 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4219 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4220 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4221 and -mstring-compare-inline-limit.
4223 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4226 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4227 for stack clash protection in a register whenever we need it to be in
4230 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4232 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4233 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4235 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4238 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4240 (mlong-double-): Ditto.
4241 * config/rs6000/sysv4.opt (msdata=): Ditto.
4242 (mtls-size=): Ditto.
4244 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4246 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4247 erroneous entries for
4248 "vector int vec_ldl (int, long int *)", and
4249 "vector unsigned int vec_ldl (int, unsigned long int *)".
4250 Add comments and entries for
4251 "vector bool char vec_ldl (int, bool char *)",
4252 "vector bool short vec_ldl (int, bool short *)",
4253 "vector bool int vec_ldl (int, bool int *)",
4254 "vector bool long long vec_ldl (int, bool long long *)",
4255 "vector pixel vec_ldl (int, pixel *)",
4256 "vector long long vec_ldl (int, long long *)",
4257 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4258 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4259 type tree bool_long_long_type_node and correct definition of
4260 bool_V2DI_type_node to make reference to this new type tree.
4261 (rs6000_mangle_type): Replace erroneous reference to
4262 bool_long_type_node with bool_long_long_type_node.
4263 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4264 comments to emphasize sign distinctions for char and int types and
4265 replace RS6000_BTI_bool_long constant with
4266 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4267 use of RS6000_BTI_pixel.
4268 (bool_long_type_node): Remove this macro definition.
4269 (bool_long_long_type_node): New macro definition
4271 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4273 PR rtl-optimization/85300
4274 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4275 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4276 simplify_unary_operation fails.
4278 2018-04-10 Martin Liska <mliska@suse.cz>
4280 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4281 cgraph_edge and ipa_ref.
4283 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4287 * config/i386/sse.md
4288 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4289 computation of the VEC_MERGE selector from mask.
4290 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4291 Fix decoding of the VEC_MERGE selector into mask.
4293 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4295 PR tree-optimization/85286
4296 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4298 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4300 * final.c (final_1): Set insn_last_address as well as
4301 insn_current_address.
4303 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4306 * explow.c (emit_stack_probe): Call validize_mem on memory location
4307 before passing it to gen_probe_stack. Create address operand and
4308 legitimize it for the probe_stack_address case.
4310 2018-04-09 Jan Hubicka <jh@suse.cz>
4313 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4314 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4315 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4317 (free_lang_data): Rebuild type inheritance graph.
4319 2018-04-09 Martin Sebor <msebor@redhat.com>
4321 * invoke.texi (-finline-small-functions): Mention other optimization
4323 (-findirect-inlining, -fpartial-inlining): Same.
4324 (-finline-functions-called-once): Same.
4325 (-freorder-blocks-and-partition): Same.
4327 2018-04-09 Jan Hubicka <jh@suse.cz>
4330 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4331 jumps; choose last target that matches the criteria (i.e.
4332 no partition changes for non-crossing jumps).
4333 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4334 support for redirecting crossing jumps to non-crossing.
4336 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4338 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4339 also for naked functions.
4341 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4343 * config/arc/arc.md (add_shift): New pattern.
4344 (add_shift2): Likewise.
4345 (sub_shift): Likewise.
4346 (sub_shift_cmp0_noout): Likewise.
4347 (compare_si_ashiftsi): Likewise.
4348 (xbfu_cmp0_noout): New combine pattern.
4349 (xbfu_cmp0"): Likewise.
4350 (movsi_set_cc_insn): Place the predicable variant first.
4351 (commutative_binary_cmp0_noout): Remove clobber.
4352 (commutative_binary_cmp0): New pattern.
4353 (noncommutative_binary_cmp0): Likewise.
4354 (noncommutative_binary_cmp0_noout): Likewise.
4355 (noncommutative_binary_comparison_result_used): Removed.
4356 (rsub_cmp0): New pattern.
4357 (rsub_cmp0_noout): Likewise.
4358 (extzvsi): Changed, keep only meaningful variants.
4359 (SQH, SEZ): New iterators.
4360 (SQH_postfix): New mode attribute.
4361 (SEZ_prefix): New code attribute.
4362 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4363 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4364 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4366 (noncommutative_operator): Check the availability of barrel
4369 2018-04-09 Richard Biener <rguenther@suse.de>
4371 PR tree-optimization/85284
4372 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4373 Only use the niter constraining form of simple_iv when the exit
4376 2018-04-09 Tom de Vries <tom@codesourcery.com>
4379 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4380 (define_expand "*memory_barrier"): New define_expand.
4381 (define_insn "memory_barrier"): New insn.
4383 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4385 PR rtl-optimization/80463
4386 PR rtl-optimization/83972
4387 PR rtl-optimization/83480
4389 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4390 correct producer for the insn.
4391 (tidy_control_flow): Fixup seqnos in case of debug insns.
4393 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4395 PR rtl-optimization/83913
4397 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4398 different sched-times when merging exprs.
4400 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4402 PR rtl-optimization/83962
4404 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4405 tidy_fallthru_edge and tidy_control_flow.
4407 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4409 PR rtl-optimization/83530
4411 * sel-sched.c (force_next_insn): New global variable.
4412 (remove_insn_for_debug): When force_next_insn is true, also leave only
4413 next insn in the ready list.
4414 (sel_sched_region): When the region wasn't scheduled, make another pass
4415 over it with force_next_insn set to 1.
4417 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4419 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4421 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4422 for interrupt control.
4423 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4424 functions for interrupt control.
4425 * config/nds32/nds32-intrinsic.md: Likewise.
4426 * config/nds32/nds32_intrinsic.h: Likewise.
4427 * config/nds32/nds32.h (nds32_builtins): Likewise.
4429 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4431 * config/nds32/nds32.c (nds32_init_machine_status,
4432 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4433 strict_aligned_p field.
4434 (nds32_expand_to_rtl_hook): New function.
4435 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4436 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4438 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4439 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4442 * config/nds32/nds32-n7.md: New file.
4443 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4444 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4446 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4447 * config/nds32/nds32.md (pipeline_model): Add n7.
4448 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4449 * config/nds32/pipelines.md: Include n7 settings.
4451 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4452 Chung-Ju Wu <jasonwucj@gmail.com>
4454 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4455 * config/nds32/nds32-e8.md: New file.
4456 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4457 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4459 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4460 * config/nds32/nds32.md (pipeline_model): Add e8.
4461 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4462 * config/nds32/pipelines.md: Include e8 settings.
4464 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4465 Chung-Ju Wu <jasonwucj@gmail.com>
4467 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4468 * config/nds32/nds32-n8.md: New file.
4469 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4470 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4472 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4473 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4474 * config/nds32/nds32.md (pipeline_model): Add n8.
4475 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4476 * config/nds32/pipelines.md: Include n8 settings.
4478 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4479 Chung-Ju Wu <jasonwucj@gmail.com>
4481 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4482 * config/nds32/nds32-n9-2r1w.md: New file.
4483 * config/nds32/nds32-n9-3r2w.md: New file.
4484 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4485 nds32_register_ports): New or modify for cpu n9.
4486 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4488 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4489 * config/nds32/nds32-utils.c: New file.
4490 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4491 TARGET_MUL_SLOW): Define.
4492 * config/nds32/nds32.md (pipeline_model): New attribute.
4493 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4494 New options that support cpu n9.
4495 * config/nds32/pipelines.md: Include n9 settings.
4496 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4498 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4500 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4501 information if necessary.
4502 (output_cond_branch_compare_zero): Likewise.
4503 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4504 (nds32_target_alignment): Refine for alignment.
4505 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4506 (FUNCTION_BOUNDARY): Modify.
4507 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4509 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4511 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4513 * config/nds32/constants.md (unspec_volatile_element): Add values for
4514 TLB operation and data prefetch.
4515 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4516 functions for TLB operation and data prefetch.
4517 * config/nds32/nds32-intrinsic.md: Likewise.
4518 * config/nds32/nds32_intrinsic.h: Likewise.
4519 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4520 (nds32_print_operand): Likewise.
4521 * config/nds32/nds32.h (nds32_builtins): Likewise.
4523 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4524 Andrew Pinski <pinsika@gcc.gnu.org>
4527 * match.pd: Use constant_boolean_node of correct type instead of
4528 boolean_true_node or boolean_false_node for simplifying
4529 pointer comparisons to zero.
4531 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4533 PR tree-optimization/80021
4534 * tree.c (verify_type_variant): Make error call in verify_variant_match
4535 translatable and remove final full stop.
4537 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4539 * config/nds32/constants.md (unspec_volatile_element): Add
4540 UNSPEC_VOLATILE_EH_RETURN.
4541 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4542 nds32_output_stack_pop): Support dwarf exception handling process.
4543 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4544 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4545 exception handling process.
4546 (nds32_compute_stack_frame): Likewise.
4547 (nds32_return_addr_rtx): Likewise.
4548 (nds32_initial_elimination_offset): Likewise.
4549 (nds32_expand_prologue): Likewise.
4550 (nds32_expand_epilogue): Likewise.
4551 (nds32_dynamic_chain_address): New function.
4552 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4554 (DYNAMIC_CHAIN_ADDRESS): Define.
4555 (EH_RETURN_DATA_REGNO): Define.
4556 (EH_RETURN_STACKADJ_RTX): Define.
4557 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4558 patterns for dwarf exception handling.
4560 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4562 * config/nds32/nds32.h: Clean up obsolete macros.
4564 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4566 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4567 Add enum values for particular instructions.
4568 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4569 particular intrinsic functions.
4570 * config/nds32/nds32-intrinsic.md: Likewise.
4571 * config/nds32/nds32_intrinsic.h: Likewise.
4572 * config/nds32/nds32.h (nds32_builtins): Likewise.
4573 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4574 (btst, ave): New patterns for particular instructions.
4576 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4578 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4579 Add enum values for atomic load/store and memory sync.
4580 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4582 * config/nds32/nds32-intrinsic.md: Likewise.
4583 * config/nds32/nds32_intrinsic.h: Likewise.
4584 * config/nds32/nds32.h (nds32_builtins): Likewise.
4586 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4588 PR tree-optimization/85257
4589 * fold-const.c (native_encode_vector): If not all elts could fit
4590 and off is -1, return 0 rather than offset.
4591 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4592 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4593 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4594 adjust buffer in native_interpret_expr call.
4596 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4598 * config/nds32/constants.md (unspec_volatile_element): Add cache
4599 control enum values.
4600 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4601 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4602 * config/nds32/nds32.c (nds32_cctl_names): New.
4603 (nds32_print_operand): Handle cache control register names.
4604 * config/nds32/nds32.h (nds32_builtins): New enum values.
4605 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4607 * config/nds32/nds32.md (type): Add mmu.
4608 * config/nds32/pipelines.md (simple_insn): Add mmu.
4610 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4612 * config/nds32/nds32.md (type): Remove call.
4613 * config/nds32/pipelines.md (simple_insn): Likewise.
4615 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4617 * config/nds32/constants.md (unspec_volatile_element): Add
4618 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4619 UNSPEC_VOLATILE_FMFCFG.
4620 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4621 description for fmfcfg and fmfcsr.
4622 (bdesc_1arg): Add fmtcsr.
4623 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4624 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4625 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4626 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4627 unspec_fmfcfg): New patterns.
4628 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4629 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4630 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4631 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4632 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4633 __nds32__fmfcfg): Define.
4635 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4637 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4638 intrinsic register names.
4639 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4640 intrinsic register enum values and macros.
4642 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4644 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4645 for load/store addressing form.
4646 (nds32_print_operand_address): Likewise.
4648 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4651 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4652 based on LABEL_REF. Remove useless assertion.
4653 (pic_address_needs_scratch): Fix formatting.
4654 (sparc_legitimize_pic_address): Minor tweaks.
4655 (sparc_delegitimize_address): Adjust assertion accordingly.
4656 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4657 into symbolic_operand.
4658 (movsi_high_pic_label_ref): Likewise.
4659 (movsi_lo_sum_pic_label_ref): Likewise.
4660 (movdi_pic_label_ref): Likewise.
4661 (movdi_high_pic_label_ref): Likewise.
4662 (movdi_lo_sum_pic_label_ref): Likewise.
4664 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4666 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4667 custom LIB_SPEC setup.
4669 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4670 Kito Cheng <kito.cheng@gmail.com>
4672 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4673 * config/riscv/freebsd.h: New.
4675 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4677 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4678 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4681 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4682 Kito Cheng <kito.cheng@gmail.com>
4684 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4685 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4686 * config/nds32/nds32-protos.h (nds32_output_call,
4687 nds32_output_return): Declare.
4688 * config/nds32/nds32.md: Refine all the call and return patterns.
4690 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4693 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4694 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4696 PR rtl-optimization/84872
4697 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4698 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4701 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4703 * expr.c (copy_blkmode_to_reg): Revert 254862.
4704 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4706 2018-04-06 Richard Biener <rguenther@suse.de>
4709 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4710 after seeing a component reference with an adjacent field. Treat
4711 refs to arrays at struct end of external decls similar to
4712 refs to unconstrained commons.
4714 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4717 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4718 look through SAVE_EXPRs with non-side-effects argument. Adjust
4720 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4723 2018-04-06 Richard Biener <rguenther@suse.de>
4726 * alias.c (find_base_term): New wrapper around find_base_term
4727 unwinding CSELIB_VAL_PTR changes.
4728 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4731 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4733 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4735 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4736 constant definitions.
4737 ("nop"): lr 0,0 -> nopr r0
4738 ("nop_lr0", "nop_lr1"): New insn definitions.
4740 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4742 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4743 NDS32_V3PUSH_AVAILABLE_P macro.
4745 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4746 Chung-Ju Wu <jasonwucj@gmail.com>
4748 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4749 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4750 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4751 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4752 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4753 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4754 * config/nds32/constraints.md: New constraints and checking for hard
4755 float configuration.
4756 * config/nds32/iterators.md: New mode iterator and attribute for hard
4757 float configuration.
4758 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4760 * config/nds32/nds32-fpu.md: New file.
4761 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4762 deal with hard float code generation.
4763 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4765 (abi_type, float_reg_number): New enum type.
4766 * config/nds32/nds32-predicates.c: New predicates for hard float.
4767 * config/nds32/nds32-protos.h: Declare functions for hard float.
4768 * config/nds32/nds32.c: Implementation for hard float configuration.
4769 * config/nds32/nds32.h: Definitions for hard float configuration.
4770 * config/nds32/nds32.md: Include hard float machine description and
4771 modify patterns for hard float configuration.
4772 * config/nds32/nds32.opt: New options for hard float configuration.
4773 * config/nds32/predicates.md: New predicates for hard float
4776 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4778 * common/config/nds32/nds32-common.c
4779 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4781 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4784 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4785 CONSTRUCTOR_ELT (ctor, ...)->value.
4787 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4790 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4792 2018-04-05 Tom de Vries <tom@codesourcery.com>
4795 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4798 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4799 Kito Cheng <kito.cheng@gmail.com>
4801 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4802 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4803 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4804 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4806 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4807 Kito Cheng <kito.cheng@gmail.com>
4809 * config/nds32/constraints.md (Ufe): New memory constraint.
4810 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4811 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4812 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4814 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4815 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4817 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4819 * config/nds32/nds32.md: Use optimize_size in the condition for
4820 alu-shift instructions.
4822 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4824 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4826 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4828 * config/nds32/nds32.md (negsi2): Refine pattern.
4830 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4831 Chung-Ju Wu <jasonwucj@gmail.com>
4833 * config/nds32/iterators.md (shift_rotate): New code iterator.
4834 (shift): New code attribute.
4835 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4836 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4837 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4838 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4839 bit-wise operations.
4840 (andsi3, *andsi3): Ditto.
4841 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4842 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4843 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4844 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4845 nds32_ior_operand, nds32_xor_operand): New predicates.
4847 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4849 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4850 (addsi3, subsi3): ... this.
4852 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4854 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4856 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4858 * config/nds32/nds32.md: Adjust indention.
4860 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4862 * config/nds32/nds32.md (feature): New attribute.
4864 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4866 * config/nds32/nds32.md (subtype): New attribute.
4868 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4871 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4872 expansion to perform a bitwise AND of the argument followed by a
4873 boolean negation of the result.
4875 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4877 PR rtl-optimization/84878
4878 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4879 the basic block. Assert the use reference is not artificial and that
4880 it has an associated insn.
4882 2018-04-04 Michael Matz <matz@suse.de>
4884 * builtins.c (compute_objsize): Pass correct operand
4885 to array_at_struct_end_p.
4887 2018-04-04 Richard Biener <rguenther@suse.de>
4890 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4891 from contexts for DINFO_LEVEL_TERSE and below.
4893 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4895 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4896 resiter_operand condition.
4897 * config/nds32/nds32.md (*move<mode>): Ditto.
4899 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4900 Monk Chiang <sh.chiang04@gmail.com>
4902 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4904 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4906 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4908 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4909 Kito Cheng <kito.cheng@gmail.com>
4911 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4912 nds32_cond_code_str, output_cond_branch,
4913 output_cond_branch_compare_zero, nds32_expand_cbranch,
4914 nds32_expand_cstore, nds32_expand_movcc,
4915 nds32_output_cbranchsi4_equality_zero,
4916 nds32_output_cbranchsi4_equality_reg,
4917 nds32_output_cbranchsi4_equality_reg_or_const_int,
4918 nds32_output_cbranchsi4_greater_less_zero: New functions.
4919 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4920 nds32_expand_cstore, nds32_expand_movcc,
4921 nds32_output_cbranchsi4_equality_zero,
4922 nds32_output_cbranchsi4_equality_reg,
4923 nds32_output_cbranchsi4_equality_reg_or_const_int,
4924 nds32_output_cbranchsi4_greater_less_zero): Declare.
4925 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4926 nds32_rimm11s_operand): New predicates.
4927 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4928 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4931 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4933 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4934 * config/nds32/nds32.md: Ditto.
4935 * config/nds32/pipelines.md: Ditto.
4937 2018-04-04 Richard Biener <rguenther@suse.de>
4939 PR tree-optimization/85168
4940 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4941 propagating abnormals.
4943 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4945 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4947 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4948 Kito Cheng <kito.cheng@gmail.com>
4950 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4951 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4952 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4953 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4954 * config/nds32/nds32.md (sibcall_internal): New.
4955 (sibcall_register): Remove.
4956 (sibcall_immediate): Remove.
4957 (sibcall_value_internal): New.
4958 (sibcall_value_register): Remove.
4959 (sibcall_value_immediate): Remove.
4960 * config/nds32/predicates.md (nds32_general_register_operand): New.
4961 (nds32_call_address_operand): New.
4963 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4965 PR rtl-optimization/85167
4966 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4967 bb_defs if *split_p, instead preinitialize it to NULL.
4969 PR tree-optimization/85156
4970 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4971 evaluating the argument multiple times.
4973 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4975 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4977 (_mm_cvtpd_ps): Likewise.
4978 (_mm_cvttpd_epi32): Likewise.
4979 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4980 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4981 vector, pixel, and bool following altivec.h include.
4983 2018-04-03 Martin Sebor <msebor@redhat.com>
4985 * doc/extend.texi (Common Function Attributes): Clarify.
4986 (const attribute): Likewise.
4987 (pure attribute): Likewise.
4989 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4992 * config/i386/i386.c (ix86_expand_vector_set): Use
4993 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4995 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4997 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4998 instructions when changing rounding bits to preserve precision bits
4999 in the x87 control word.
5001 2018-04-03 Martin Liska <mliska@suse.cz>
5003 PR tree-optimization/82491
5004 * rtl.h (strip_offset_and_add): Replace += suboffset with
5005 poly_uint64 () + suboffset.
5007 2018-03-29 Martin Liska <mliska@suse.cz>
5008 Martin Jambor <mjambor@suse.cz>
5011 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5012 param_type is not an integral or pointer type.
5014 2018-04-03 Richard Biener <rguenther@suse.de>
5016 * sese.h (recompute_all_dominators): Remove.
5018 2018-04-02 Martin Sebor <msebor@redhat.com>
5020 * doc/invoke.texi (-Wrestrict): Fix typos.
5022 2018-04-02 Jim Wilson <jimw@sifive.com>
5024 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5025 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5026 (<optab>di3, <optab>si3_extend): Likewise.
5027 (<optab>si3_mask, <optab>si3_mask_1): New.
5028 (<optab>di3_mask, <optab>di3_mask_1): New.
5029 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5030 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5031 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5033 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5035 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5038 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5040 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5041 (nds32_canonicalize_comparison): New function.
5043 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5044 Kito Cheng <kito.cheng@gmail.com>
5045 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5047 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5048 * config/nds32/constants.md (unspec_volatile_element): Add
5049 UNSPEC_VOLATILE_RELAX_GROUP.
5050 * config/nds32/nds32-relax-opt.c: New file.
5051 * config/nds32/nds32-predicates.c
5052 (nds32_symbol_load_store_p): New function.
5053 * config/nds32/nds32-protos.h
5054 (nds32_symbol_load_store_p): Declare function.
5055 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5056 * config/nds32/nds32.c
5057 (nds32_register_pass): New function to register pass.
5058 (nds32_register_passes): New function to register passes.
5059 * config/nds32/nds32.md (relax_group): New pattern.
5060 * config/nds32/nds32.opt (mrelax-hint): New option.
5061 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5063 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5065 * config/nds32/t-nds32: Modify files dependency.
5067 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5069 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5070 (PROFILE_HOOK): Define its implementation.
5072 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5074 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5075 type and 32-bit size.
5077 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5080 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5081 (V_128_256): New mode iterator.
5082 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5083 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5084 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5086 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5087 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5089 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5092 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5093 NaN inputs correctly.
5095 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5098 * config/rs6000/vsx.md (??r): New mode attribute.
5099 (*vsx_mov<mode>_64bit): Use it.
5100 (*vsx_mov<mode>_32bit): Likewise.
5102 2018-03-30 Martin Sebor <msebor@redhat.com>
5104 PR tree-optimization/84818
5105 * builtins.c (check_access): Use warning_n.
5107 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5110 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5112 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5115 2018-03-30 Julia Koval <julia.koval@intel.com>
5118 * x86-tune.def (movx, partial_reg_dependency): Enable for
5121 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5124 * lra-constraints.c (process_alt_operands): Move setting
5125 this_alternative_matches below.
5127 2018-03-29 Martin Liska <mliska@suse.cz>
5130 * doc/invoke.texi: Document how LTO works with debug info.
5131 Describe auto-load support of binutils. Mention 'x86-64'
5132 as valid option value of -march option.
5134 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5136 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5139 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5140 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5141 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5144 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5147 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5148 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5149 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5150 for RS6000_BTM_POWERPC64.
5151 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5152 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5153 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5158 2018-03-28 Carl Love <cel@us.ibm.com>
5161 2017-09-27 Carl Love <cel@us.ibm.com>
5163 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5164 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5165 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5168 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5170 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5171 instead of __vector bool.
5172 (_mm_max_pu8): Likewise.
5173 (_mm_min_pi16): Likewise.
5175 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5178 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5179 (DIVWEUO): Likewise.
5181 (DIVDEUO): Likewise.
5182 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5183 DIVWEUO and DIVDEUO.
5184 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5185 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5186 (div_extend): Likewise.
5187 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5189 (__builtin_divweuo): Likewise.
5190 (__builtin_divdeo): Likewise.
5191 (__builtin_divdeuo): Likewise.
5193 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5196 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5197 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5199 PR tree-optimization/82004
5200 * gimple-match-head.c (optimize_pow_to_exp): New function.
5201 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5202 Don't fold to exp if optimize_pow_to_exp is false.
5204 2018-03-28 Martin Liska <mliska@suse.cz>
5207 * calls.c (initialize_argument_information): Fix trailing space.
5208 * common.opt: Fix typo and provide better explanation for
5209 -fsanitize-coverage option.
5210 * config/i386/i386.opt: Fix typo.
5212 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5213 Martin Liska <mliska@suse.cz>
5216 * gimplify.c (asan_poison_variable): Don't do the check for
5217 gimplify_omp_ctxp here.
5218 (gimplify_decl_expr): Do it here.
5219 (gimplify_target_expr): Likewise.
5221 2018-03-28 Martin Liska <mliska@suse.cz>
5224 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5225 chkp_type_bounds_count if MPX is not enabled.
5227 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5229 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5231 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5234 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5235 function to create the function decl for complex long double
5236 multiply and divide for -mabi=ieeelongdouble.
5237 (init_float128_ieee): Call it.
5239 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5242 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5243 -fcf-protection=branch -mibt.
5244 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5246 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5249 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5251 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5254 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5255 extern array declarations.
5257 2018-03-27 Richard Biener <rguenther@suse.de>
5260 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5261 explicit single_use checks.
5263 2018-03-27 Richard Biener <rguenther@suse.de>
5265 PR tree-optimization/85082
5266 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5269 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5271 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5272 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5273 Turn on fasynchronous-unwind-tables and funwind-tables.
5275 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5278 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5279 (*bmi_blsr_<mode>_ccz): Ditto.
5281 2018-03-26 Tom de Vries <tom@codesourcery.com>
5283 PR tree-optimization/85063
5284 * omp-general.c (offloading_function_p): New function. Factor out
5286 * omp-offload.c (pass_omp_target_link::gate): ... here.
5287 * omp-general.h (offloading_function_p): Declare.
5288 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5289 with attribute omp declare target for offloading functions.
5291 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5293 PR tree-optimization/84005
5294 * tree-data-ref.h (get_base_for_alignment): Declare.
5295 * tree-data-ref.c (get_base_for_alignment_1): New function.
5296 (get_base_for_alignment): Likewise.
5297 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5298 get_base_for_alignment to find a suitable base object, instead
5299 of always using drb->base_address.
5301 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5304 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5305 known size by default.
5307 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5310 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5311 and non BLKmode operands.
5313 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5316 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5317 Clean up attributes.
5319 2018-03-23 Richard Biener <rguenther@suse.de>
5322 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5323 we are going to emit early debug for LTO.
5325 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5328 * function.c (match_asm_constraints_1): Don't optimize if input
5329 doesn't satisfy general_operand predicate for output's mode.
5332 * alias.c (write_dependence_p): Don't require for x_canonicalized
5333 non-VOIDmode if x has VOIDmode.
5336 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5337 just don't try to optimize it rather than assert it never happens.
5339 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5341 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5342 macro expansions for definition of ST_INTERNAL_<mode> and
5343 LD_INTERNAL_<mode> builtins.
5344 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5346 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5348 (altivec_expand_st_builtin): Likewise.
5349 (altivec_expand_builtin): Remove calls to deleted functions.
5350 (rs6000_address_for_altivec): Delete this function.
5351 * config/rs6000/vector.md: Remove expands for
5352 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5354 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5357 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5358 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5359 re-computing once computed.
5360 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5361 (arm_init_machine_status): Initialize
5362 machine->static_chain_stack_bytes.
5364 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5367 * doc/extend.texi: Add four new prototypes for vec_ld.
5368 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5369 definitions for more logical presentation.
5370 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5371 entries for V1TI variants of __builtin_altivec_ld builtin.
5372 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5373 handling of V1TI variant of LVX icode pattern.
5374 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5375 (rs6000_gimple_fold_builtin): Likewise.
5376 (altivec_init_builtins): Add code to define
5377 __builtin_altivec_lvx_v1ti function.
5379 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5382 * function.c (match_asm_constraints_1): Don't do the optimization
5383 if input isn't a REG, SUBREG, MEM or constant.
5385 2018-03-22 Tom de Vries <tom@codesourcery.com>
5387 PR tree-optimization/84956
5388 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5389 bb_has_abnormal_pred.
5391 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5394 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5395 DECL_INITIAL (decl) to decl at the end.
5396 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5399 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5401 * doc/extend.texi (__builtin_tgmath): Document when complex
5402 integer types are treated as _Complex _Float64.
5404 2018-03-21 Tom de Vries <tom@codesourcery.com>
5406 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5408 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5410 PR tree-optimization/84960
5411 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5412 if it is ENTRY block, move them into single succ of ENTRY in that case.
5414 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5416 PR tree-optimization/84811
5417 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5418 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5419 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5420 (known_subrange_p): Update accordingly. Cast each value involved
5421 in the size comparison, rather than casting the result of the
5424 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5426 PR tree-optimization/84982
5427 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5428 by flipping the least significant bit rather than all bits from
5429 bitpos to bitpos + bitsize - 1.
5431 2018-03-21 Nathan Sidwell <nathan@acm.org>
5433 * doc/extend.texi (Deprecated Features): Remove mention of
5434 long-deleted deprecations.
5436 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5439 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5440 * configure: Regenerate.
5442 2018-03-21 Tom de Vries <tom@codesourcery.com>
5444 PR tree-optimization/83126
5445 * tree-parloops.c (num_phis): New function.
5446 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5448 2018-03-21 Nathan Sidwell <nathan@acm.org>
5450 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5451 mention anon-struct/union members and trailing attributes.
5453 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5455 PR tree-optimization/84969
5456 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5457 builtin memset partitions if they set different rhs values.
5459 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5461 PR rtl-optimization/84989
5462 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5463 VEC_DUPLICATE with scalar result mode.
5465 2018-03-21 Martin Liska <mliska@suse.cz>
5468 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5469 not intended return statement.
5471 2018-03-21 Martin Liska <mliska@suse.cz>
5474 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5475 (chkp_find_bound_slots_1): Limit number of iterations.
5477 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5480 * Minor grammar fixes for x86 options.
5482 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5485 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5486 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5488 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5491 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5492 (altivec_lvx_<mode>_1op): Likewise.
5493 (altivec_stvx_<mode>_2op): Likewise.
5494 (altivec_stvx_<mode>_1op): Likewise.
5495 (altivec_lvx_<VM2:mode>): New define_expand.
5496 (altivec_stvx_<VM2:mode>): Likewise.
5497 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5498 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5499 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5500 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5501 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5502 (rs6000_gen_lvx): Likewise.
5503 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5504 (altivec_expand_stv_builtin): Likewise.
5505 (altivec_expand_builtin): Likewise.
5506 * config/rs6000/vector.md: Likewise.
5508 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5511 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5514 2018-03-20 Richard Biener <rguenther@suse.de>
5517 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5518 sign-conversions as zero, fall back to standard scalar_stmt
5521 2018-03-20 Martin Liska <mliska@suse.cz>
5524 * predict.c (rebuild_frequencies): Handle case when we have
5525 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5527 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5530 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5531 flag_section_anchors.
5532 * varasm.c (use_blocks_for_decl_p): Remove hack for
5533 dw2_force_const_mem.
5536 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5538 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5539 be created, use lowpart_subreg of operands[0] rather than operands[0]
5541 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5542 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5543 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5544 and n constraint instead of aarch64_shift_imm_di and Usd.
5545 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5546 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5548 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5551 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5552 to favor GPR over NEON registers.
5553 (<shift>di3_neon): Likewise.
5555 2018-03-20 Tom de Vries <tom@codesourcery.com>
5558 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5559 (nvptx_process_pars): Emit bar.sync asap and alap.
5561 2018-03-20 Tom de Vries <tom@codesourcery.com>
5564 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5565 seen_label if seen_label is already set.
5567 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5570 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5571 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5572 Use 1U instead of 1. Formatting fixes.
5575 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5576 instead of TREE_TYPE (s1) for the return value.
5578 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5580 PR tree-optimization/84946
5581 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5582 bitsize + bitsize in poly_uint64 rather than poly_int64.
5585 * dwarf2asm.c: Include fold-const.c.
5586 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5587 of decl rather than decl itself.
5589 PR rtl-optimization/84643
5590 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5592 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5595 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5596 calling assemble_variable.
5598 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5601 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5602 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5604 2018-03-19 Jim Wilson <jimw@sifive.com>
5607 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5608 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5609 (riscv_first_stack_step): Likewise.
5610 (riscv_option_override): Use STACK_BOUNDARY instead of
5612 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5614 (BIGGEST_ALIGNMENT): Set to 128.
5615 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5616 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5619 2018-03-19 Richard Biener <rguenther@suse.de>
5621 PR tree-optimization/84933
5622 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5623 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5625 2018-03-19 Richard Biener <rguenther@suse.de>
5627 PR tree-optimization/84859
5628 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5629 (cond_if_else_store_replacement): Perform sinking operation on
5630 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5631 Generalize what a BB with a single eligible store is.
5633 2018-03-19 Richard Biener <rguenther@suse.de>
5635 PR tree-optimization/84929
5636 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5637 chrec_is_positive against non-chrec arg.
5639 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5642 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5644 2018-03-18 Martin Liska <mliska@suse.cz>
5646 PR rtl-optimization/84635
5647 * regrename.c (build_def_use): Use matches_mode only when
5650 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5652 PR tree-optimization/84913
5653 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5654 vectorize chains of COND_EXPRs.
5656 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5658 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5660 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5662 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5664 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5666 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5668 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5669 Kito Cheng <kito.cheng@gmail.com>
5671 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5672 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5673 (nds32_adjust_reg_alloc_order): New function.
5674 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5676 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5678 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5679 nds32_print_operand, nds32_print_operand_address): Use
5680 HOST_WIDE_INT_PRINT_DEC instead.
5682 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5684 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5686 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5689 * config/i386/i386.c (initial_ix86_tune_features,
5690 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5692 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5693 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5694 rather than 1u << ix86_tune. Formatting fix.
5695 (ix86_option_override_internal): Change ix86_arch_mask from
5696 unsigned int to unsigned HOST_WIDE_INT, initialize to
5697 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5698 (ix86_function_specific_restore): Likewise.
5700 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5703 * postreload.c (reload_combine_recognize_pattern): Perform
5704 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5705 truncate_int_for_mode the result for the destination's mode.
5708 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5709 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5711 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5714 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5717 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5718 regno_allocno_class_array and sorted_pseudos.
5719 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5720 insns where regno is used.
5722 2018-03-16 Martin Liska <mliska@suse.cz>
5725 * multiple_target.c (create_dispatcher_calls): Redirect
5726 reference in the symbol table.
5728 2018-03-16 Martin Liska <mliska@suse.cz>
5731 * multiple_target.c (create_dispatcher_calls): Redirect also
5734 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5738 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5739 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5740 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5742 2018-03-16 Julia Koval <julia.koval@intel.com>
5744 * doc/invoke.texi (Skylake Server): Add CLWB.
5745 Cannonlake): Remove CLWB.
5747 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5749 PR tree-optimization/84841
5750 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5752 (FLOAT_ONE_CONST_TYPE): Define.
5753 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5754 (sort_by_operand_rank): Put entries with higher constant_type last
5755 rather than first to match comments.
5757 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5759 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5762 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5765 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5766 check and use address of target always.
5768 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5771 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5772 (indirect_thunk_bnd_needed): Likewise.
5773 (indirect_thunks_used): Likewise.
5774 (indirect_thunks_bnd_used): Likewise.
5775 (indirect_return_needed): New.
5776 (indirect_return_bnd_needed): Likewise.
5777 (output_indirect_thunk_function): Add a bool argument for
5779 (output_indirect_thunk_function): Don't generate alias for
5780 function return thunk.
5781 (ix86_code_end): Call output_indirect_thunk_function to generate
5782 function return thunks.
5783 (ix86_output_function_return): Set indirect_return_bnd_needed
5784 and indirect_return_needed instead of indirect_thunk_bnd_needed
5785 and indirect_thunk_needed.
5787 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5789 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5790 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5791 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5793 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5794 Paul Hua <paul.hua.gm@gmail.com>
5797 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5799 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5801 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5802 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5805 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5808 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5809 instead of GET_MODE_SIZE when comparing Units.
5811 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5814 * varasm.c (hash_section): Return an unchangeble hash value
5815 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5816 Return !aarch64_can_use_per_function_literal_pools_p ().
5818 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5821 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5822 rather than address of cmode as last argument to prepare_cmp_insn.
5824 2018-03-15 Julia Koval <julia.koval@intel.com>
5826 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5827 F_AVX512VNNI, F_AVX512BITALG): New.
5829 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5832 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5833 insn for floating-point loads and stores.
5835 2018-03-14 Carl Love <cel@us.ibm.com>
5837 * config/rs6000/rs6000-c.c: Add macro definitions for
5838 ALTIVEC_BUILTIN_VEC_PERMXOR.
5839 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5840 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5841 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5842 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5844 * config/doc/extend.texi: Add prototypes for vec_permxor.
5846 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5849 * diagnostic-show-locus.c (class layout_point): Convert m_line
5850 from int to linenum_type.
5851 (line_span::comparator): Use linenum "compare" function when
5852 comparing line numbers.
5853 (test_line_span): New function.
5854 (layout_range::contains_point): Convert param "row" from int to
5856 (layout_range::intersects_line_p): Likewise.
5857 (layout::will_show_line_p): Likewise.
5858 (layout::print_source_line): Likewise.
5859 (layout::should_print_annotation_line_p): Likewise.
5860 (layout::print_annotation_line): Likewise.
5861 (layout::print_leading_fixits): Likewise.
5862 (layout::annotation_line_showed_range_p): Likewise.
5863 (struct line_corrections): Likewise for field m_row.
5864 (line_corrections::line_corrections): Likewise for param "row".
5865 (layout::print_trailing_fixits): Likewise.
5866 (layout::get_state_at_point): Likewise.
5867 (layout::get_x_bound_for_row): Likewise.
5868 (layout::print_line): Likewise.
5869 (diagnostic_show_locus): Likewise for locals "last_line" and
5871 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5872 * input.c (selftest::test_linenum_comparisons): New function.
5873 (selftest::input_c_tests): Call it.
5874 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5875 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5876 * selftest.h (ASSERT_GT): New macro.
5877 (ASSERT_GT_AT): New macro.
5878 (ASSERT_LT): New macro.
5879 (ASSERT_LT_AT): New macro.
5881 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5883 PR rtl-optimization/84780
5884 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5886 2018-03-14 Martin Liska <mliska@suse.cz>
5888 * tree.c (record_node_allocation_statistics): Use
5889 get_stats_node_kind.
5890 (get_stats_node_kind): New function extracted from
5891 record_node_allocation_statistics.
5892 (free_node): Use get_stats_node_kind.
5894 2018-03-14 Richard Biener <rguenther@suse.de>
5896 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5897 that the value-set of ANTIC_IN doesn't grow.
5900 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5902 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5903 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5905 2018-03-14 Julia Koval <julia.koval@intel.com>
5907 * config.gcc (icelake-client, icelake-server): New.
5909 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5910 (initial_ix86_arch_features): Ditto.
5911 (PTA_SKYLAKE): Add SGX.
5912 (PTA_ICELAKE): Remove.
5913 (PTA_ICELAKE_CLIENT): New.
5914 (PTA_ICELAKE_SERVER): New.
5915 (ix86_option_override_internal): Split up icelake on icelake client and
5917 (get_builtin_code_for_version): Ditto.
5918 (fold_builtin_cpu): Ditto.
5919 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5920 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5921 * config/i386/i386.h (processor_type): Ditto.
5922 * doc/invoke.texi: Ditto.
5924 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5927 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5928 INTEGER_CST offset, add it together with bitpos / 8 and
5929 sign extend based on POINTER_SIZE.
5933 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5936 * config/i386/constraints.md (Yc): New register constraint.
5937 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5938 Use Yc constraint for alternative 2 of operand 0. Remove
5939 preferred_for_speed attribute.
5941 2018-03-14 Richard Biener <rguenther@suse.de>
5943 PR tree-optimization/84830
5944 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5945 with the old one to avoid oscillations.
5947 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5950 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5952 (assign_by_spills): Return a flag of reload assignment failure.
5953 Do not process the reload assignment failures. Do not spill other
5954 reload pseudos if they has the same reg class. Update n if
5956 (lra_assign): Add a return arg. Set up from the result of
5957 assign_by_spills call.
5958 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5959 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5960 usage_insns if it is not NULL.
5961 (spill_hard_reg_in_range): New function.
5962 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5963 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5964 function prototypes.
5965 (lra_assign): Change prototype.
5966 * lra.c (lra): Add code to deal with fails by splitting hard reg
5969 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5971 * config/riscv/riscv.opt (mrelax): New option.
5972 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5973 "norelax" when riscv_mrelax is disabled.
5974 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5976 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5979 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5980 reassociation for int modes.
5982 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5984 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5985 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5987 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5988 * config/aarch64/aarch64-sve.md
5989 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5990 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5991 (*extend<mode><Vwide>2): Rename to...
5992 (aarch64_sve_extend<mode><Vwide>2): ...this.
5993 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5994 renaming the old pattern to...
5995 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5997 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5998 define_expand, renaming the old pattern to...
5999 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6000 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6001 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6002 account when deciding which SVE instruction the optab should use.
6003 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6005 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6007 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6008 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6009 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6010 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6011 (tlsdesc_small_<mode>): Turn a define_expand and use
6012 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6013 (tlsdesc_small_advsimd_<mode>): ...this.
6014 (tlsdesc_small_sve_<mode>): New pattern.
6016 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6018 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6019 (UNSPEC_UMUL_HIGHPART): New constants.
6020 (MUL_HIGHPART): New int iteraor.
6021 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6022 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6024 (*<su>mul<mode>3_highpart): New define_insn.
6026 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6029 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6032 2018-03-13 Martin Liska <mliska@suse.cz>
6035 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6037 (sem_item_optimizer::~sem_item_optimizer): Release it.
6038 (sem_item_optimizer::merge_classes): Register variable aliases.
6039 (sem_item_optimizer::fixup_pt_set): New function.
6040 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6041 * ipa-icf.h: Declare new variables and functions.
6043 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6046 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6047 integer_pow2p@2 and test integer_pow2p in condition.
6048 (A < 0 ? C : 0): Similarly for @1.
6051 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6052 characters starting at p contain '\0' character, don't look beyond
6056 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6057 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6060 * reg-stack.c (change_stack): Change update_end var from int to
6061 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6062 also call set_block_for_insn on the newly added insns and rescan.
6065 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6066 on the last operand.
6069 * tree.c (stabilize_reference_1): Return save_expr (e) for
6070 STATEMENT_LIST even if it doesn't have side-effects.
6072 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6074 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6076 2018-03-12 Renlin Li <renlin.li@arm.com>
6078 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6079 aarch64_output_scalar_simd_mov_immediate.
6081 2018-03-12 Martin Sebor <msebor@redhat.com>
6083 PR tree-optimization/83456
6084 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6085 for perfectly overlapping calls to memcpy.
6086 (gimple_fold_builtin_memory_chk): Same.
6087 (gimple_fold_builtin_strcpy): Handle no-warning.
6088 (gimple_fold_builtin_stxcpy_chk): Same.
6089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6091 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6093 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6094 parameter. Use it for SFmode.
6095 (rs6000_function_arg_advance_1): Adjust.
6096 (rs6000_function_arg): Adjust.
6097 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6099 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6101 PR rtl-optimization/84169
6102 PR rtl-optimization/84780
6103 * combine.c (can_combine_p): Check for a 2-insn combination whether
6104 the destination register is used between the two insns, too.
6106 2018-03-12 Richard Biener <rguenther@suse.de>
6108 PR tree-optimization/84803
6109 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6110 for refs DR analysis didn't process.
6112 2018-03-12 Richard Biener <rguenther@suse.de>
6114 PR tree-optimization/84777
6115 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6116 force-vectorize loops ignore whether we are optimizing for size.
6118 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6120 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6121 (TARGET_MD_ASM_ADJUST): Define.
6123 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6124 Kito Cheng <kito.cheng@gmail.com>
6125 Chung-Ju Wu <jasonwucj@gmail.com>
6127 * config/nds32/nds32.c (nds32_compute_stack_frame,
6128 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6129 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6130 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6131 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6132 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6133 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6134 * config/nds32/nds32.md (prologue, epilogue): Use macro
6135 NDS32_V3PUSH_AVAILABLE_P to do checking.
6137 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6140 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6141 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6142 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6143 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6144 addition of most attributes on !orig_type_die or the attribute not
6145 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6147 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6148 Chung-Ju Wu <jasonwucj@gmail.com>
6150 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6152 * config/nds32/nds32.opt (mvh): New option.
6154 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6155 Chung-Ju Wu <jasonwucj@gmail.com>
6157 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6159 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6160 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6163 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6164 Chung-Ju Wu <jasonwucj@gmail.com>
6166 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6168 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6169 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6171 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6172 Kito Cheng <kito.cheng@gmail.com>
6173 Chung-Ju Wu <jasonwucj@gmail.com>
6175 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6176 UNSPEC_FFMISM and UNSPEC_FLMISM.
6177 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6178 for ffb, ffmism and flmism.
6179 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6180 (unspec_ffmism): Ditto.
6181 (unspec_flmism): Ditto.
6182 (nds32_expand_builtin_impl): Check if string extension is available.
6183 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6184 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6186 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6189 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6192 * lra-assigns.c (assign_by_spills): Return a flag of reload
6193 assignment failure. Do not process the reload assignment
6194 failures. Do not spill other reload pseudos if they has the same
6196 (lra_assign): Add a return arg. Set up from the result of
6197 assign_by_spills call.
6198 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6199 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6200 usage_insns if it is not NULL.
6201 (spill_hard_reg_in_range): New function.
6202 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6203 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6204 function prototypes.
6205 (lra_assign): Change prototype.
6206 * lra.c (lra): Add code to deal with fails by splitting hard reg
6209 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6212 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6214 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6217 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6218 (dw_val_node): Add val_symbolic_view.
6219 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6220 (symview_upper_bound): New.
6221 (new_line_info_table): Initialize symviews_since_reset.
6222 (dwarf2out_source_line): Count symviews_since_reset and set
6223 symview_upper_bound.
6224 (dw_val_equal_p): Handle symview.
6225 (add_AT_symview): New.
6226 (print_dw_val): Handle symview.
6227 (attr_checksum, attr_checksum_ordered): Likewise.
6228 (same_dw_val_p, size_of_die): Likewise.
6229 (value_format, output_die): Likewise.
6230 (add_high_low_attributes): Use add_AT_symview for entry_view.
6231 (dwarf2out_finish): Reset symview_upper_bound, clear
6234 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6237 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6238 Add strict argument and use it.
6239 (rs6000_split_multireg_move): Update for new strict argument.
6240 (mem_operand_gpr): Disallow all non-offsettable addresses.
6241 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6243 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6246 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6247 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6248 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6251 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6252 decl, use remap_type if we want to use the type.
6254 2018-03-09 Martin Sebor <msebor@redhat.com>
6256 PR tree-optimization/84526
6257 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6259 (builtin_access::generic_overlap): Be prepared to handle non-array
6262 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6264 PR rtl-optimization/84682
6265 * lra-constraints.c (process_address_1): Check is_address flag
6266 for address constraints.
6267 (process_alt_operands): Likewise.
6268 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6269 preprocess_constraints.
6270 * recog.h (preprocess_constraints): Add oploc parameter.
6272 * recog.c (preprocess_constraints): Test address_operand for
6273 CT_ADDRESS constraints.
6275 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6278 * lra-assigns.c (assign_by_spills): Return a flag of reload
6279 assignment failure. Do not process the reload assignment
6280 failures. Do not spill other reload pseudos if they has the same
6282 (lra_assign): Add a return arg. Set up from the result of
6283 assign_by_spills call.
6284 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6285 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6286 usage_insns if it is not NULL.
6287 (spill_hard_reg_in_range): New function.
6288 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6289 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6290 function prototypes.
6291 (lra_assign): Change prototype.
6292 * lra.c (lra): Add code to deal with fails by splitting hard reg
6295 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6298 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6299 Accept complain bool parameter. Only emit errors if it is true.
6300 (arm_parse_cpu_option_name): Likewise.
6301 (arm_target_thumb_only): Adjust callers of the above.
6302 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6303 prototype to take a default true bool parameter.
6304 (arm_parse_arch_option_name): Likewise.
6306 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6307 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6311 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6312 * configure: Regenerate.
6313 * configure.ac ("linker --version-script option"): New.
6314 ("linker soname option"): New.
6316 2018-03-09 Richard Biener <rguenther@suse.de>
6318 PR tree-optimization/84775
6319 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6320 immediate uses of predicate stmts and mark them modified.
6323 PR tree-optimization/84178
6324 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6326 (version_loop_for_if_conversion): Delay update_ssa call.
6327 (tree_if_conversion): Delay update_ssa until after predicate
6330 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6333 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6334 when the function accesses prior frames.
6336 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6339 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6340 gen_llsym, otherwise call maybe_gen_llsym.
6343 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6344 has ',' character inside of it.
6346 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6349 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6350 as clobbering CC_REGNUM.
6352 2018-03-08 Richard Biener <rguenther@suse.de>
6355 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6356 (follow_copies_to_constant): Do not follow SSA names registered
6359 2018-03-08 Richard Biener <rguenther@suse.de>
6361 PR tree-optimization/84178
6362 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6364 (version_loop_for_if_conversion): Delay update_ssa call.
6365 (tree_if_conversion): Delay update_ssa until after predicate
6368 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6370 PR tree-optimization/84178
6371 * tree-if-conv.c (release_bb_predicate): Remove the
6372 the assertion that the stmts have NULL use_ops.
6373 Discard the statements, asserting that they haven't
6374 yet been added to a BB.
6376 2018-03-08 Richard Biener <rguenther@suse.de>
6378 PR tree-optimization/84746
6379 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6380 (phi_translate): Pass in destination ANTIC_OUT set.
6381 (phi_translate_1): Likewise. For a simplified result lookup
6382 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6383 (phi_translate_set): Adjust.
6384 (do_pre_regular_insertion): Likewise.
6385 (do_pre_partial_partial_insertion): Likewise.
6387 2018-03-08 Martin Liska <mliska@suse.cz>
6389 PR gcov-profile/84735
6390 * doc/gcov.texi: Document usage of profile files.
6391 * gcov-io.h: Document changes in the format.
6393 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6397 * dwarf2out.c (struct dw_line_info_table): Update comments for
6399 (FORCE_RESET_NEXT_VIEW): New.
6400 (FORCE_RESETTING_VIEW_P): New.
6401 (RESETTING_VIEW_P): Check for -1 too.
6402 (ZERO_VIEW_P): Likewise.
6403 (new_line_info_table): Force-reset next view.
6404 (dwarf2out_begin_function): Likewise.
6405 (dwarf2out_source_line): Simplify zero_view_p initialization.
6406 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6407 view directly. Omit view when omitting .loc at line 0.
6409 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6411 PR tree-optimization/84740
6412 * tree-switch-conversion.c (process_switch): Call build_constructors
6413 only if info.phi_count is non-zero.
6415 PR tree-optimization/84739
6416 * tree-tailcall.c (find_tail_calls): Check call arguments against
6417 DECL_ARGUMENTS (current_function_decl) rather than
6418 DECL_ARGUMENTS (func) when checking for tail recursion.
6420 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6422 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6423 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6424 Volker Reichelt's entry and add entries for people that perform
6425 GCC fuzzy testing and report numerous bugs.
6427 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6430 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6431 readonly data in sdata, if that is disabled.
6432 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6433 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6434 -mreadonly-in-sdata option.
6436 2018-03-07 Martin Sebor <msebor@redhat.com>
6438 PR tree-optimization/84468
6439 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6440 basic block when looking for nul assignment.
6442 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6445 * except.h (output_function_exception_table): Adjust prototype.
6446 * except.c (output_function_exception_table): Remove FNNAME parameter
6447 and add SECTION parameter. Ouput one part of the table at a time.
6448 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6449 the first part of the exception table and emit unwind directives.
6450 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6451 (i386_pe_seh_cold_init): Likewise.
6452 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6453 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6454 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6455 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6456 * config/i386/winnt.c: Include except.h.
6457 (struct seh_frame_state): Add reg_offset, after_prologue and
6458 in_cold_section fields.
6459 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6460 (i386_pe_seh_cold_init): New function.
6461 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6462 to seh->in_cold_section.
6463 (seh_emit_push): Record the offset of the push.
6464 (seh_emit_save): Record the offet of the save.
6465 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6466 Test seh->after_prologue to disregard the epilogue.
6467 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6468 (i386_pe_end_cold_function): New function.
6470 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6473 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6474 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6477 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6479 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6480 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6483 * multiple_target.c: Include tree-inline.h and intl.h.
6484 (expand_target_clones): Diagnose and fail if node->definition and
6485 !tree_versionable_function_p (node->decl).
6487 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6489 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6491 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6492 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6493 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6495 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6498 * combine.c (try_combine): Use reg_or_subregno instead of handling
6499 just paradoxical SUBREGs and REGs.
6501 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6503 * config/arc/arc.c (arc_finalize_pic): Remove function.
6504 (arc_must_save_register): We use single base PIC register, remove
6505 checks to save/restore the PIC register.
6506 (arc_expand_prologue): Likewise.
6507 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6509 (arc_verify_short): Likewise.
6510 (arc_attr_type): Likewise.
6511 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6512 (walk_stores): Likewise.
6513 (arc_address_cost): Make it static.
6514 (arc_verify_short): Likewise.
6515 (branch_dest): Likewise.
6516 (arc_attr_type): Likewise.
6517 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6518 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6519 (arc_final_prescan_insn): Remove inserting the nops due to
6520 hardware hazards. It is done in reorg step.
6521 (insn_length_variant_t): Remove.
6522 (insn_length_parameters_t): Likewise.
6523 (arc_insn_length_parameters): Likewise.
6524 (arc_get_insn_variants): Likewise.
6525 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6527 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6530 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6533 PR tree-optimization/84687
6534 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6536 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6538 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6540 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6541 Rename to ppc_speculation_barrier.
6542 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6543 __builtin_ppc_speculation_barrier.
6545 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6548 * combine.c (combine_simplify_rtx): Don't try to simplify if
6549 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6552 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6554 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6555 to 32 bytes when compiling for POWER9.
6557 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6560 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6561 regparm >= 3 with no arg reg available also for calls with
6562 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6565 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6567 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6569 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6572 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6574 2018-03-05 Richard Biener <rguenther@suse.de>
6576 PR tree-optimization/84486
6577 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6578 When inserting a __builtin_assume_aligned call set the LHS
6579 SSA name alignment info accordingly.
6581 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6583 PR tree-optimization/84114
6584 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6585 Avoid reassociation of FLOAT_MODE addition.
6587 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6589 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6590 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6591 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6592 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6593 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6594 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6595 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6597 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6598 __builtin_ia32_wbinvd): New builtins.
6599 (SPECIAL_ARGS2): New.
6600 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6601 (SPECIAL_ARGS2): New.
6602 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6603 (ix86_valid_target_attribute_inner_p): Ditto.
6604 (ix86_init_mmx_sse_builtins): Add special_args2.
6605 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6606 TARGET_WBNOINVD_P): New.
6607 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6608 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6609 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6610 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6611 * config/i386/pconfigintrin.h: New file.
6612 * config/i386/wbnoinvdintrin.h: Ditto.
6613 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6614 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6616 2018-03-05 Richard Biener <rguenther@suse.de>
6618 PR tree-optimization/84670
6619 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6621 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6622 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6623 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6624 doesn't grow if all successors (recursively) were visited at least
6627 2018-03-05 Richard Biener <rguenther@suse.de>
6629 PR tree-optimization/84650
6630 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6631 if executed in the loop pipeline.
6633 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6635 * doc/configfiles.texi (Configuration Files): Move info about
6636 conditionalizing $target-protos.h to...
6637 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6638 differs from $target-protos.h.
6640 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6641 Chung-Ju Wu <jasonwucj@gmail.com>
6643 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6644 * config/nds32/nds32-multiple.md (setmemsi): Define.
6645 * config/nds32/nds32-memory-manipulation.c
6646 (nds32_gen_dup_4_byte_to_word_value): New.
6647 (emit_setmem_word_loop): New.
6648 (emit_setmem_byte_loop): New.
6649 (nds32_expand_setmem_loop): New.
6650 (nds32_expand_setmem_loop_v3m): New.
6651 (nds32_expand_setmem_unroll): New.
6652 (nds32_expand_setmem): New.
6654 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6655 Chung-Ju Wu <jasonwucj@gmail.com>
6657 * config/nds32/nds32-memory-manipulation.c
6658 (nds32_emit_load_store): New.
6659 (nds32_emit_post_inc_load_store): New.
6660 (nds32_emit_mem_move): New.
6661 (nds32_emit_mem_move_block): New.
6662 (nds32_expand_movmemsi_loop_unknown_size): New.
6663 (nds32_expand_movmemsi_loop_known_size): New.
6664 (nds32_expand_movmemsi_loop): New.
6665 (nds32_expand_movmemsi_unroll): New.
6666 (nds32_expand_movmemqi): Rename ...
6667 (nds32_expand_movmemsi): ... to this.
6668 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6669 (movmemsi): ... to this.
6670 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6671 (nds32_expand_movmemsi): ... to this.
6673 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6674 Monk Chiang <sh.chiang04@gmail.com>
6675 Chung-Ju Wu <jasonwucj@gmail.com>
6677 * config/nds32/nds32-protos.h
6678 (nds32_expand_load_multiple): New arguments.
6679 (nds32_expand_store_multiple): Ditto.
6680 (nds32_valid_multiple_load_store): Rename ...
6681 (nds32_valid_multiple_load_store_p): ... to this.
6682 * config/nds32/nds32-memory-manipulation.c
6683 (nds32_expand_load_multiple): Refine implementation.
6684 (nds32_expand_store_multiple): Ditto.
6685 * config/nds32/nds32-multiple.md
6686 (load_multiple): Update nds32_expand_load_multiple interface.
6687 (store_multiple): Update nds32_expand_store_multiple interface.
6688 * config/nds32/nds32-predicates.c
6689 (nds32_valid_multiple_load_store): Rename ...
6690 (nds32_valid_multiple_load_store_p): ... to this and refine
6692 * config/nds32/predicates.md
6693 (nds32_load_multiple_and_update_address_operation): New predicate.
6694 (nds32_store_multiple_and_update_address_operation): New predicate.
6696 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6697 Chung-Ju Wu <jasonwucj@gmail.com>
6699 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6700 (combo): New attribute.
6701 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6703 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6705 * config/nds32/nds32.opt: Change -mcmodel= default value.
6707 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6708 Monk Chiang <sh.chiang04@gmail.com>
6709 Chung-Ju Wu <jasonwucj@gmail.com>
6711 * config/nds32/constants.md (unspec_element): New enum.
6712 * config/nds32/constraints.md (Umw): New constraint.
6713 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6714 * config/nds32/nds32-intrinsic.md: Likewise.
6715 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6716 (nds32_valid_smw_lwm_base_p): New.
6717 (nds32_output_smw_single_word): New.
6718 (nds32_output_lmw_single_word): New.
6719 (nds32_expand_unaligned_load): New.
6720 (nds32_expand_unaligned_store): New.
6721 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6722 (nds32_output_smw_single_word): Declare.
6723 (nds32_output_lmw_single_word): Declare.
6724 (nds32_expand_unaligned_load): Declare.
6725 (nds32_expand_unaligned_store): Declare.
6726 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6727 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6728 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6729 NDS32_BUILTIN_UASTORE_DW.
6730 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6733 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6734 Kito Cheng <kito.cheng@gmail.com>
6735 Chung-Ju Wu <jasonwucj@gmail.com>
6737 * config/nds32/nds32-intrinsic.c
6738 (nds32_expand_builtin_null_ftype_reg): Delete.
6739 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6740 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6741 (nds32_read_argument): New.
6742 (nds32_legitimize_target): Ditto.
6743 (nds32_legitimize_argument): Ditto.
6744 (nds32_check_constant_argument): Ditto.
6745 (nds32_expand_unop_builtin): Ditto.
6746 (nds32_expand_unopimm_builtin): Ditto.
6747 (nds32_expand_binop_builtin): Ditto.
6748 (nds32_builtin_decl_impl): Ditto.
6749 (builtin_description): Ditto.
6750 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6751 (nds32_init_builtins_impl): Ditto.
6752 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6753 (nds32_builtin_decl): New.
6754 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6755 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6757 2018-03-02 Jeff Law <law@redhat.com>
6759 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6760 (redundant_insn, fill_simple_delay_slots): Likewise.
6761 (fill_slots_from_thread): Likewise.
6762 * resource.c (mark_referenced_resources): Likewise.
6763 (mark_set_resources, find_dead_or_set_registers): Likewise.
6765 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6767 * substring-locations.h (format_warning_va): Formatting fix for
6769 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6771 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6772 * substring-locations.c: Include intl.h.
6773 (format_warning_va): Turned into small wrapper around
6774 format_warning_n_va, renamed to ...
6775 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6776 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6778 (format_warning_at_substring_n): New function.
6779 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6780 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6781 format_warning_at_substring with just a shorter name instead of
6782 const function pointer.
6783 (fmtwarn_n): New function.
6784 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6785 appropriate, get rid of all the fmtstr temporaries, move conditionals
6786 with G_() wrapped string literals directly into fmtwarn arguments,
6787 cast dir.len to (int), formatting fixes.
6789 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6791 * doc/invoke.texi: Remove "Cilk Plus" references.
6793 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6794 Richard Biener <rguenther@suse.de>
6797 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6798 for error or warning attributes if CALL_FROM_THUNK_P is set.
6801 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6804 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6805 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6808 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6809 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6811 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6812 instead of -1U in last predictors element's probability member.
6814 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6817 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6818 arguments if they are comparable.
6820 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6822 PR tree-optimization/84634
6823 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6824 masks and masked_loop_p with a single loop_masks, making sure it's
6825 null for bb vectorization.
6827 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6829 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6830 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6831 loop->force_vectorize to check whether there is no alias.
6833 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6836 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6838 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6840 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6841 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6842 instead of a loop around prev_real_insn.
6843 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6847 * config/i386/i386.c (ix86_print_operand): Use conditional
6848 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6851 2018-03-02 Richard Biener <rguenther@suse.de>
6853 PR tree-optimization/84427
6854 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6855 (bitmap_set_subtract_values): Rewrite to handle multiple
6858 (prune_clobbered_mems): Likewise.
6859 (phi_translate): Take edge instead of pred/phiblock.
6860 (phi_translate_1): Likewise.
6861 (phi_translate_set): Likewise. Insert all translated
6862 exprs for a value into the set, keeping possibly multiple
6863 expressions per value.
6864 (compute_antic_aux): Adjust for phi_translate changes.
6865 When intersecting union the expressions and prune those
6866 not in the final value set, keeping possibly multiple
6867 expressions per value. Do not use value-insertion
6868 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6869 all expressions. Add verification that the value-sets
6870 only shrink during iteration.
6871 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6872 (do_pre_regular_insertion): Likewise.
6873 (do_pre_partial_partial_insertion): Likewise.
6875 2018-03-02 Richard Biener <rguenther@suse.de>
6878 * config/darwin.c (saved_debug_info_level): New static global.
6879 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6880 (darwin_asm_lto_end): Restore debug info generation settings.
6882 2018-03-01 Martin Liska <mliska@suse.cz>
6885 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6888 2018-03-01 Richard Biener <rguenther@suse.de>
6891 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6892 type annotation with LTO when debug was disabled at compile-time.
6894 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6896 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6898 (mips_final_postscan_insn): Likewise.
6900 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6902 PR rtl-optimization/84528
6903 * alias.c (init_alias_target): Add commentary.
6904 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6905 a unique base value if the frame pointer is not eliminated
6906 to the stack pointer.
6908 2018-03-01 Tom de Vries <tom@codesourcery.com>
6910 PR rtl-optimization/83327
6911 * lra-int.h (hard_regs_spilled_into): Declare.
6912 * lra.c (hard_regs_spilled_into): Define.
6913 (init_reg_info): Init hard_regs_spilled_into.
6914 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6915 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6916 (process_bb_lives): Handle hard_regs_spilled_into.
6917 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6918 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6920 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6922 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6923 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6924 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6925 * config/rs6000/aix72.h: New file.
6927 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6929 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6930 instead of warning_at with conditional singular and plural messages
6934 * stor-layout.c (update_alignment_for_field): For
6935 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6936 && !DECL_PACKED (field), do the alignment update, just use
6937 only desired_align instead of MAX (type_align, desired_align)
6939 (place_field): Don't do known_align < desired_align handling
6940 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6941 is non-NULL, instead do it after rli->prev_field handling and
6942 only if not within a bitfield word. For DECL_PACKED (field)
6943 use type_align of BITS_PER_UNIT.
6945 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6947 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6948 superfluous parentheses and trailing spaces.
6950 2018-02-28 Richard Biener <rguenther@suse.de>
6952 PR tree-optimization/84584
6953 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6954 SCoPs with fake exit edge.
6956 2018-02-28 Martin Liska <mliska@suse.cz>
6959 * timevar.c (timer::print): Fix format to properly print 100%
6962 2018-02-28 Richard Biener <rguenther@suse.de>
6965 * genmatch.c (capture_info::walk_match): Do not mark
6966 captured expressions without operands as expr_p given
6967 they act more like predicates and should be subject to
6968 "lost tail" side-effect preserving.
6970 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6972 PR rtl-optimization/81611
6973 * auto-inc-dec.c (attempt_change): Move dead note from
6974 mem_insn if it's the next use of regno
6975 (find_address): Take address use of reg holding
6976 non-incremented value. Add parm to limit search to the named
6978 (merge_in_block): Attempt to use a mem insn that is the next
6979 use of the original regno.
6981 2018-02-27 Martin Sebor <msebor@redhat.com>
6984 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6985 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6987 2018-02-27 Martin Sebor <msebor@redhat.com>
6989 PR translation/84207
6990 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6991 n argument to unsigned HOST_WIDE_INT.
6992 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6993 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6994 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6995 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6997 2018-02-27 Richard Biener <rguenther@suse.de>
6999 PR tree-optimization/84512
7000 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7001 Do not use the estimate returned from record_stmt_cost for
7002 the scalar iteration cost but sum properly using add_stmt_cost.
7004 2018-02-27 Richard Biener <rguenther@suse.de>
7006 PR tree-optimization/84466
7007 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7008 Adjust last change to less strictly validate use operands.
7010 2018-02-27 Martin Liska <mliska@suse.cz>
7012 PR gcov-profile/84548
7013 * gcov.c (process_file): Allow partial overlap and consider it
7014 also as group functions.
7015 (output_lines): Properly calculate range of lines for a group.
7017 2018-02-27 Martin Liska <mliska@suse.cz>
7019 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7020 'ggc' suffixes. Change first column width.
7021 (timer::print): Fix formatting of the column.
7023 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7025 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7026 preserve inline entry blocks for the sake of debug inline
7027 entry point markers alone.
7028 (remove_unused_locals): Suggest in comments a better place to
7029 force the preservation of inline entry blocks that are
7030 otherwise unused, but do not preserve them.
7032 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7034 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7036 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7039 * config/i386/constraints.md (Bs): Replace
7040 ix86_indirect_branch_register with
7041 TARGET_INDIRECT_BRANCH_REGISTER.
7043 * config/i386/i386.md (indirect_jump): Likewise.
7044 (tablejump): Likewise.
7045 (*sibcall_memory): Likewise.
7046 (*sibcall_value_memory): Likewise.
7047 Peepholes of indirect call and jump via memory: Likewise.
7048 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7049 (*sibcall_value_GOT_32): Likewise.
7050 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7051 (GOT_memory_operand): Likewise.
7052 (call_insn_operand): Likewise.
7053 (sibcall_insn_operand): Likewise.
7054 (GOT32_symbol_operand): Likewise.
7055 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7057 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7059 PR rtl-optimization/83496
7060 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7061 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7062 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7063 redundant insn, if any.
7064 (relax_delay_slots): Likewise.
7065 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7067 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7069 PR tree-optimization/83965
7070 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7071 that grouped statements are part of a reduction chain. Return
7072 true if the statement is not marked as a reduction itself but
7074 (vect_recog_dot_prod_pattern): Don't check whether the statement
7075 is part of a group here.
7076 (vect_recog_sad_pattern): Likewise.
7077 (vect_recog_widen_sum_pattern): Likewise.
7079 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7082 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7084 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7087 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7089 (ix86_output_indirect_function_return): New prototype.
7090 (ix86_split_simple_return_pop_internal): Likewise.
7091 * config/i386/i386.c (indirect_return_via_cx): New.
7092 (indirect_return_via_cx_bnd): Likewise.
7093 (indirect_thunk_name): Handle return va CX_REG.
7094 (output_indirect_thunk_function): Create alias for
7095 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7096 (ix86_output_indirect_jmp): Remove the bool argument.
7097 (ix86_output_indirect_function_return): New function.
7098 (ix86_split_simple_return_pop_internal): Likewise.
7099 * config/i386/i386.md (*indirect_jump): Don't pass false
7100 to ix86_output_indirect_jmp.
7101 (*tablejump_1): Likewise.
7102 (simple_return_pop_internal): Change it to define_insn_and_split.
7103 Call ix86_split_simple_return_pop_internal to split it for
7105 (simple_return_indirect_internal): Call
7106 ix86_output_indirect_function_return instead of
7107 ix86_output_indirect_jmp.
7109 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7112 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7113 memset and value initialization afterwards.
7115 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7117 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7119 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7122 * common/config/aarch64/aarch64-common.c
7123 (aarch_option_optimization_table[]): Switch
7124 off fomit-frame-pointer
7126 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7127 Chung-Ju Wu <jasonwucj@gmail.com>
7129 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7131 (store_multiple): Ditto.
7133 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7135 * config.gcc: Add --with-cpu support for nds32 target.
7136 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7137 * config/nds32/nds32.opt: Add -mcpu= option.
7139 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7141 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7142 isel=yes): Warn for these deprecated options.
7144 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7146 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7147 ISA_2_5_MASKS_EMBEDDED.
7149 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7151 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7152 p->max as pointers rather than using iterative_hash_expr.
7154 2018-02-23 Carl Love <cel@us.ibm.com>
7156 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7157 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7159 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7160 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7161 P8V_BUILTIN_VEC_VUNSIGNED2.
7163 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7166 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7167 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7169 * lra-constraints.c (curr_insn_transform): Set up
7170 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7172 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7173 * lra-eliminations.c (spill_pseudos): Ditto.
7174 (process_insn_for_elimination): Ditto.
7175 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7176 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7177 LRA_NON_CLOBBERED_ALT.
7179 2018-02-22 Martin Sebor <msebor@redhat.com>
7181 PR tree-optimization/84480
7182 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7183 to maybe_diag_stxncpy_trunc. Call it.
7184 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7185 from gimple_fold_builtin_strcpy. Print inlining stack.
7186 (handle_builtin_stxncpy): Print inlining stack.
7187 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7189 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7192 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7193 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7194 and -fcheck-pointer-bounds are used together.
7195 (indirect_thunk_prefix): New enum.
7196 (indirect_thunk_need_prefix): New function.
7197 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7198 "_nt" instead of "_bnd" for NOTRACK prefix.
7199 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7200 (output_indirect_thunk_function): Likewise.
7202 (ix86_code_end): Update output_indirect_thunk_function calls.
7203 (ix86_output_indirect_branch_via_reg): Replace
7204 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7205 (ix86_output_indirect_branch_via_push): Likewise.
7206 (ix86_output_function_return): Likewise.
7207 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7208 incompatible with -fcf-protection=branch and
7209 -fcheck-pointer-bounds.
7211 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7214 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7215 Change gcc_assert call to output_operand_lossage.
7217 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7219 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7221 2018-02-22 DJ Delorie <dj@redhat.com>
7222 Sebastian Perta <sebastian.perta@renesas.com>
7223 Oleg Endo <olegendo@gcc.gnu.org>
7225 * config/rx/rx.c (rx_rtx_costs): New function.
7226 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7228 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7230 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7232 2018-02-22 Martin Liska <mliska@suse.cz>
7235 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7236 Add "native" as a possible value.
7238 2018-02-22 Martin Liska <mliska@suse.cz>
7241 * config/i386/i386.c (ix86_option_override_internal):
7242 Add "native" as a possible value for -march and -mtune.
7244 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7247 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7248 to all type variants.
7250 PR tree-optimization/84503
7251 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7252 width as info->bitpos + info->bitsize - start.
7253 (merged_store_group::merge_overlapping): Simplify width computation.
7254 (check_no_overlap): New function.
7255 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7256 start + width and last_order of the group, fail if check_no_overlap
7258 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7259 to group if check_no_overlap fails.
7261 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7263 * config/rs6000/altivec.md: Delete contraint arguments to
7264 define_expand, define_split, and define_peephole2, and in
7265 define_insn_and_split if always unused.
7266 * config/rs6000/darwin.md: Ditto.
7267 * config/rs6000/dfp.md: Ditto.
7268 * config/rs6000/rs6000.md: Ditto.
7269 * config/rs6000/sync.md: Ditto.
7270 * config/rs6000/vector.md: Ditto.
7271 * config/rs6000/vsx.md: Ditto.
7273 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7275 * config/rs6000/altivec.md: Write output control strings as braced
7276 blocks instead of double-quoted strings.
7277 * config/rs6000/darwin.md: Ditto.
7278 * config/rs6000/rs6000.md: Ditto.
7279 * config/rs6000/vector.md: Ditto.
7280 * config/rs6000/vsx.md: Ditto.
7282 2018-02-21 Jason Merrill <jason@redhat.com>
7284 PR c++/84314 - ICE with templates and fastcall attribute.
7285 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7287 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7289 * ipa-cp.c (determine_versionability): Fix comment typos.
7291 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7294 * ipa-cp.c (determine_versionability): Do not version functions caling
7297 2018-02-21 Martin Liska <mliska@suse.cz>
7300 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7301 Add "native" as a possible value.
7302 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7303 the macro when native cpu detection is available.
7305 2018-02-21 Martin Liska <mliska@suse.cz>
7308 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7309 Add "native" as a possible value.
7310 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7311 when native cpu detection is available.
7313 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7314 Martin Sebor <msebor@redhat.com>
7316 PR tree-optimization/84478
7317 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7319 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7320 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7321 support which is conservatively correct, for 2 only stay conservative
7322 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7323 argument to the 2 argument get_range_strlen, adjust 6 arg
7324 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7326 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7327 (gimple_fold_builtin_strlen): Pass true as last argument to
7330 2018-02-20 Martin Sebor <msebor@redhat.com>
7333 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7334 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7335 (builtin_memref::builtin_memref): Factor out parts into
7336 set_base_and_offset and call it.
7338 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7341 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7342 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7343 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7344 search at the associated MODE_INT.
7346 2018-02-20 Jeff Law <law@redhat.com>
7349 PR tree-optimization/81592
7351 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7352 for range data rather than using global data.
7353 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7354 range data rather than using global data.
7355 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7356 pass it to children as needed.
7357 (struct directive::fmtresult): Similarly.
7358 (struct directive::set_width): Similarly.
7359 (struct directive::set_precision): Similarly.
7360 (format_integer, format_directive, parse_directive): Similarly.
7361 (format_none): Accept unnamed vr_values parameter.
7362 (format_percent, format_floating, format_character): Similarly.
7363 (format_string, format_plain): Similarly.
7364 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7365 the EVRP range analyzer for range data rather than using global data.
7366 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7367 gimple-ssa-evrp-analyze.h
7368 (class sprintf_dom_walker): Add after_dom_children member function.
7369 Add evrp_range_analyzer member.
7370 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7371 range analyzer as needed.
7372 (sprintf_dom_walker::after_dom_children): New member function.
7373 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7375 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7376 (evrp_range_analyzer::pop_to_marker): Likewise.
7378 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7380 PR tree-optimization/84419
7381 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7382 with the required type if its current type is compatible but
7385 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7388 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7389 after vectorization.
7391 2018-02-20 Martin Liska <mliska@suse.cz>
7394 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7395 possible values if we don't have a hint.
7397 2018-02-20 Martin Liska <mliska@suse.cz>
7401 * final.c (shorten_branches): Build align_tab array with one
7403 * opts.c (finish_options): Add alignment option limit check.
7404 (MAX_CODE_ALIGN): Likewise.
7405 (MAX_CODE_ALIGN_VALUE): Likewise.
7406 * doc/invoke.texi: Document maximum allowed option value for
7407 all -falign-* options.
7409 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7412 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7413 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7414 * var-tracking.c (emit_note_insn_var_location): Remove all references
7415 to NOTE_INSN_CALL_ARG_LOCATION.
7416 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7417 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7418 Use copy_rtx_if_shared.
7419 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7420 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7421 (dwarf2out_var_location): Remove handling of
7422 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7424 * final.c (final_scan_insn): Remove all references to
7425 NOTE_INSN_CALL_ARG_LOCATION.
7426 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7427 before dumping final insns.
7428 * except.c (emit_note_eh_region_end): Remove all references to
7429 NOTE_INSN_CALL_ARG_LOCATION.
7430 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7431 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7432 * config/arc/arc.c (hwloop_optimize): Likewise.
7433 * config/arm/arm.c (create_fix_barrier): Likewise.
7434 * config/s390/s390.c (s390_chunkify_start): Likewise.
7435 * config/sh/sh.c (find_barrier): Likewise.
7436 * config/i386/i386.c (rest_of_insert_endbranch,
7437 ix86_seh_fixup_eh_fallthru): Likewise.
7438 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7439 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7440 * config/frv/frv.c (frv_function_prologue): Likewise.
7441 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7443 (note_outside_basic_block_p): Remove all references to
7444 NOTE_INSN_CALL_ARG_LOCATION.
7445 * gengtype.c (adjust_field_rtx_def): Likewise.
7446 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7448 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7449 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7452 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7455 PR tree-optimization/84452
7456 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7457 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7460 2018-02-19 Martin Liska <mliska@suse.cz>
7463 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7465 2018-02-19 Martin Liska <mliska@suse.cz>
7466 Richard Sandiford <richard.sandiford@linaro.org>
7468 PR tree-optimization/82491
7469 * gimple-fold.c (get_base_constructor): Make earlier bail out
7472 2018-02-19 Carl Love <cel@us.ibm.com>
7474 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7475 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7477 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7478 P8V_BUILTIN_VEC_NEG.
7480 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7482 * config/rl78/rl78.md (movdf): New define expand.
7484 2018-02-19 Martin Liska <mliska@suse.cz>
7487 * doc/invoke.texi: Fix typo.
7488 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7490 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7492 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7493 handle rs6000_single_float and rs6000_double_float specially for
7496 2018-02-16 Jeff Law <law@redhat.com>
7498 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7501 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7504 * ipa-inline.c (inline_small_functions): Fix a typo.
7506 2018-02-16 Nathan Sidwell <nathan@acm.org>
7508 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7510 2018-02-16 Carl Love <cel@us.ibm.com>
7512 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7513 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7514 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7515 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7516 expansion to P8V_BUILTIN_VEC_FLOAT2.
7518 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7520 PR rtl-optimization/70023
7521 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7522 src_regno into account.
7524 2018-02-16 Carl Love <cel@us.ibm.com>
7526 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7527 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7528 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7529 * config/rs6000/rs6000.c: Remove case statements for
7530 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7531 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7532 and P9V_BUILTIN_VEC_VINSERT4B.
7533 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7534 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7535 * config/rs6000/vsx.md:
7536 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7539 2018-02-16 Carl Love <cel@us.ibm.com>
7541 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7543 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7545 * config/rs6000/rs6000-c.c: Add the definitions for
7546 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7547 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7548 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7549 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7550 definition for insert4b and define insn *insert3b_internal.
7551 * doc/extend.texi: Add documentation for vec_extract4b.
7553 2018-02-16 Nathan Sidwell <nathan@acm.org>
7555 * doc/extend.texi (Backwards Compatibility): Mention friend
7556 injection. Note for-scope is deprecated.
7557 * doc/invoke.texi (-ffriend-injection): Deprecate.
7559 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7561 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7562 that moved to I2, also allow destinations that are a paradoxical
7563 subreg (instead of a normal reg).
7565 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7568 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7571 2018-02-16 Richard Biener <rguenther@suse.de>
7573 PR tree-optimization/84037
7574 PR tree-optimization/84016
7576 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7577 Adjust vec_construct for the fact we need additional higher latency
7578 128bit inserts for AVX256 and AVX512 vector builds.
7579 (ix86_add_stmt_cost): Scale vector construction cost for
7582 2018-02-16 Richard Biener <rguenther@suse.de>
7584 PR tree-optimization/84417
7585 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7586 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7587 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7589 2018-02-16 Martin Liska <mliska@suse.cz>
7592 * internal-fn.def (ASAN_CHECK): Set proper flags.
7593 (ASAN_MARK): Likewise.
7595 2018-02-16 Julia Koval <julia.koval@intel.com>
7597 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7598 from PTA_CANNONLAKE.
7600 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7603 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7604 Use ++iter rather than iter++ for std::list iterators.
7605 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7606 defer deleting them until all nodes in the forest are processed. Do
7607 free even leaf nodes. Change to_process into auto_vec.
7610 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7611 * vec.h (vec_default_construct): Use memset instead of placement new
7612 if BROKEN_VALUE_INITIALIZATION is defined.
7613 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7614 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7617 PR rtl-optimization/83723
7618 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7619 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7620 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7622 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7624 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7626 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7628 PR rtl-optimization/81443
7629 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7630 from inner REGs to paradoxical SUBREGs.
7632 2018-02-16 Richard Biener <rguenther@suse.de>
7634 PR tree-optimization/84399
7635 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7636 For operands we can analyze at their definition make sure we can
7637 analyze them at each use as well.
7639 2018-02-16 Richard Biener <rguenther@suse.de>
7641 PR tree-optimization/84190
7642 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7643 volatile accesses if the decl isn't volatile.
7645 2018-02-15 Jason Merrill <jason@redhat.com>
7647 PR c++/84314 - ICE with templates and fastcall attribute.
7648 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7649 TYPE_CANONICAL on an existing type.
7651 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7653 PR tree-optimization/84383
7654 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7655 dstoff nor call operand_equal_p if dstbase is NULL.
7657 PR tree-optimization/84334
7658 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7659 also a CONSTANT_CLASS_P, punt.
7661 2018-02-14 Jim Wilson <jimw@sifive.com>
7663 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7664 first SMALL_OPERAND check. New local min_second_step. Move assert
7665 to where locals are set. Add TARGET_RVC support.
7666 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7668 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7670 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7672 2018-02-14 Martin Sebor <msebor@redhat.com>
7674 PR tree-optimization/83698
7675 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7676 arrays constrain the offset range to their bounds.
7677 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7678 (builtin_access::overlap): Avoid setting the size of overlap if it's
7680 (maybe_diag_overlap): Also consider arrays when deciding what values
7681 of offsets to include in diagnostics.
7683 2018-02-14 Martin Sebor <msebor@redhat.com>
7686 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7687 that correspond to the kind of a declaration.
7689 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7692 * config/pa/pa.md: Load address of PIC label using the linkage table
7693 if the label is nonlocal.
7695 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7697 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7698 warning message if user requests -maltivec=be.
7699 * doc/invoke.texi: Document deprecation of -maltivec=be.
7701 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7704 * config/rs6000/rs6000-c.c: Update definitions for
7705 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7706 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7708 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7711 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7712 add _get_ssp intrinsics. Remove argument from
7713 __builtin_ia32_rdssp[d|q].
7714 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7715 * config/i386/i386-builtin.def: Remove argument from
7716 __builtin_ia32_rdssp[d|q].
7717 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7718 ix86_expand_special_args_builtin for _rdssp[d|q].
7719 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7720 Clear register before usage.
7721 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7722 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7724 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7726 PR tree-optimization/84357
7727 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7728 operand 1 of an ARRAY_REF too.
7730 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7733 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7734 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7736 (set_of_reg): New struct.
7737 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7738 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7739 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7741 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7742 Split into bitclr, bitset, bitinvert patterns if appropriate.
7743 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7744 use rx_fuse_in_memory_bitop.
7745 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7746 to named insn, correct maximum insn length.
7748 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7751 * machmode.def: Define a complex mode for PARTIAL_INT.
7752 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7754 * doc/rtl.texi: Document CSPImode.
7755 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7757 (msp430_hard_regno_nregs_with_padding): Likewise.
7759 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7762 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7764 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7766 PR rtl-optimization/84169
7767 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7768 we generated a parallel as new i3 and we split that to new i2 and i3
7769 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7770 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7771 those to i2, not i1. Partially rewrite this scan code.
7773 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7776 * stor-layout.c (place_field): For variable length fields, adjust
7777 offset_align afterwards not just based on the field's alignment,
7778 but also on the size.
7781 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7782 of exps and logs in the use_exp2 case.
7784 2018-02-13 Jeff Law <law@redhat.com>
7786 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7789 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7792 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7796 * common.opt (gas-loc-support, gas-locview-support): New.
7797 (ginline-points, ginternal-reset-location-views): New.
7798 * doc/invoke.texi: Document them. Use @itemx where intended.
7799 (gvariable-location-views): Adjust.
7800 * target.def (reset_location_view): New.
7801 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7802 (TARGET_RESET_LOCATION_VIEW): New.
7803 * doc/tm.texi: Rebuilt.
7804 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7805 (dwarf2out_default_as_locview_support): New.
7806 (output_asm_line_debug_info): Use option variables.
7807 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7808 (output_loc_list): Likewise.
7809 (add_high_low_attributes): Check option variables.
7810 Don't output entry view attribute in strict mode.
7811 (gen_inlined_subroutine_die): Check option variables.
7812 (dwarf2out_inline_entry): Likewise.
7813 (init_sections_and_labels): Likewise.
7814 (dwarf2out_early_finish): Likewise.
7815 (maybe_reset_location_view): New, from...
7816 (dwarf2out_var_location): ... here. Call it.
7817 * debug.h (dwarf2out_default_as_loc_support): Declare.
7818 (dwarf2out_default_as_locview_support): Declare.
7819 * hooks.c (hook_int_rtx_insn_0): New.
7820 * hooks.h (hook_int_rtx_insn_0): Declare.
7821 * toplev.c (process_options): Take -gas-loc-support and
7822 -gas-locview-support from dwarf2out. Enable
7823 -gvariable-location-views by default only with locview
7824 assembler support. Enable -ginternal-reset-location-views by
7825 default only if the target defines the corresponding hook.
7826 Enable -ginline-points by default if location views are
7827 enabled; force it disabled if statement frontiers are
7829 * tree-inline.c (expand_call_inline): Check option variables.
7830 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7832 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7834 PR tree-optimization/84321
7835 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7836 handling. Also check whether the anti-range contains any values
7837 that satisfy the mask; switch to a VR_RANGE if not.
7839 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7842 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7844 2018-02-13 Martin Jambor <mjambor@suse.cz>
7847 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7848 of call statements, also set location of a load to a temporary.
7850 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7852 * config/rl78/rl78.c (add_vector_labels): New function.
7853 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7854 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7855 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7856 which checks that no arguments are passed.
7857 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7858 * doc/extend.texi: Documentation for the new attribute.
7860 2018-02-13 Andreas Schwab <schwab@suse.de>
7862 * config/riscv/linux.h (CPP_SPEC): Define.
7864 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7867 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7868 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7869 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7870 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7871 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7872 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7873 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7874 temporarily for AES and PCLMUL builtins.
7876 PR tree-optimization/84339
7877 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7878 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7882 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7883 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7884 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7886 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7888 * omp-simd-clone.h: New file.
7889 * omp-simd-clone.c: Include omp-simd-clone.h.
7890 (expand_simd_clones): No longer static.
7891 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7892 cgraph.h and omp-simd-clone.h.
7893 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7894 (vect_recog_widen_shift_pattern): Formatting fix.
7895 (vect_pattern_recog_1): Don't check optab for calls.
7898 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7899 operands[2] into a REG before using gen_lowpart on it.
7901 2018-02-12 Jeff Law <law@redhat.com>
7904 * config/sh/sh.c (find_barrier): Consider a sibling call
7907 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7908 successfully back substituting a reg.
7910 2018-02-12 Richard Biener <rguenther@suse.de>
7912 PR tree-optimization/84037
7913 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7914 parameter, move visited init to caller.
7915 (vect_slp_analyze_operations): Separate cost from validity
7916 check, initialize visited once for all instances.
7917 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7919 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7921 (vect_model_promotion_demotion_cost): Likewise.
7922 (vectorizable_bswap): Guard cost modeling with !slp_node
7923 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7925 (vectorizable_call): Likewise.
7926 (vectorizable_conversion): Likewise.
7927 (vectorizable_assignment): Likewise.
7928 (vectorizable_shift): Likewise.
7929 (vectorizable_operation): Likewise.
7930 (vectorizable_store): Likewise.
7931 (vectorizable_load): Likewise.
7932 (vectorizable_condition): Likewise.
7933 (vectorizable_comparison): Likewise.
7935 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7938 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7939 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7940 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7942 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7945 * params.def (inline-min-speedup): Increase from 8 to 15.
7946 (max-inline-insns-auto): Decrease from 40 to 30.
7947 * ipa-split.c (consider_split): Add some buffer for function to
7948 be considered inlining candidate.
7949 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7952 2018-02-12 Richard Biener <rguenther@suse.de>
7954 PR tree-optimization/84037
7955 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7956 matched stmts if we cannot swap the non-matched ones.
7958 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7960 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7961 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7962 _mm_maskz_scalef_round_ss): New intrinsics.
7963 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7964 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7965 __builtin_ia32_scalefss_round): Remove.
7966 (__builtin_ia32_scalefsd_mask_round,
7967 __builtin_ia32_scalefss_mask_round): New intrinsics.
7968 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7969 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7970 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7971 "<round_constraint>")): Changed to ...
7972 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7973 "<round_scalar_constraint>")): ... this.
7974 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7975 %0, %1, %2<round_op3>}"): Changed to ...
7976 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7977 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7978 %2<round_scalar_mask_op3>}"): ... this.
7979 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7981 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7983 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7984 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7985 (_mm_maskz_sqrt_round_ss): New intrinsics.
7986 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7987 (__builtin_ia32_sqrtsd_mask_round)
7988 (__builtin_ia32_sqrtss_mask_round): New builtins.
7989 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7990 (__builtin_ia32_sqrtss_round): Remove.
7991 (__builtin_ia32_sqrtsd_mask_round)
7992 (__builtin_ia32_sqrtss_mask_round): New builtins.
7993 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7994 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7995 ((match_operand:VF_128 1 "vector_operand"
7996 "xBm,<round_constraint>")): Changed to ...
7997 ((match_operand:VF_128 1 "vector_operand"
7998 "xBm,<round_scalar_constraint>")): ... this.
7999 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8000 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8001 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8002 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8003 %<iptr>1<round_scalar_mask_op3>}): ... this.
8004 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8005 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8007 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8010 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8011 Cast vec_cmpeq result to correct type.
8012 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8013 Cast vec_cmpgt result to correct type.
8015 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8017 * final.c (final_scan_insn_1): Renamed from...
8018 (final_scan_insn): ... this. New wrapper, to recover
8019 seen from the outermost call in recursive ones.
8020 * config/sparc/sparc.c (output_return): Drop seen from call.
8021 (output_sibcall): Likewise.
8022 * config/visium/visium.c (output_branch): Likewise.
8024 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8026 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8029 2018-02-10 Alan Modra <amodra@gmail.com>
8032 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8033 Specify LR as an input.
8035 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8038 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8039 remove_member_access_dummy_vars): New functions.
8040 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8041 lower_omp_1, execute_lower_omp): Use them.
8043 PR rtl-optimization/84308
8044 * shrink-wrap.c (spread_components): Release todo vector.
8046 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8048 PR rtl-optimization/57193
8049 * ira-color.c (struct allocno_color_data): Add member
8050 conflict_allocno_hard_prefs.
8051 (update_conflict_allocno_hard_prefs): New.
8052 (bucket_allocno_compare_func): Add a preference based on
8053 conflict_allocno_hard_prefs.
8054 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8055 (color_allocnos): Remove a dead code. Initiate
8056 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8058 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8061 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8062 constraint from =wa to wa. Avoid a subreg on the output operand,
8063 instead use a pseudo and subreg it in a move.
8064 (p9_xxbrd_<mode>): Changed to ...
8065 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8066 (p9_xxbrd_v2df): New expander.
8067 (p9_xxbrw_<mode>): Changed to ...
8068 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8069 (p9_xxbrw_v4sf): New expander.
8071 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8073 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8075 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8078 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8079 multiply in 32-bit mode.
8080 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8081 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8084 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8086 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8087 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8088 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8089 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8091 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8094 * dwarf2out.c (is_trivial_indirect_ref): New function.
8095 (dwarf2out_late_global_decl): Do not generate a location
8096 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8097 and that are not defined in the current unit.
8099 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8101 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8102 instead of a libcall for UNORDERED.
8104 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8107 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8108 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8110 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8113 * config/s390/s390.c (s390_set_current_function): Invoke
8114 s390_indirect_branch_settings also if fndecl didn't change.
8116 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8118 * config/rs6000/rs6000.md (blockage): Set length to zero.
8120 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8122 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8124 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8127 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8128 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8132 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8133 PARALLEL incoming that failed vt_get_decl_and_offset check.
8136 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8137 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8139 (get_variable_section): For decls in named .bss* sections pass true as
8140 second argument to bss_initializer_p.
8142 2018-02-09 Marek Polacek <polacek@redhat.com>
8143 Jakub Jelinek <jakub@redhat.com>
8146 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8147 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8148 Sync some changes from cxx_fold_indirect_ref.
8150 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8152 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8154 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8155 (BLOCK_INLINE_ENTRY_LABEL): New.
8156 (dwarf2out_var_location): Disregard inline entry markers.
8157 (inline_entry_data): New struct.
8158 (inline_entry_data_hasher): New hashtable type.
8159 (inline_entry_data_hasher::hash): New.
8160 (inline_entry_data_hasher::equal): New.
8161 (inline_entry_data_table): New variable.
8162 (add_high_low_attributes): Add DW_AT_entry_pc and
8163 DW_AT_GNU_entry_view attributes if a pending entry is found
8164 in inline_entry_data_table. Add old entry_pc attribute only
8165 if debug nonbinding markers are disabled.
8166 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8167 markers are enabled.
8168 (block_within_block_p, dwarf2out_inline_entry): New.
8169 (dwarf2out_finish): Check that no entries remained in
8170 inline_entry_data_table.
8171 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8172 (final_scan_insn, notice_source_line): Likewise.
8173 (rest_of_clean_state): Skip inline entry markers.
8174 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8176 * gimple.c (gimple_build_debug_inline_entry): New.
8177 * gimple.h (enum gimple_debug_subcode): Add
8178 GIMPLE_DEBUG_INLINE_ENTRY.
8179 (gimple_build_debug_inline_entry): Declare.
8180 (gimple_debug_inline_entry_p): New.
8181 (gimple_debug_nonbind_marker_p): Adjust.
8182 * insn-notes.def (INLINE_ENTRY): New.
8183 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8184 inline entry marker notes.
8185 (print_insn): Likewise.
8186 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8187 (INSN_DEBUG_MARKER_KIND): Likewise.
8188 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8189 * tree-inline.c (expand_call_inline): Build and insert
8190 debug_inline_entry stmt.
8191 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8192 inline entry blocks early, if nonbind markers are enabled.
8193 (dump_scope_block): Dump fragment info.
8194 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8195 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8196 (gimple_build_debug_inline_entry): New.
8197 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8198 Enable/disable inline entry points too.
8199 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8200 (DEBUG_INSN): Describe inline entry markers.
8202 * common.opt (gvariable-location-views): New.
8203 (gvariable-location-views=incompat5): New.
8204 * config.in: Rebuilt.
8205 * configure: Rebuilt.
8206 * configure.ac: Test assembler for view support.
8207 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8208 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8209 * dwarf2out.c (var_loc_view): New typedef.
8210 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8211 (dwarf2out_locviews_in_attribute): New.
8212 (dwarf2out_locviews_in_loclist): New.
8213 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8214 (enum dw_line_info_opcode): Add LI_adv_address.
8215 (struct dw_line_info_table): Add view.
8216 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8217 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8218 (zero_view_p): New variable.
8219 (ZERO_VIEW_P): New macro.
8220 (output_asm_line_debug_info): New.
8221 (struct var_loc_node): Add view.
8222 (add_AT_view_list, AT_loc_list): New.
8223 (add_var_loc_to_decl): Add view param. Test it against last.
8224 (new_loc_list): Add view params. Record them.
8225 (AT_loc_list_ptr): Handle loc and view lists.
8226 (view_list_to_loc_list_val_node): New.
8227 (print_dw_val): Handle dw_val_class_view_list.
8228 (size_of_die): Likewise.
8229 (value_format): Likewise.
8230 (loc_list_has_views): New.
8231 (gen_llsym): Set vl_symbol too.
8232 (maybe_gen_llsym, skip_loc_list_entry): New.
8233 (dwarf2out_maybe_output_loclist_view_pair): New.
8234 (output_loc_list): Output view list or entries too.
8235 (output_view_list_offset): New.
8236 (output_die): Handle dw_val_class_view_list.
8237 (output_dwarf_version): New.
8238 (output_compilation_unit_header): Use it.
8239 (output_skeleton_debug_sections): Likewise.
8240 (output_rnglists, output_line_info): Likewise.
8241 (output_pubnames, output_aranges): Update version comments.
8242 (output_one_line_info_table): Output view numbers in asm comments.
8243 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8244 Call maybe_gen_llsym.
8245 (loc_list_from_tree_1): Adjust.
8246 (add_AT_location_description): Create view list attribute if
8247 needed, check it's absent otherwise.
8248 (convert_cfa_to_fb_loc_list): Adjust.
8249 (maybe_emit_file): Call output_asm_line_debug_info for test.
8250 (dwarf2out_var_location): Reset views as needed. Precompute
8251 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8252 attribute. Set view.
8253 (new_line_info_table): Reset next view.
8254 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8255 (dwarf2out_source_line): Likewise. Output view resets and labels to
8256 the assembler, or select appropriate line info opcodes.
8257 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8258 (optimize_string_length): Catch it. Adjust.
8259 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8260 dw_val_class_view_list, and remove it if no longer needed.
8261 (hash_loc_list): Hash view numbers.
8262 (loc_list_hasher::equal): Compare them.
8263 (optimize_location_lists): Check whether a view list symbol is
8264 needed, and whether the locview attribute is present, and
8265 whether they match. Remove the locview attribute if no longer
8267 (index_location_lists): Call skip_loc_list_entry for test.
8268 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8269 Use output_dwarf_version.
8270 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8271 (struct dw_val_node): Add val_view_list.
8272 * final.c (SEEN_NEXT_VIEW): New.
8273 (set_next_view_needed): New.
8274 (clear_next_view_needed): New.
8275 (maybe_output_next_view): New.
8276 (final_start_function): Rename to...
8277 (final_start_function_1): ... this. Take pointer to FIRST,
8278 add SEEN parameter. Emit param bindings in the initial view.
8279 (final_start_function): Reintroduce SEEN-less interface.
8280 (final): Rename to...
8281 (final_1): ... this. Take SEEN parameter. Output final pending
8282 next view at the end.
8283 (final): Reintroduce seen-less interface.
8284 (final_scan_insn): Output pending next view before switching
8285 sections or ending a block. Mark the next view as needed when
8286 outputting variable locations. Notify debug backend of section
8287 changes, and of location view changes.
8288 (rest_of_handle_final): Adjust.
8289 * toplev.c (process_options): Autodetect value for debug variable
8290 location views option. Warn on incompat5 without -gdwarf-5.
8291 * doc/invoke.texi (gvariable-location-views): New.
8292 (gvariable-location-views=incompat5): New.
8293 (gno-variable-location-views): New.
8295 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8297 PR tree-optimization/84136
8298 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8299 that the result of find_edge is non-NULL.
8301 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8304 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8305 storing integer register in SImode. Fix cost of 256 and 512
8306 byte aligned SSE register store.
8308 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8310 * config/i386/i386.c (ix86_multiplication_cost): Fix
8311 multiplication cost for TARGET_AVX512DQ.
8313 2018-02-08 Marek Polacek <polacek@redhat.com>
8315 PR tree-optimization/84238
8316 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8319 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8321 PR tree-optimization/84265
8322 * tree-vect-stmts.c (vectorizable_store): Don't treat
8323 VMAT_CONTIGUOUS accesses as grouped.
8324 (vectorizable_load): Likewise.
8326 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8328 PR tree-optimization/81635
8329 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8330 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8331 (test_round_for_mask): New functions.
8332 (wide_int_cc_tests): Call test_round_for_mask.
8333 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8334 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8335 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8336 range returned by get_range_info.
8338 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8341 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8342 * symtab.c: Include builtins.h
8343 (symtab_node::output_to_lto_symbol_table_p): Move here
8344 from lto-streamer-out.c:output_symbol_p.
8345 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8346 (output_symbol_p): Move all logic to symtab.c
8347 (produce_symtab): Update.
8349 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8351 * config/s390/s390-opts.h (enum indirect_branch): Define.
8352 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8353 (s390_indirect_branch_via_thunk)
8354 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8355 (enum s390_indirect_branch_type): Define.
8356 * config/s390/s390.c (struct s390_frame_layout, struct
8357 machine_function): Remove.
8358 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8359 (indirect_branch_table_label_no, indirect_branch_table_name):
8361 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8362 (enum s390_indirect_branch_option): Define.
8363 (s390_return_addr_from_memory): New function.
8364 (s390_handle_string_attribute): New function.
8365 (s390_attribute_table): Add new attribute handler.
8366 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8367 (s390_indirect_branch_via_thunk): New function.
8368 (s390_indirect_branch_via_inline_thunk): New function.
8369 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8370 sibling call optimization for non z10 compiles.
8371 (s390_emit_call): Force indirect branch target to be a single
8372 register. Add r1 clobber for non-z10 compiles.
8373 (s390_emit_epilogue): Emit return jump via return_use expander.
8374 (s390_reorg): Handle JUMP_INSNs as execute targets.
8375 (s390_option_override_internal): Perform validity checks for the
8376 new command line options.
8377 (s390_indirect_branch_attrvalue): New function.
8378 (s390_indirect_branch_settings): New function.
8379 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8380 (s390_output_indirect_thunk_function): New function.
8381 (s390_code_end): Implement target hook.
8382 (s390_case_values_threshold): Implement target hook.
8383 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8385 * config/s390/s390.h (struct s390_frame_layout)
8386 (struct machine_function): Move here from s390.c.
8387 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8388 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8389 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8390 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8391 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8392 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8393 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8394 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8395 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8396 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8397 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8398 (mnemonic attribute): Add values which aren't recognized
8400 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8401 pattern for branch conversion. Fix mnemonic attribute.
8402 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8403 indirect branch via thunk if requested.
8404 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8405 ("*indirect_jump"): Disable for branch conversion using out of
8407 ("indirect_jump_via_thunk<mode>_z10")
8408 ("indirect_jump_via_thunk<mode>")
8409 ("indirect_jump_via_inlinethunk<mode>_z10")
8410 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8411 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8412 ("casesi_jump_via_inlinethunk<mode>_z10")
8413 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8414 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8415 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8416 ("*indirect2_jump"): Disable for branch conversion.
8417 ("casesi_jump"): Turn into expander and expand patterns for branch
8419 ("return_use"): New expander.
8420 ("*return"): Emit return via thunk and rename it to ...
8421 ("*return<mode>"): ... this one.
8422 * config/s390/s390.opt: Add new options and and enum for the
8425 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8427 * lra-constraints.c (match_reload): Unconditionally use
8428 gen_lowpart_SUBREG, rather than selecting between that
8429 and equivalent gen_rtx_SUBREG code.
8431 2018-02-08 Richard Biener <rguenther@suse.de>
8433 PR tree-optimization/84233
8434 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8435 changed flag instead of boguously re-using phi_inserted.
8437 2018-02-08 Martin Jambor <mjambor@suse.cz>
8439 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8440 static local variables.
8442 2018-02-08 Richard Biener <rguenther@suse.de>
8444 PR tree-optimization/84278
8445 * tree-vect-stmts.c (vectorizable_store): When looking for
8446 smaller vector types to perform grouped strided loads/stores
8447 make sure the mode is supported by the target.
8448 (vectorizable_load): Likewise.
8450 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8452 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8453 Increase LDP/STP opportunities by adding adjacent callee-saves.
8455 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8457 PR rtl-optimization/84068
8458 PR rtl-optimization/83459
8459 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8461 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8463 PR tree-optimization/84224
8464 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8465 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8468 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8471 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8472 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8473 count, remove one USE.
8475 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8477 * doc/install.texi (Configuration): Document the
8478 --with-long-double-format={ibm,ieee} PowerPC configuration
8482 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8483 Convert from define_expand to be define_insn_and_split. Rework
8484 float/double/_Float128 conversions to QI/HI/SImode to work with
8485 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8486 conversions to QI/HImode types did a store and then a load to
8487 truncate the value. For conversions to VSX registers, don't split
8488 the insn, instead emit the code directly. Use the code iterator
8489 any_fix to combine signed and unsigned conversions.
8490 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8491 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8492 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8493 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8494 (fix_<mode>di2_hw): Likewise.
8495 (fixuns_<mode>di2_hw): Likewise.
8496 (fix_<mode>si2_hw): Likewise.
8497 (fixuns_<mode>si2_hw): Likewise.
8498 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8499 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8500 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8501 fix<uns>_trunc<SFDF:mode>si2_p8.
8502 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8504 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8505 (fix<uns>_<mode>_mem): Likewise.
8506 (fctiw<u>z_<mode>_mem): Likewise.
8507 (fix<uns>_<mode>_mem): Likewise.
8508 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8509 the register allocator from doing a direct move to the GPRs to do
8510 a store, and instead use the ISA 3.0 store byte/half-word from
8511 vector register instruction. For IEEE 128-bit floating point,
8512 also optimize stores of 32-bit ints.
8513 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8515 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8517 * genextract.c (push_pathstr_operand): New function to support
8519 (walk_rtx): Call push_pathstr_operand.
8520 (print_path): Support [a-zA-Z].
8522 2018-02-07 Richard Biener <rguenther@suse.de>
8524 PR tree-optimization/84037
8525 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8526 (cse_and_gimplify_to_preheader): Declare.
8527 (vect_get_place_in_interleaving_chain): Likewise.
8528 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8530 (_loop_vec_info::~_loop_vec_info): Delete it.
8531 (cse_and_gimplify_to_preheader): New function.
8532 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8533 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8534 (vectorizable_load): Likewise. For grouped stores always base
8535 the IV on the first element.
8536 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8537 condition before gimplifying.
8539 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8541 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8542 *DIV_EXPR and *MOD_EXPR.
8544 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8547 * config/i386/i386.c (ix86_option_override_internal): Mask out
8548 the CF_SET bit when checking -fcf-protection.
8550 2018-02-07 Tom de Vries <tom@codesourcery.com>
8553 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8556 2018-02-07 Richard Biener <rguenther@suse.de>
8558 PR tree-optimization/84204
8559 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8562 PR tree-optimization/84205
8563 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8564 special-case isl_ast_op_zdiv_r.
8566 PR tree-optimization/84223
8567 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8568 Only add conditions from within the region.
8569 (gather_bbs::after_dom_children): Adjust.
8571 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8574 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8575 * config/avr/avr.md: Only post-reload split REG-REG moves if
8576 either register is GENERAL_REG_P.
8578 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8580 PR tree-optimization/84235
8581 * tree-ssa-scopedtables.c
8582 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8583 if the subtraction is performed in floating point type where NaNs are
8584 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8585 build 1. Formatting fix.
8587 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8590 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8591 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8592 and skip it regardless of bb boundaries. Use CALL_P macro,
8593 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8594 unnecessarily, formatting fix.
8596 2018-02-06 Michael Collison <michael.collison@arm.com>
8598 * config/arm/thumb2.md:
8599 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8600 (*thumb_mov_notscc): Ditto.
8602 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8605 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8606 unsigned_fix, not "s".
8608 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8610 * configure.ac (gcc_fn_eh_frame_ro): New function.
8611 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8612 correct .eh_frame permissions.
8613 * configure: Regenerate.
8615 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8617 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8620 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8622 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8623 Display warning message for -mno-speculate-indirect-jumps.
8625 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8627 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8629 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8631 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8633 PR tree-optimization/84225
8634 * tree-eh.c (find_trapping_overflow): Only call
8635 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8637 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8640 * config/i386/i386.c: Reimplement the check of possible options
8641 -mibt/-mshstk conbination. Change error messages.
8642 * doc/invoke.texi: Fix a typo: remove extra '='.
8644 2018-02-06 Marek Polacek <polacek@redhat.com>
8646 PR tree-optimization/84228
8647 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8649 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8652 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8653 emitted arch directives.
8654 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8655 __ARM_FEATURE_COPROC before changing architectures.
8657 2018-02-06 Richard Biener <rguenther@suse.de>
8659 * config/i386/i386.c (print_reg): Fix typo.
8660 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8662 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8664 * configure: Regenerate.
8666 2018-02-05 Martin Sebor <msebor@redhat.com>
8668 PR tree-optimization/83369
8669 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8672 2018-02-05 Martin Liska <mliska@suse.cz>
8674 * doc/invoke.texi: Cherry-pick upstream r323995.
8676 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8678 * ira.c (ira_init_register_move_cost): Adjust comment.
8680 2018-02-05 Martin Liska <mliska@suse.cz>
8682 PR gcov-profile/84137
8683 * doc/gcov.texi: Fix typo in documentation.
8685 2018-02-05 Martin Liska <mliska@suse.cz>
8687 PR gcov-profile/83879
8688 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8689 using dlopen functionality.
8691 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8693 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8694 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8695 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8696 _mm_maskz_range_round_ss): New intrinsics.
8697 (__builtin_ia32_rangesd128_round)
8698 (__builtin_ia32_rangess128_round): Remove.
8699 (__builtin_ia32_rangesd128_mask_round,
8700 __builtin_ia32_rangess128_mask_round): New builtins.
8701 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8702 __builtin_ia32_rangess128_round): Remove.
8703 (__builtin_ia32_rangesd128_mask_round,
8704 __builtin_ia32_rangess128_mask_round): New builtins.
8705 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8706 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8707 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8708 "<round_saeonly_constraint>")): Changed to ...
8709 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8710 "<round_saeonly_scalar_constraint>")): ... this.
8711 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8712 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8713 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8714 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8715 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8717 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8719 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8721 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8722 Remove all values except native, 8540 and 8548.
8724 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8726 * config/i386/i386.c (ix86_output_function_return): Pass
8727 INVALID_REGNUM, instead of -1, as invalid register number to
8728 indirect_thunk_name and output_indirect_thunk.
8730 2018-02-02 Julia Koval <julia.koval@intel.com>
8732 * config.gcc: Add -march=icelake.
8733 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8734 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8735 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8736 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8737 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8738 (processor_target_table): Add icelake.
8739 (ix86_option_override_internal): Handle new PTAs.
8740 (get_builtin_code_for_version): Handle icelake.
8741 (M_INTEL_COREI7_ICELAKE): New.
8742 (fold_builtin_cpu): Handle icelake.
8743 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8744 * doc/invoke.texi: Add -march=icelake.
8746 2018-02-02 Julia Koval <julia.koval@intel.com>
8748 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8749 to wide_int_bitmask.
8750 * wide-int-bitmask.h: New.
8752 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8755 * config/i386/i386.md: Replace Pmode with word_mode in
8756 builtin_setjmp_setup and builtin_longjmp to support x32.
8758 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8762 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8764 (rs6000_supported_cpu_names): New static variable.
8765 (linux_cpu_translation_table): Likewise.
8766 (elf_platform) <cpu>: Define new static variable and use it.
8767 Translate kernel AT_PLATFORM name to canonical name if needed.
8768 Error if platform name is unknown.
8770 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8773 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8775 2018-02-01 Jeff Law <law@redhat.com>
8778 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8779 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8780 the scratch if RELEASE_VIA_POP is false.
8781 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8782 If we have to save a temporary register, decrement SIZE appropriately.
8783 Pass new arguments to release_scratch_register_on_entry.
8784 (ix86_adjust_stack_and_probe): Likewise.
8785 (ix86_emit_probe_stack_range): Pass new arguments to
8786 release_scratch_register_on_entry.
8788 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8790 PR rtl-optimization/84157
8791 * combine.c (change_zero_ext): Use REG_P predicate in
8792 front of HARD_REGISTER_P predicate.
8794 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8796 * config/avr/avr.c (avr_option_override): Move disabling of
8797 -fdelete-null-pointer-checks to...
8798 * common/config/avr/avr-common.c (avr_option_optimization_table):
8801 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8803 PR tree-optimization/81635
8804 * tree-data-ref.c (split_constant_offset_1): For types that
8805 wrap on overflow, try to use range info to prove that wrapping
8808 2018-02-01 Renlin Li <renlin.li@arm.com>
8811 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8813 (aarch64_register_move_cost): Likewise.
8814 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8816 (REG_CLASS_NAMES): Likewise.
8817 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8818 TAILCALL_ADDR_REGS. Remove IP registers.
8819 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8821 2018-02-01 Richard Biener <rguenther@suse.de>
8823 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8824 for specifying RPO order and allow NULL for that.
8825 * domwalk.c (dom_walker::dom_walker): Likewise.
8826 (dom_walker::walk): Handle NULL RPO order.
8827 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8829 (rewrite_update_dom_walker): Likewise.
8830 (mark_def_dom_walker): Likewise.
8832 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8834 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8835 (aarch64_maybe_expand_sve_subreg_move): Declare.
8836 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8837 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8839 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8840 that are semantically a reverse operation.
8841 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8842 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8843 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8845 (aarch64_can_change_mode_class): For big-endian, forbid changes
8846 between two SVE modes if they have different element sizes.
8848 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8850 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8851 the TImode handling for big-endian targets.
8853 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8855 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8856 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8858 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8859 Remove BSWAP handing for big-endian targets and use the form of
8860 LD1RQ appropariate for the mode.
8862 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8864 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8865 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8868 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8871 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8872 check for operands that need to go through aarch64_sve_reload_be.
8874 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8876 PR tree-optimization/81661
8877 PR tree-optimization/84117
8878 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8879 * tree-eh.c: Include gimplify.h.
8880 (find_trapping_overflow, replace_trapping_overflow,
8881 rewrite_to_non_trapping_overflow): New functions.
8882 * tree-vect-loop.c: Include tree-eh.h.
8883 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8884 * tree-data-ref.c: Include tree-eh.h.
8885 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8887 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8889 PR rtl-optimization/84123
8890 * combine.c (change_zero_ext): Check if hard register satisfies
8891 can_change_dest_mode before calling gen_lowpart_SUBREG.
8893 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8896 * ira.c (ira_init_register_move_cost): Remove assert.
8898 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8900 PR rtl-optimization/84071
8901 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8902 * doc/tm.texi: Regenerate.
8904 2018-01-31 Richard Biener <rguenther@suse.de>
8906 PR tree-optimization/84132
8907 * tree-data-ref.c (analyze_miv_subscript): Properly
8908 check whether evolution_function_is_affine_multivariate_p
8909 before calling gcd_of_steps_may_divide_p.
8911 2018-01-31 Julia Koval <julia.koval@intel.com>
8914 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8915 * config/i386/i386.md (rdpid_rex64) New.
8916 (rdpid): Make 32bit only.
8918 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8921 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8922 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8924 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8927 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8929 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8931 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8933 PR rtl-optimization/84071
8934 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8935 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8937 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8939 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8940 (arc_attribute_table): Add 'aux' attribute.
8941 (arc_in_small_data_p): Consider aux like variables.
8942 (arc_is_aux_reg_p): New function.
8943 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8944 (arc_get_aux_arg): New function.
8945 (prepare_move_operands): Handle aux-register access.
8946 (arc_handle_aux_attribute): New function.
8947 * doc/extend.texi (ARC Variable attributes): Add subsection.
8949 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8951 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8952 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8953 (arc_attribute_table): Add 'uncached' attribute.
8954 (arc_print_operand): Print '.di' flag for uncached memory
8956 (arc_in_small_data_p): Do not consider for small data the uncached
8958 (arc_is_uncached_mem_p): New function.
8959 * config/arc/predicates.md (compact_store_memory_operand): Check
8960 for uncached memory accesses.
8961 (nonvol_nonimm_operand): Likewise.
8962 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8964 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8967 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8968 falign-loops=): Add Optimization flag.
8970 2018-01-30 Jeff Law <law@redhat.com>
8973 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8974 INT_REGISTERS_SAVED. Check it prior to calling
8975 get_scratch_register_on_entry.
8976 (ix86_adjust_stack_and_probe): Similarly.
8977 (ix86_emit_probe_stack_range): Similarly.
8978 (ix86_expand_prologue): Corresponding changes.
8980 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8983 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8984 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8986 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8989 * lra-constraints.c (curr_insn_transform): Process AND in the
8992 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8994 PR rtl-optimization/83986
8995 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8996 dependence against last_pending_memory_flush in addition to
8999 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9001 PR tree-optimization/81611
9002 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9005 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9008 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9011 2018-01-30 Richard Biener <rguenther@suse.de>
9012 Jakub Jelinek <jakub@redhat.com>
9014 PR tree-optimization/84111
9015 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9016 inner loops added during recursion, as they don't have up-to-date
9019 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9022 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9023 (can_inline_edge_by_limits_p): ... here.
9024 (can_early_inline_edge_p, check_callers,
9025 update_caller_keys, update_callee_keys, recursive_inlining,
9026 add_new_edges_to_heap, speculation_useful_p,
9027 inline_small_functions,
9028 inline_small_functions, flatten_function,
9029 inline_to_all_callers_1): Update.
9031 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9033 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9036 2018-01-30 Richard Biener <rguenther@suse.de>
9038 PR tree-optimization/83008
9039 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9040 invariant and constant vector uses in stmts when they need
9043 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9046 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9047 * configure: Regenerate.
9049 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9051 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9053 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9054 Use gen_rtx_REG rather than gen_lowpart.
9056 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9058 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9059 rather than 0 when creating partial subregs.
9061 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9063 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9066 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9069 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9070 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9071 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9072 flags. This restores the settings used before the 2017-07-24.
9073 Turning off pre increment/decrement/modify allows IVOPTS to
9074 optimize DF/SF loops where the index is an int.
9076 2018-01-29 Richard Biener <rguenther@suse.de>
9077 Kelvin Nilsen <kelvin@gcc.gnu.org>
9080 * tree-vect-stmts.c (vectorizable_call): Don't call
9081 targetm.vectorize_builtin_md_vectorized_function if callee is
9084 2018-01-22 Carl Love <cel@us.ibm.com>
9086 * doc/extend.tex: Fix typo in second arg in
9087 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9089 2018-01-29 Richard Biener <rguenther@suse.de>
9091 PR tree-optimization/84086
9092 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9093 (flush_ssaname_freelist): When SSA names were released reset
9094 the SCEV hash table.
9096 2018-01-29 Richard Biener <rguenther@suse.de>
9098 PR tree-optimization/84057
9099 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9100 removed paths when removing edges.
9102 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9104 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9105 -mfunction-return=@var{choice}.
9107 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9110 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9112 (layout::print_source_line): Likewise.
9113 (test_get_line_width_without_trailing_whitespace): Add test cases.
9115 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9118 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9121 2018-01-26 Jim Wilson <jimw@sifive.com>
9123 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9125 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9128 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9130 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9131 and CMP + SUB-immediate -> SUBS.
9133 2018-01-26 Martin Sebor <msebor@redhat.com>
9135 PR tree-optimization/83896
9136 * tree-ssa-strlen.c (get_string_len): Rename...
9137 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9138 Avoid assuming length is constant.
9139 (handle_char_store): Use HOST_WIDE_INT for string length.
9141 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9144 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9145 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9147 2018-01-26 Richard Biener <rguenther@suse.de>
9149 PR rtl-optimization/84003
9150 * dse.c (record_store): Only record redundant stores when
9151 the earlier store aliases at least all accesses the later one does.
9153 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9155 PR rtl-optimization/83985
9156 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9157 REG_CFA_RESTORE insns.
9158 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9159 REG_CFA_RESTORE insns here.
9162 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9163 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9165 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9167 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9168 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9169 (arc_init): Likewise.
9170 (arc_override_options): Likewise.
9171 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9173 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9175 * config/arc/arc.h (TARGET_DBNZ): Define.
9176 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9177 properly set the tune attribute.
9178 (dbnz): Use TARGET_DBNZ guard.
9179 * config/arc/arc.opt (mtune): Add core3 option.
9181 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9183 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9184 recognize new pic like addresses.
9185 (arc_delegitimize_address): Clean up.
9187 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9189 * config/arc/arc-arches.def: Option mrf16 valid for all
9191 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9192 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9193 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9194 * config/arc/arc-tables.opt: Regenerate.
9195 * config/arc/arc.c (arc_conditional_register_usage): Handle
9196 reduced register file case.
9197 (arc_file_start): Set must have build attributes.
9198 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9200 * config/arc/arc.opt (mrf16): Add new option.
9201 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9202 * config/arc/genmultilib.awk: Handle new mrf16 option.
9203 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9204 * config/arc/t-multilib: Regenerate.
9205 * doc/invoke.texi (ARC Options): Document mrf16 option.
9207 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9209 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9210 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9211 (arc_attribute_table): Add 'secure_call' attribute.
9212 (arc_print_operand): Print secure call operand.
9213 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9215 (arc_is_secure_call_p): New function. * config/arc/arc.md
9216 (call_i): Add support for sjli instruction.
9217 (call_value_i): Likewise.
9218 * config/arc/constraints.md (Csc): New constraint.
9220 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9221 John Eric Martin <John.Martin@emmicro-us.com>
9223 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9224 * config/arc/arc.c (_arc_jli_section): New struct.
9225 (arc_jli_section): New type.
9226 (rc_jli_sections): New static variable.
9227 (arc_handle_jli_attribute): New function.
9228 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9229 (arc_file_end): New function.
9230 (TARGET_ASM_FILE_END): Define.
9231 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9232 (arc_add_jli_section): New function.
9233 (jli_call_scan): Likewise.
9234 (arc_reorg): Call jli_call_scan.
9235 (arc_output_addsi): Remove 'S' from printing asm operand.
9236 (arc_is_jli_call_p): New function.
9237 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9239 (movhi_insn): Likewise.
9240 (movsi_insn): Likewise.
9241 (movsi_set_cc_insn): Likewise.
9242 (loadqi_update): Likewise.
9243 (load_zeroextendqisi_update): Likewise.
9244 (load_signextendqisi_update): Likewise.
9245 (loadhi_update): Likewise.
9246 (load_zeroextendhisi_update): Likewise.
9247 (load_signextendhisi_update): Likewise.
9248 (loadsi_update): Likewise.
9249 (loadsf_update): Likewise.
9250 (movsicc_insn): Likewise.
9251 (bset_insn): Likewise.
9252 (bxor_insn): Likewise.
9253 (bclr_insn): Likewise.
9254 (bmsk_insn): Likewise.
9255 (bicsi3_insn): Likewise.
9256 (cmpsi_cc_c_insn): Likewise.
9257 (movsi_ne): Likewise.
9258 (movsi_cond_exec): Likewise.
9259 (clrsbsi2): Likewise.
9270 (call_i): Remove 'S' asm letter, add jli instruction.
9271 (call_value_i): Likewise.
9272 * config/arc/arc.op (mjli-always): New option.
9273 * config/arc/constraints.md (Cji): New constraint.
9274 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9276 (subsf3_fpx): Likewise.
9277 (mulsf3_fpx): Likewise.
9278 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9280 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9282 * doc/invoke.texi (ARC): Document mjli-always option.
9284 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9286 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9287 avoid addition with 0 and use incw and decw where possible.
9289 2018-01-26 Richard Biener <rguenther@suse.de>
9291 PR tree-optimization/81082
9292 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9293 association if it requires casting to unsigned.
9294 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9295 from fold_plusminus_mult_expr to catch important cases late when
9296 range info is available.
9298 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9300 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9301 * configure.ac (hidden_linkonce): New test.
9302 * configure: Regenerate.
9303 * config.in: Regenerate.
9305 2018-01-26 Julia Koval <julia.koval@intel.com>
9307 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9308 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9309 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9310 _mm_mask_bitshuffle_epi64_mask): Fix type.
9311 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9312 USI_FTYPE_V4DI_V4DI_USI): Remove.
9313 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9314 __builtin_ia32_vpshufbitqmb256_mask,
9315 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9316 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9317 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9319 2018-01-26 Alan Modra <amodra@gmail.com>
9322 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9323 UNSPEC_VBPERMQ. Sort other unspecs.
9325 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9327 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9329 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9332 * predict.c (drop_profile): Do not push/pop cfun; update also
9334 (handle_missing_profiles): Fix logic looking for zero profiles.
9336 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9339 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9340 on functions with #pragma omp declare simd or functions with simd
9342 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9343 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9344 Remove trailing \n from warning_at calls.
9346 2018-01-25 Tom de Vries <tom@codesourcery.com>
9349 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9350 for neutered workers.
9352 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9355 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9356 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9358 2018-01-24 Jeff Law <law@redhat.com>
9361 * i386.c (get_probe_interval): Move to earlier point.
9362 (ix86_compute_frame_layout): If -fstack-clash-protection and
9363 the frame is larger than the probe interval, then use pushes
9364 to save registers rather than reg->mem moves.
9365 (ix86_expand_prologue): Remove conditional for int_registers_saved
9368 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9371 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9372 min/max for never referenced object.
9374 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9377 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9379 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9380 attributes from DECL_ATTRIBUTES (decl) without affecting
9381 DECL_ATTRIBUTES (current_function_decl).
9382 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9383 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9385 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9387 PR tree-optimization/83979
9388 * fold-const.c (fold_comparison): Use constant_boolean_node
9389 instead of boolean_{true,false}_node.
9391 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9393 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9396 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9398 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9399 Simplify the clause that sets the length attribute.
9400 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9401 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9402 clause that sets the length attribute.
9403 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9405 2018-01-24 Tom de Vries <tom@codesourcery.com>
9408 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9409 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9410 Add strict parameter.
9411 (prevent_branch_around_nothing): Insert dummy insn between branch to
9412 label and label with no ptx insn inbetween.
9413 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9415 2018-01-24 Tom de Vries <tom@codesourcery.com>
9418 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9419 for neutered threads in warp.
9420 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9422 2018-01-24 Richard Biener <rguenther@suse.de>
9424 PR tree-optimization/83176
9425 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9428 2018-01-24 Richard Biener <rguenther@suse.de>
9430 PR tree-optimization/82819
9431 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9432 code generating pluses that are no-ops in the target precision.
9434 2018-01-24 Richard Biener <rguenther@suse.de>
9437 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9439 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9441 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9442 to merge probabilities.
9443 * predict.c (probably_never_executed): Also mark as cold functions
9444 with global 0 profile and guessed local profile.
9445 * profile-count.c (profile_probability::combine_with_count): New
9447 * profile-count.h (profile_probability::operator*,
9448 profile_probability::operator*=, profile_probability::operator/,
9449 profile_probability::operator/=): Reduce precision to adjusted
9450 and set value to guessed on contradictory divisions.
9451 (profile_probability::combine_with_freq): Remove.
9452 (profile_probability::combine_wiht_count): Declare.
9453 (profile_count::force_nonzero):: Set to adjusted.
9454 (profile_count::probability_in):: Set quality to adjusted.
9455 * tree-ssa-tail-merge.c (replace_block_by): Use
9458 2018-01-23 Andrew Waterman <andrew@sifive.com>
9459 Jim Wilson <jimw@sifive.com>
9461 * config/riscv/riscv.c (riscv_stack_boundary): New.
9462 (riscv_option_override): Set riscv_stack_boundary. Handle
9463 riscv_preferred_stack_boundary_arg.
9464 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9465 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9466 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9467 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9468 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9469 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9471 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9474 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9475 of struct ix86_frame.
9476 (ix86_expand_epilogue): Likewise. Add a local variable for
9477 the reg_save_offset field in struct ix86_frame.
9479 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9481 PR tree-optimization/82604
9482 * tree-loop-distribution.c (enum partition_kind): New enum item
9483 PKIND_PARTIAL_MEMSET.
9484 (partition_builtin_p): Support above new enum item.
9485 (generate_code_for_partition): Ditto.
9486 (compute_access_range): Differentiate cases that equality can be
9487 proven at all loops, the innermost loops or no loops.
9488 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9489 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9490 (finalize_partitions, distribute_loop): Don't fuse partition of
9491 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9492 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9495 2018-01-23 Martin Liska <mliska@suse.cz>
9497 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9498 order to ignore the predictor.
9499 (PRED_POLYMORPHIC_CALL): Likewise.
9500 (PRED_RECURSIVE_CALL): Likewise.
9502 2018-01-23 Martin Liska <mliska@suse.cz>
9504 * tree-profile.c (tree_profiling): Print function header to
9505 aware reader which function we are working on.
9506 * value-prof.c (gimple_find_values_to_profile): Do not print
9507 not interesting value histograms.
9509 2018-01-23 Martin Liska <mliska@suse.cz>
9511 * profile-count.h (enum profile_quality): Add
9512 profile_uninitialized as the first value. Do not number values
9513 as they are zero based.
9514 (profile_count::verify): Update sanity check.
9515 (profile_probability::verify): Likewise.
9517 2018-01-23 Nathan Sidwell <nathan@acm.org>
9519 * doc/invoke.texi (ffor-scope): Deprecate.
9521 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9523 PR tree-optimization/83510
9524 * domwalk.c (set_all_edges_as_executable): New function.
9525 (dom_walker::dom_walker): Convert bool param
9526 "skip_unreachable_blocks" to enum reachability. Move setup of
9527 edge flags to set_all_edges_as_executable and only do it when
9528 reachability is REACHABLE_BLOCKS.
9529 * domwalk.h (enum dom_walker::reachability): New enum.
9530 (dom_walker::dom_walker): Convert bool param
9531 "skip_unreachable_blocks" to enum reachability.
9532 (set_all_edges_as_executable): New decl.
9533 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9534 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9536 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9537 but converting true to REACHABLE_BLOCKS.
9538 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9540 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9541 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9542 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9544 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9545 if check_all_array_refs will be called.
9547 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9549 * tree.c (selftest::test_location_wrappers): Add more test
9552 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9554 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9555 (selftest::test_bit_in_range): Likewise.
9557 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9560 * doc/sourcebuild.texi (vect_float): Say that the selector
9561 only describes the situation when -funsafe-math-optimizations is on.
9562 (vect_float_strict): Document.
9564 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9566 PR tree-optimization/83965
9567 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9568 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9569 instead of checking only for a reduction.
9570 (vect_recog_widen_sum_pattern): Likewise.
9572 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9574 * predict.c (probably_never_executed): Only use precise profile info.
9575 (compute_function_frequency): Skip after inlining hack since we now
9576 have quality checking.
9578 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9580 * profile-count.h (profile_probability::very_unlikely,
9581 profile_probability::unlikely, profile_probability::even): Set
9582 precision to guessed.
9584 2018-01-23 Richard Biener <rguenther@suse.de>
9586 PR tree-optimization/83963
9587 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9588 Properly terminate dominator walk when crossing the exit edge not
9589 when visiting its source block.
9591 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9594 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9595 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9597 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9599 PR tree-optimization/83957
9600 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9601 semicolon after for body surrounded by braces.
9603 PR tree-optimization/83081
9604 * profile-count.h (profile_probability::split): New method.
9605 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9606 Use profile_probability::split.
9607 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9608 when splitting a single conditional jump into 2.
9610 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9612 PR tree-optimization/69452
9613 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9616 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9618 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9619 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9620 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9622 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9624 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9625 * config/rl78/rl78.md (movdi): New define_expand.
9626 * config/rl78/rl78.c (rl78_split_movdi): New function.
9628 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9631 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9633 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9634 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9635 128-bit to produce an UNSPEC move to get the double word with the
9636 signbit and then a shift directly to do signbit.
9637 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9638 implementation with a new version that just does either a direct
9639 move or a regular move. Move memory interface to separate insns.
9640 Move insns so they are next to the expander.
9641 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9642 with signbit move. Split big and little endian case.
9643 (signbit<mode>2_dm_mem_le): Likewise.
9644 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9645 (signbit<mode>2_dm2): Likewise.
9647 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9649 * config/rl78/rl78.md (anddi3): New define_expand.
9651 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9653 * config/rl78/rl78.md (umindi3): New define_expand.
9655 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9657 * config/rl78/rl78.md (smindi3): New define_expand.
9659 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9661 * config/rl78/rl78.md (smaxdi3): New define_expand.
9663 2018-01-22 Carl Love <cel@us.ibm.com>
9665 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9666 LVX_V1TI): Add macro expansion.
9667 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9668 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9669 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9670 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9671 Change check to determine if the instruction is a byte reversing
9672 entry. Fix typo in comment.
9673 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9674 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9675 Add def_builtin calls for new builtins.
9676 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9677 Add define_insn expansion.
9679 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9681 * config/rl78/rl78.md (umaxdi3): New define_expand.
9683 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9685 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9686 for non-QImode registers.
9688 2018-01-22 Richard Biener <rguenther@suse.de>
9690 PR tree-optimization/83963
9691 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9692 including the loop exit block.
9693 (scop_detection::merge_sese): Likewise.
9694 (scop_detection::add_scop): Do it here instead.
9696 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9698 * doc/sourcebuild.texi (arm_softfloat): Document.
9700 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9703 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9704 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9705 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9707 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9708 David Edelsohn <dje.gcc@gmail.com>
9711 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9712 Change "crset eq" to "crset 2".
9713 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9714 (*call_indirect_aix<mode>_nospec): Likewise.
9715 (*call_value_indirect_aix<mode>_nospec): Likewise.
9716 (*call_indirect_elfv2<mode>_nospec): Likewise.
9717 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9718 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9719 change assembly output from . to $.
9720 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9721 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9722 (*tablejump<mode>_internal1_nospec): Likewise.
9724 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9727 * config/sh/sh_optimize_sett_clrt.cc:
9728 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9730 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9732 PR tree-optimization/83940
9733 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9734 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9735 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9736 use it to pass back the definition type.
9737 (vect_check_store_rhs): Likewise rhs_dt_out.
9738 (vect_build_gather_load_calls): Add a mask_dt argument and use
9739 it instead of a call to vect_is_simple_use.
9740 (vectorizable_store): Update calls to vect_check_load_store_mask
9741 and vect_check_store_rhs. Use the dt returned by the latter instead
9742 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9743 instead of calls to vect_is_simple_use. Pass the scalar rather
9744 than the vector operand to vect_is_simple_use when handling
9745 second and subsequent copies of an rhs value.
9746 (vectorizable_load): Update calls to vect_check_load_store_mask
9747 and vect_build_gather_load_calls. Use the cached mask_dt and
9748 gs_info.offset_dt instead of calls to vect_is_simple_use.
9750 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9753 * tree-emutls.c: Include gimplify.h.
9754 (lower_emutls_2): New function.
9755 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9756 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9757 it before further processing.
9760 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9761 UINTVAL (trueop1) instead of INTVAL (op1).
9763 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9767 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9768 INCOMING_FRAME_SP_OFFSET if not defined.
9769 (scan_trace): Add ENTRY argument. If true and
9770 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9771 emit a note to adjust the CFA offset.
9772 (create_cfi_notes): Adjust scan_trace callers.
9773 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9774 INCOMING_FRAME_SP_OFFSET in the CIE.
9775 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9776 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9778 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9779 * doc/tm.texi: Regenerated.
9781 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9783 PR rtl-optimization/83147
9784 * lra-constraints.c (remove_inheritance_pseudos): Use
9785 lra_substitute_pseudo_within_insn.
9787 2018-01-19 Tom de Vries <tom@codesourcery.com>
9788 Cesar Philippidis <cesar@codesourcery.com>
9791 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9793 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9796 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9797 spaces for function labels.
9799 2018-01-19 Martin Liska <mliska@suse.cz>
9801 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9802 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9803 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9804 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9805 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9806 (PRED_CONST_RETURN): Change from 69 to 65.
9807 (PRED_NULL_RETURN): Change from 91 to 71.
9808 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9809 (PRED_LOOP_GUARD): Change from 66 to 73.
9811 2018-01-19 Martin Liska <mliska@suse.cz>
9813 * predict.c (predict_insn_def): Add new assert.
9814 (struct branch_predictor): Change type to signed integer.
9815 (test_prediction_value_range): Amend test to cover
9817 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9818 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9819 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9820 (PRED_LOOP_IV_COMPARE): Likewise.
9821 * predict.h (PROB_UNINITIALIZED): Define new constant.
9823 2018-01-19 Martin Liska <mliska@suse.cz>
9825 * predict.c (dump_prediction): Add new format for
9826 analyze_brprob.py script which is enabled with -details
9828 * profile-count.h (precise_p): New function.
9830 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9832 PR tree-optimization/83922
9833 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9834 there are no statements that need masking.
9835 (vect_active_double_reduction_p): New function.
9836 (vect_analyze_loop_operations): Use it when handling phis that
9837 are not in the loop header.
9839 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9841 PR tree-optimization/83914
9842 * tree-vect-loop.c (vectorizable_induction): Don't convert
9843 init_expr or apply the peeling adjustment for inductions
9844 that are nested within the vectorized loop.
9846 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9848 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9851 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9855 * function.h (gimplify_parameters): Add gimple_seq * argument.
9856 * function.c: Include gimple.h and options.h.
9857 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9858 for the added local temporaries if needed.
9859 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9860 if there are any parameter cleanups, wrap whole body into a
9861 try/finally with the cleanups.
9863 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9866 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9867 Use GET_MODE_CLASS for scalar floating point.
9869 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9873 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9874 Fix call of call_cgraph_insertion_hooks.
9876 2018-01-18 Martin Sebor <msebor@redhat.com>
9878 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9880 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9883 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9886 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9889 * common.opt: (-ffile-prefix-map): New option.
9890 * opts.c (common_handle_option): Defer it.
9891 * opts-global.c (handle_common_deferred_options): Handle it.
9892 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9893 * file-prefix-map.h: New file.
9894 (remap_debug_filename, add_debug_prefix_map): ...here.
9895 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9896 * final.c (debug_prefix_map, add_debug_prefix_map
9897 remap_debug_filename): Move to...
9898 * file-prefix-map.c: New file.
9899 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9900 generalize, get rid of alloca(), use strrchr() instead of strchr().
9901 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9902 Implement in terms of add_prefix_map().
9903 (remap_macro_filename, remap_debug_filename): Implement in term of
9905 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9906 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9907 * dbxout.c: Include file-prefix-map.h.
9908 * varasm.c: Likewise.
9909 * vmsdbgout.c: Likewise.
9910 * xcoffout.c: Likewise.
9911 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9912 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9913 * doc/invoke.texi (-ffile-prefix-map): Document.
9914 (-fdebug-prefix-map): Update description.
9916 2018-01-18 Martin Liska <mliska@suse.cz>
9918 * config/i386/i386.c (indirect_thunk_name): Document that also
9920 (output_indirect_thunk): Document why both instructions
9921 (pause and lfence) are generated.
9923 2018-01-18 Richard Biener <rguenther@suse.de>
9925 PR tree-optimization/83887
9926 * graphite-scop-detection.c
9927 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9928 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9929 (scop_detection::merge_sese): Re-implement with a flood-fill
9930 algorithm that properly finds a SESE region if it exists.
9932 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9935 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9936 pointer_diff optimizations use view_convert instead of convert.
9938 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9940 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9941 Generate different code for -mno-speculate-indirect-jumps.
9942 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9943 (*call_indirect_aix<mode>): Disable for
9944 -mno-speculate-indirect-jumps.
9945 (*call_indirect_aix<mode>_nospec): New define_insn.
9946 (*call_value_indirect_aix<mode>): Disable for
9947 -mno-speculate-indirect-jumps.
9948 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9949 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9950 -mno-speculate-indirect-jumps.
9951 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9953 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9955 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9956 long double type, set the flags for noting the default long double
9957 type, even if we don't pass or return a long double type.
9959 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9962 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9965 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9967 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9968 support for merge[hl].
9969 (fold_mergehl_helper): New helper function.
9970 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9971 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9972 (altivec_vmrglw_direct): Add xxmrglw insn.
9974 2018-01-17 Andrew Waterman <andrew@sifive.com>
9976 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9977 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9979 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9982 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9983 call the lto_location_cache before reading the
9984 DECL_SOURCE_LOCATION of the types.
9986 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9987 Richard Sandiford <richard.sandiford@linaro.org>
9989 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9990 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9991 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9992 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9993 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9995 * config/aarch64/constraints.md (aarch64_movti_operand):
9997 * config/aarch64/predicates.md (Uti): Add new constraint.
9999 2018-01-17 Carl Love <cel@us.ibm.com>
10001 * config/rs6000/vsx.md (define_expand xl_len_r,
10002 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10003 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10005 (define_expand, define_insn): Move the shift left from the
10006 define_insn to the define_expand for lxvl and stxvl instructions.
10007 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10008 and XL_LEN_R definitions to PURE.
10010 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10012 * config/i386/i386.c (indirect_thunk_name): Declare regno
10013 as unsigned int. Compare regno with INVALID_REGNUM.
10014 (output_indirect_thunk): Ditto.
10015 (output_indirect_thunk_function): Ditto.
10016 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10017 in the call to output_indirect_thunk_function.
10019 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10021 PR middle-end/83884
10022 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10023 rather than the size of inner_type to determine the stack slot size
10024 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10026 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10029 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10032 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10034 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10035 endian Linux systems to optionally enable multilibs for selecting
10036 the long double type if the user configured an explicit type.
10037 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10038 have no long double multilibs if not defined.
10039 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10040 warn if the user used -mabi={ieee,ibm}longdouble and we built
10041 multilibs for long double.
10042 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10043 appropriate multilib option.
10044 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10046 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10047 for building long double multilibs.
10048 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10050 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10052 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10055 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10057 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10060 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10063 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10066 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10068 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10069 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10071 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10073 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10074 different rtl trees depending on TARGET_64BIT.
10075 (rs6000_gen_lvx): Likewise.
10077 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10079 * config/visium/visium.md (nop): Tweak comment.
10080 (hazard_nop): Likewise.
10082 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10084 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10085 -mspeculate-indirect-jumps.
10086 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10087 for -mno-speculate-indirect-jumps.
10088 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10089 (*call_value_indirect_elfv2<mode>): Disable for
10090 -mno-speculate-indirect-jumps.
10091 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10092 (indirect_jump): Emit different RTL for
10093 -mno-speculate-indirect-jumps.
10094 (*indirect_jump<mode>): Disable for
10095 -mno-speculate-indirect-jumps.
10096 (*indirect_jump<mode>_nospec): New define_insn.
10097 (tablejump): Emit different RTL for
10098 -mno-speculate-indirect-jumps.
10099 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10100 (tablejumpsi_nospec): New define_expand.
10101 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10102 (tablejumpdi_nospec): New define_expand.
10103 (*tablejump<mode>_internal1): Disable for
10104 -mno-speculate-indirect-jumps.
10105 (*tablejump<mode>_internal1_nospec): New define_insn.
10106 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10109 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10111 * caller-save.c (insert_save): Drop unnecessary parameter. All
10114 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10115 Richard Biener <rguenth@suse.de>
10118 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10119 return early, inline manually is_gimple_sizepos. Make sure if we
10120 call gimplify_expr we don't end up with a gimple constant.
10121 * tree.c (variably_modified_type_p): Don't return true for
10122 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10123 * gimplify.h (is_gimple_sizepos): Remove.
10125 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10127 PR tree-optimization/83857
10128 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10129 vectorizable_live_operation for pure SLP statements.
10130 (vectorizable_live_operation): Handle PHIs.
10132 2018-01-16 Richard Biener <rguenther@suse.de>
10134 PR tree-optimization/83867
10135 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10136 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10138 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10141 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10142 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10143 If off is not INTEGER_CST, issue a may not be aligned warning
10144 rather than isn't aligned. Use isn%'t rather than isn't.
10145 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10147 <case MULT_EXPR>: Improve the case when bottom and one of the
10148 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10149 operand, in that case check if the other operand is multiple of
10150 bottom divided by the INTEGER_CST operand.
10152 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10155 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10156 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10157 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10158 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10159 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10160 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10161 (pa_function_arg_size): New function.
10163 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10165 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10166 in a separate statement.
10168 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10170 PR tree-optimization/83847
10171 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10172 group gathers and scatters.
10174 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10176 PR rtl-optimization/86620
10177 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10179 PR rtl-optimization/83213
10180 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10181 to last if both are JUMP_INSNs.
10183 PR tree-optimization/83843
10184 * gimple-ssa-store-merging.c
10185 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10186 store_immediate_info for bswap/nop orig_stores.
10188 2018-01-15 Andrew Waterman <andrew@sifive.com>
10190 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10192 <UDIV>: Increase cost if !TARGET_DIV.
10194 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10196 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10197 (define_attr "cr_logical_3op"): New.
10198 (cceq_ior_compare): Adjust.
10199 (cceq_ior_compare_complement): Adjust.
10200 (*cceq_rev_compare): Adjust.
10201 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10202 (is_cracked_insn): Adjust.
10203 (insn_must_be_first_in_group): Adjust.
10204 * config/rs6000/40x.md: Adjust.
10205 * config/rs6000/440.md: Adjust.
10206 * config/rs6000/476.md: Adjust.
10207 * config/rs6000/601.md: Adjust.
10208 * config/rs6000/603.md: Adjust.
10209 * config/rs6000/6xx.md: Adjust.
10210 * config/rs6000/7450.md: Adjust.
10211 * config/rs6000/7xx.md: Adjust.
10212 * config/rs6000/8540.md: Adjust.
10213 * config/rs6000/cell.md: Adjust.
10214 * config/rs6000/e300c2c3.md: Adjust.
10215 * config/rs6000/e500mc.md: Adjust.
10216 * config/rs6000/e500mc64.md: Adjust.
10217 * config/rs6000/e5500.md: Adjust.
10218 * config/rs6000/e6500.md: Adjust.
10219 * config/rs6000/mpc.md: Adjust.
10220 * config/rs6000/power4.md: Adjust.
10221 * config/rs6000/power5.md: Adjust.
10222 * config/rs6000/power6.md: Adjust.
10223 * config/rs6000/power7.md: Adjust.
10224 * config/rs6000/power8.md: Adjust.
10225 * config/rs6000/power9.md: Adjust.
10226 * config/rs6000/rs64.md: Adjust.
10227 * config/rs6000/titan.md: Adjust.
10229 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10231 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10232 ix86_indirect_branch_register logic.
10234 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10236 * config/i386/constraints.md (Bs): Update
10237 ix86_indirect_branch_register check. Don't check
10238 ix86_indirect_branch_register with GOT_memory_operand.
10240 * config/i386/predicates.md (GOT_memory_operand): Don't check
10241 ix86_indirect_branch_register here.
10242 (GOT32_symbol_operand): Likewise.
10244 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10246 * config/i386/predicates.md (constant_call_address_operand):
10247 Rewrite ix86_indirect_branch_register logic.
10248 (sibcall_insn_operand): Likewise.
10250 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10252 * config/i386/constraints.md (Bs): Replace
10253 ix86_indirect_branch_thunk_register with
10254 ix86_indirect_branch_register.
10256 * config/i386/i386.md (indirect_jump): Likewise.
10257 (tablejump): Likewise.
10258 (*sibcall_memory): Likewise.
10259 (*sibcall_value_memory): Likewise.
10260 Peepholes of indirect call and jump via memory: Likewise.
10261 * config/i386/i386.opt: Likewise.
10262 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10263 (GOT_memory_operand): Likewise.
10264 (call_insn_operand): Likewise.
10265 (sibcall_insn_operand): Likewise.
10266 (GOT32_symbol_operand): Likewise.
10268 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10270 PR middle-end/83837
10271 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10272 type rather than type addr's type points to.
10273 (expand_omp_atomic_mutex): Likewise.
10274 (expand_omp_atomic): Likewise.
10276 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10279 * config/i386/i386.c (output_indirect_thunk_function): Use
10280 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10281 for __x86_return_thunk.
10283 2018-01-15 Richard Biener <rguenther@suse.de>
10285 PR middle-end/83850
10286 * expmed.c (extract_bit_field_1): Fix typo.
10288 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10291 * config/arm/iterators.md (VF): New mode iterator.
10292 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10293 Remove integer-related logic from pattern.
10294 (neon_vabd<mode>_3): Likewise.
10296 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10298 PR middle-end/82694
10299 * common.opt (fstrict-overflow): No longer an alias.
10300 (fwrapv-pointer): New option.
10301 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10302 also for pointer types based on flag_wrapv_pointer.
10303 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10304 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10305 opts->x_flag_wrapv got set.
10306 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10307 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10308 POINTER_TYPE_OVERFLOW_UNDEFINED.
10309 * match.pd: Likewise in address comparison pattern.
10310 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10312 2018-01-15 Richard Biener <rguenther@suse.de>
10315 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10316 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10317 Reset type names to their identifier if their TYPE_DECL doesn't
10318 have linkage (and thus is used for ODR and devirt).
10319 (save_debug_info_for_decl): Remove.
10320 (save_debug_info_for_type): Likewise.
10321 (add_tree_to_fld_list): Adjust.
10322 * tree-pretty-print.c (dump_generic_node): Make dumping of
10323 type names more robust.
10325 2018-01-15 Richard Biener <rguenther@suse.de>
10327 * BASE-VER: Bump to 8.0.1.
10329 2018-01-14 Martin Sebor <msebor@redhat.com>
10332 * builtins.c (check_access): Avoid warning when the no-warning bit
10335 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10337 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10338 * ira-color (allocno_hard_regs_compare): Likewise.
10340 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10343 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10344 Use .pushsection/.popsection.
10346 2018-01-14 Martin Sebor <msebor@redhat.com>
10349 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10351 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10353 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10354 entry from extra_headers.
10355 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10356 extra_headers, make the list bitwise identical to the i?86-*-* one.
10358 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10360 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10361 -mcmodel=large with -mindirect-branch=thunk,
10362 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10363 -mfunction-return=thunk-extern.
10364 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10365 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10366 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10368 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10370 * config/i386/i386.c (print_reg): Print the name of the full
10371 integer register without '%'.
10372 (ix86_print_operand): Handle 'V'.
10373 * doc/extend.texi: Document 'V' modifier.
10375 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10377 * config/i386/constraints.md (Bs): Disallow memory operand for
10378 -mindirect-branch-register.
10380 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10381 (GOT_memory_operand): Likewise.
10382 (call_insn_operand): Likewise.
10383 (sibcall_insn_operand): Likewise.
10384 (GOT32_symbol_operand): Likewise.
10385 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10386 for -mindirect-branch-register.
10387 (tablejump): Likewise.
10388 (*sibcall_memory): Likewise.
10389 (*sibcall_value_memory): Likewise.
10390 Disallow peepholes of indirect call and jump via memory for
10391 -mindirect-branch-register.
10392 (*call_pop): Replace m with Bw.
10393 (*call_value_pop): Likewise.
10394 (*sibcall_pop_memory): Replace m with Bs.
10395 * config/i386/i386.opt (mindirect-branch-register): New option.
10396 * doc/invoke.texi: Document -mindirect-branch-register option.
10398 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10400 * config/i386/i386-protos.h (ix86_output_function_return): New.
10401 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10402 set function_return_type.
10403 (indirect_thunk_name): Add ret_p to indicate thunk for function
10405 (output_indirect_thunk_function): Pass false to
10406 indirect_thunk_name.
10407 (ix86_output_indirect_branch_via_reg): Likewise.
10408 (ix86_output_indirect_branch_via_push): Likewise.
10409 (output_indirect_thunk_function): Create alias for function
10410 return thunk if regno < 0.
10411 (ix86_output_function_return): New function.
10412 (ix86_handle_fndecl_attribute): Handle function_return.
10413 (ix86_attribute_table): Add function_return.
10414 * config/i386/i386.h (machine_function): Add
10415 function_return_type.
10416 * config/i386/i386.md (simple_return_internal): Use
10417 ix86_output_function_return.
10418 (simple_return_internal_long): Likewise.
10419 * config/i386/i386.opt (mfunction-return=): New option.
10420 (indirect_branch): Mention -mfunction-return=.
10421 * doc/extend.texi: Document function_return function attribute.
10422 * doc/invoke.texi: Document -mfunction-return= option.
10424 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10426 * config/i386/i386-opts.h (indirect_branch): New.
10427 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10428 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10429 with local indirect jump when converting indirect call and jump.
10430 (ix86_set_indirect_branch_type): New.
10431 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10432 (indirectlabelno): New.
10433 (indirect_thunk_needed): Likewise.
10434 (indirect_thunk_bnd_needed): Likewise.
10435 (indirect_thunks_used): Likewise.
10436 (indirect_thunks_bnd_used): Likewise.
10437 (INDIRECT_LABEL): Likewise.
10438 (indirect_thunk_name): Likewise.
10439 (output_indirect_thunk): Likewise.
10440 (output_indirect_thunk_function): Likewise.
10441 (ix86_output_indirect_branch_via_reg): Likewise.
10442 (ix86_output_indirect_branch_via_push): Likewise.
10443 (ix86_output_indirect_branch): Likewise.
10444 (ix86_output_indirect_jmp): Likewise.
10445 (ix86_code_end): Call output_indirect_thunk_function if needed.
10446 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10448 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10449 (ix86_attribute_table): Add indirect_branch.
10450 * config/i386/i386.h (machine_function): Add indirect_branch_type
10451 and has_local_indirect_jump.
10452 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10454 (tablejump): Likewise.
10455 (*indirect_jump): Use ix86_output_indirect_jmp.
10456 (*tablejump_1): Likewise.
10457 (simple_return_indirect_internal): Likewise.
10458 * config/i386/i386.opt (mindirect-branch=): New option.
10459 (indirect_branch): New.
10462 (thunk-inline): Likewise.
10463 (thunk-extern): Likewise.
10464 * doc/extend.texi: Document indirect_branch function attribute.
10465 * doc/invoke.texi: Document -mindirect-branch= option.
10467 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10470 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10472 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10474 * ipa-inline.c (want_inline_small_function_p): Return false if
10475 inlining has already failed with CIF_FINAL_ERROR.
10476 (update_caller_keys): Call want_inline_small_function_p before
10478 (update_callee_keys): Likewise.
10480 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10482 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10484 (rs6000_quadword_masked_address_p): Likewise.
10485 (quad_aligned_load_p): Likewise.
10486 (quad_aligned_store_p): Likewise.
10487 (const_load_sequence_p): Add comment to describe the outer-most loop.
10488 (mimic_memory_attributes_and_flags): New function.
10489 (rs6000_gen_stvx): Likewise.
10490 (replace_swapped_aligned_store): Likewise.
10491 (rs6000_gen_lvx): Likewise.
10492 (replace_swapped_aligned_load): Likewise.
10493 (replace_swapped_load_constant): Capitalize argument name in
10494 comment describing this function.
10495 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10496 and stores that access quad-word aligned addresses and replace
10497 with stvx or lvx instructions when appropriate.
10498 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10499 New function prototype.
10500 (rs6000_quadword_masked_address_p): Likewise.
10501 (rs6000_gen_lvx): Likewise.
10502 (rs6000_gen_stvx): Likewise.
10503 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10504 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10505 when memory address is aligned.
10506 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10507 this split to select lvx instruction when memory address is aligned.
10508 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10509 instruction when memory address is aligned.
10510 (*vsx_le_perm_load_v16qi): Likewise.
10511 (four unnamed splitters): Modify to select the stvx instruction
10512 when memory is aligned.
10514 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10516 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10517 which appears in the queue multiple times.
10519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10520 Alan Hayward <alan.hayward@arm.com>
10521 David Sherwood <david.sherwood@arm.com>
10523 * tree-vectorizer.h (vec_lower_bound): New structure.
10524 (_loop_vec_info): Add check_nonzero and lower_bounds.
10525 (LOOP_VINFO_CHECK_NONZERO): New macro.
10526 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10527 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10528 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10529 fields. Make seg_len the distance travelled, not including the
10531 (dr_direction_indicator): Declare.
10532 (dr_zero_step_indicator): Likewise.
10533 (dr_known_forward_stride_p): Likewise.
10534 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10536 (runtime_alias_check_p): Allow runtime alias checks with
10538 (operator ==): Compare access_size and align.
10539 (prune_runtime_alias_test_list): Rework for new distinction between
10540 the access_size and seg_len.
10541 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10543 (get_segment_min_max): New function.
10544 (create_intersect_range_checks): Use it.
10545 (dr_step_indicator): New function.
10546 (dr_direction_indicator): Likewise.
10547 (dr_zero_step_indicator): Likewise.
10548 (dr_known_forward_stride_p): Likewise.
10549 * tree-loop-distribution.c (data_ref_segment_size): Return
10550 DR_STEP * (niters - 1).
10551 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10553 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10554 (vect_preserves_scalar_order_p): New function, split out from...
10555 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10556 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10557 (vect_vfa_access_size): New function.
10558 (vect_vfa_align): Likewise.
10559 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10560 (dump_lower_bound): New function.
10561 (vect_check_lower_bound): Likewise.
10562 (vect_small_gap_p): Likewise.
10563 (vectorizable_with_step_bound_p): Likewise.
10564 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10565 depencies if the vectorization factor is 1. Convert the checks
10566 for nonzero steps into checks on the bounds of DR_STEP. Try using
10567 a bunds check for variable steps if the minimum required step is
10568 relatively small. Update calls to the dr_with_seg_len
10569 constructor and to vect_compile_time_alias.
10570 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10572 (vect_loop_versioning): Call it.
10573 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10575 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10578 Alan Hayward <alan.hayward@arm.com>
10579 David Sherwood <david.sherwood@arm.com>
10581 * doc/sourcebuild.texi (vect_scatter_store): Document.
10582 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10584 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10586 * genopinit.c (main): Add supports_vec_scatter_store and
10587 supports_vec_scatter_store_cached to target_optabs.
10588 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10589 IFN_MASK_SCATTER_STORE.
10590 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10592 * internal-fn.h (internal_store_fn_p): Declare.
10593 (internal_fn_stored_value_index): Likewise.
10594 * internal-fn.c (scatter_store_direct): New macro.
10595 (expand_scatter_store_optab_fn): New function.
10596 (direct_scatter_store_optab_supported_p): New macro.
10597 (internal_store_fn_p): New function.
10598 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10599 IFN_MASK_SCATTER_STORE.
10600 (internal_fn_mask_index): Likewise.
10601 (internal_fn_stored_value_index): New function.
10602 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10603 for scatter stores.
10604 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10605 * optabs-query.c (supports_vec_scatter_store_p): New function.
10606 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10607 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10608 true for scatter stores.
10609 (vect_gather_scatter_fn_p): Handle scatter stores too.
10610 (vect_check_gather_scatter): Consider using scatter stores if
10611 supports_vec_scatter_store_p.
10612 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10613 scatter stores too.
10614 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10615 internal_fn_stored_value_index.
10616 (check_load_store_masking): Handle scatter stores too.
10617 (vect_get_store_rhs): Make public.
10618 (vectorizable_call): Use internal_store_fn_p.
10619 (vectorizable_store): Handle scatter store internal functions.
10620 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10621 when deciding whether the end of the group has been reached.
10622 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10623 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10624 (mask_scatter_store<mode>): New insns.
10626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10627 Alan Hayward <alan.hayward@arm.com>
10628 David Sherwood <david.sherwood@arm.com>
10630 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10631 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10632 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10634 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10635 Use vect_truncate_gather_scatter_offset if we can't treat the
10636 operation as a normal gather load or scatter store.
10637 (get_group_load_store_type): Take the gather_scatter_info
10638 as argument. Try using a gather load or scatter store for
10639 single-element groups.
10640 (get_load_store_type): Update calls to get_group_load_store_type
10641 and vect_use_strided_gather_scatters_p.
10643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10644 Alan Hayward <alan.hayward@arm.com>
10645 David Sherwood <david.sherwood@arm.com>
10647 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10648 optional tree argument.
10649 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10651 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10652 but continue to use the current value as a fallback.
10653 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10654 to compare the updates.
10655 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10656 (get_load_store_type): Use it when handling a strided access.
10657 (vect_get_strided_load_store_ops): New function.
10658 (vect_get_data_ptr_increment): Likewise.
10659 (vectorizable_load): Handle strided gather loads. Always pass
10660 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10663 Alan Hayward <alan.hayward@arm.com>
10664 David Sherwood <david.sherwood@arm.com>
10666 * doc/md.texi (gather_load@var{m}): Document.
10667 (mask_gather_load@var{m}): Likewise.
10668 * genopinit.c (main): Add supports_vec_gather_load and
10669 supports_vec_gather_load_cached to target_optabs.
10670 * optabs-tree.c (init_tree_optimization_optabs): Use
10671 ggc_cleared_alloc to allocate target_optabs.
10672 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10673 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10675 * internal-fn.h (internal_load_fn_p): Declare.
10676 (internal_gather_scatter_fn_p): Likewise.
10677 (internal_fn_mask_index): Likewise.
10678 (internal_gather_scatter_fn_supported_p): Likewise.
10679 * internal-fn.c (gather_load_direct): New macro.
10680 (expand_gather_load_optab_fn): New function.
10681 (direct_gather_load_optab_supported_p): New macro.
10682 (direct_internal_fn_optab): New function.
10683 (internal_load_fn_p): Likewise.
10684 (internal_gather_scatter_fn_p): Likewise.
10685 (internal_fn_mask_index): Likewise.
10686 (internal_gather_scatter_fn_supported_p): Likewise.
10687 * optabs-query.c (supports_at_least_one_mode_p): New function.
10688 (supports_vec_gather_load_p): Likewise.
10689 * optabs-query.h (supports_vec_gather_load_p): Declare.
10690 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10691 and memory_type field.
10692 (NUM_PATTERNS): Bump to 15.
10693 * tree-vect-data-refs.c: Include internal-fn.h.
10694 (vect_gather_scatter_fn_p): New function.
10695 (vect_describe_gather_scatter_call): Likewise.
10696 (vect_check_gather_scatter): Try using internal functions for
10697 gather loads. Recognize existing calls to a gather load function.
10698 (vect_analyze_data_refs): Consider using gather loads if
10699 supports_vec_gather_load_p.
10700 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10701 (vect_get_gather_scatter_offset_type): Likewise.
10702 (vect_convert_mask_for_vectype): Likewise.
10703 (vect_add_conversion_to_patterm): Likewise.
10704 (vect_try_gather_scatter_pattern): Likewise.
10705 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10706 (vect_vect_recog_func_ptrs): Add it.
10707 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10708 internal_fn_mask_index and internal_gather_scatter_fn_p.
10709 (check_load_store_masking): Take the gather_scatter_info as an
10710 argument and handle gather loads.
10711 (vect_get_gather_scatter_ops): New function.
10712 (vectorizable_call): Check internal_load_fn_p.
10713 (vectorizable_load): Likewise. Handle gather load internal
10715 (vectorizable_store): Update call to check_load_store_masking.
10716 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10717 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10718 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10719 (aarch64_gather_scale_operand_d): New predicates.
10720 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10721 (mask_gather_load<mode>): New insns.
10723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10724 Alan Hayward <alan.hayward@arm.com>
10725 David Sherwood <david.sherwood@arm.com>
10727 * optabs.def (fold_left_plus_optab): New optab.
10728 * doc/md.texi (fold_left_plus_@var{m}): Document.
10729 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10730 * internal-fn.c (fold_left_direct): Define.
10731 (expand_fold_left_optab_fn): Likewise.
10732 (direct_fold_left_optab_supported_p): Likewise.
10733 * fold-const-call.c (fold_const_fold_left): New function.
10734 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10735 * tree-parloops.c (valid_reduction_p): New function.
10736 (gather_scalar_reductions): Use it.
10737 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10738 (vect_finish_replace_stmt): Declare.
10739 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10740 (needs_fold_left_reduction_p): New function, split out from...
10741 (vect_is_simple_reduction): ...here. Accept reductions that
10742 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10743 (vect_force_simple_reduction): Also store the reduction type in
10744 the assignment's STMT_VINFO_REDUC_TYPE.
10745 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10746 (merge_with_identity): New function.
10747 (vect_expand_fold_left): Likewise.
10748 (vectorize_fold_left_reduction): Likewise.
10749 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10750 scalar phi in place for it. Check for target support and reject
10751 cases that would reassociate the operation. Defer the transform
10752 phase to vectorize_fold_left_reduction.
10753 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10754 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10755 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10759 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10760 call to ifc_temp_var.
10762 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10763 Alan Hayward <alan.hayward@arm.com>
10764 David Sherwood <david.sherwood@arm.com>
10766 * target.def (legitimize_address_displacement): Take the original
10767 offset as a poly_int.
10768 * targhooks.h (default_legitimize_address_displacement): Update
10770 * targhooks.c (default_legitimize_address_displacement): Likewise.
10771 * doc/tm.texi: Regenerate.
10772 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10773 as an argument, moving assert of ad->disp == ad->disp_term to...
10774 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10775 Try calling targetm.legitimize_address_displacement before expanding
10776 the address rather than afterwards, and adjust for the new interface.
10777 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10778 Match the new hook interface. Handle SVE addresses.
10779 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10780 new hook interface.
10782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10784 * Makefile.in (OBJS): Add early-remat.o.
10785 * target.def (select_early_remat_modes): New hook.
10786 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10787 * doc/tm.texi: Regenerate.
10788 * targhooks.h (default_select_early_remat_modes): Declare.
10789 * targhooks.c (default_select_early_remat_modes): New function.
10790 * timevar.def (TV_EARLY_REMAT): New timevar.
10791 * passes.def (pass_early_remat): New pass.
10792 * tree-pass.h (make_pass_early_remat): Declare.
10793 * early-remat.c: New file.
10794 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10796 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10799 Alan Hayward <alan.hayward@arm.com>
10800 David Sherwood <david.sherwood@arm.com>
10802 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10803 vfm1 with a bound_epilog parameter.
10804 (vect_do_peeling): Update calls accordingly, and move the prologue
10805 call earlier in the function. Treat the base bound_epilog as 0 for
10806 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10807 this base when peeling for gaps.
10808 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10809 with fully-masked loops.
10810 (vect_estimate_min_profitable_iters): Handle the single peeled
10811 iteration in that case.
10813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10814 Alan Hayward <alan.hayward@arm.com>
10815 David Sherwood <david.sherwood@arm.com>
10817 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10818 single-element interleaving even if the size is not a power of 2.
10819 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10820 accesses for single-element interleaving if the group size is
10823 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10824 Alan Hayward <alan.hayward@arm.com>
10825 David Sherwood <david.sherwood@arm.com>
10827 * doc/md.texi (fold_extract_last_@var{m}): Document.
10828 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10829 * optabs.def (fold_extract_last_optab): New optab.
10830 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10831 * internal-fn.c (fold_extract_direct): New macro.
10832 (expand_fold_extract_optab_fn): Likewise.
10833 (direct_fold_extract_optab_supported_p): Likewise.
10834 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10835 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10836 EXTRACT_LAST_REDUCTION.
10837 (get_initial_def_for_reduction): Do not create an initial vector
10838 for EXTRACT_LAST_REDUCTION reductions.
10839 (vectorizable_reduction): Leave the scalar phi in place for
10840 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10841 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10842 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10843 transform phase to vectorizable_condition.
10844 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10846 (vect_finish_stmt_generation): ...here.
10847 (vect_finish_replace_stmt): New function.
10848 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10849 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10851 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10854 Alan Hayward <alan.hayward@arm.com>
10855 David Sherwood <david.sherwood@arm.com>
10857 * doc/md.texi (extract_last_@var{m}): Document.
10858 * optabs.def (extract_last_optab): New optab.
10859 * internal-fn.def (EXTRACT_LAST): New internal function.
10860 * internal-fn.c (cond_unary_direct): New macro.
10861 (expand_cond_unary_optab_fn): Likewise.
10862 (direct_cond_unary_optab_supported_p): Likewise.
10863 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10864 loops using EXTRACT_LAST.
10865 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10866 (extract_last_<mode>): ...this optab.
10867 (vec_extract<mode><Vel>): Update accordingly.
10869 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10870 Alan Hayward <alan.hayward@arm.com>
10871 David Sherwood <david.sherwood@arm.com>
10873 * target.def (empty_mask_is_expensive): New hook.
10874 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10875 * doc/tm.texi: Regenerate.
10876 * targhooks.h (default_empty_mask_is_expensive): Declare.
10877 * targhooks.c (default_empty_mask_is_expensive): New function.
10878 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10879 if the target says that empty masks are expensive.
10880 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10882 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10885 Alan Hayward <alan.hayward@arm.com>
10886 David Sherwood <david.sherwood@arm.com>
10888 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10889 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10890 (vect_use_loop_mask_for_alignment_p): New function.
10891 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10892 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10893 niters_skip argument. Make sure that the first niters_skip elements
10894 of the first iteration are inactive.
10895 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10896 Update call to vect_set_loop_masks_directly.
10897 (get_misalign_in_elems): New function, split out from...
10898 (vect_gen_prolog_loop_niters): ...here.
10899 (vect_update_init_of_dr): Take a code argument that specifies whether
10900 the adjustment should be added or subtracted.
10901 (vect_update_init_of_drs): Likewise.
10902 (vect_prepare_for_masked_peels): New function.
10903 (vect_do_peeling): Skip prologue peeling if we're using a mask
10904 instead. Update call to vect_update_inits_of_drs.
10905 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10907 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10908 alignment. Do not include the number of peeled iterations in
10909 the minimum threshold in that case.
10910 (vectorizable_induction): Adjust the start value down by
10911 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10912 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10913 Take the number of skipped iterations into account when calculating
10915 * tree-vect-stmts.c (vect_gen_while_not): New function.
10917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10918 Alan Hayward <alan.hayward@arm.com>
10919 David Sherwood <david.sherwood@arm.com>
10921 * doc/sourcebuild.texi (vect_fully_masked): Document.
10922 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10923 default value to 0.
10924 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10926 (vect_analyze_loop_2): ...here. Don't check the vectorization
10927 factor against the number of loop iterations if the loop is
10930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10931 Alan Hayward <alan.hayward@arm.com>
10932 David Sherwood <david.sherwood@arm.com>
10934 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10935 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10936 (dump_groups): Update accordingly.
10937 (iv_use::mem_type): New member variable.
10938 (address_p): New function.
10939 (record_use): Add a mem_type argument and initialize the new
10941 (record_group_use): Add a mem_type argument. Use address_p.
10942 Remove obsolete null checks of base_object. Update call to record_use.
10943 (find_interesting_uses_op): Update call to record_group_use.
10944 (find_interesting_uses_cond): Likewise.
10945 (find_interesting_uses_address): Likewise.
10946 (get_mem_type_for_internal_fn): New function.
10947 (find_address_like_use): Likewise.
10948 (find_interesting_uses_stmt): Try find_address_like_use before
10949 calling find_interesting_uses_op.
10950 (addr_offset_valid_p): Use the iv mem_type field as the type
10951 of the addressed memory.
10952 (add_autoinc_candidates): Likewise.
10953 (get_address_cost): Likewise.
10954 (split_small_address_groups_p): Use address_p.
10955 (split_address_groups): Likewise.
10956 (add_iv_candidate_for_use): Likewise.
10957 (autoinc_possible_for_pair): Likewise.
10958 (rewrite_groups): Likewise.
10959 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10960 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10961 (get_alias_ptr_type_for_ptr_address): New function.
10962 (rewrite_use_address): Rewrite address uses in calls that were
10963 identified by find_address_like_use.
10965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10966 Alan Hayward <alan.hayward@arm.com>
10967 David Sherwood <david.sherwood@arm.com>
10969 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10971 * gimple-expr.h (is_gimple_addressable: Likewise.
10972 * gimple-expr.c (is_gimple_address): Likewise.
10973 * internal-fn.c (expand_call_mem_ref): New function.
10974 (expand_mask_load_optab_fn): Use it.
10975 (expand_mask_store_optab_fn): Likewise.
10977 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10978 Alan Hayward <alan.hayward@arm.com>
10979 David Sherwood <david.sherwood@arm.com>
10981 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10982 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10983 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10984 (cond_umax@var{mode}): Document.
10985 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10986 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10987 (cond_umin_optab, cond_umax_optab): New optabs.
10988 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10989 (COND_IOR, COND_XOR): New internal functions.
10990 * internal-fn.h (get_conditional_internal_fn): Declare.
10991 * internal-fn.c (cond_binary_direct): New macro.
10992 (expand_cond_binary_optab_fn): Likewise.
10993 (direct_cond_binary_optab_supported_p): Likewise.
10994 (get_conditional_internal_fn): New function.
10995 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10996 Cope with reduction statements that are vectorized as calls rather
10998 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10999 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11000 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11001 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11002 (UNSPEC_COND_EOR): New unspecs.
11003 (optab): Add mappings for them.
11004 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11005 (sve_int_op, sve_fp_op): New int attributes.
11007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11008 Alan Hayward <alan.hayward@arm.com>
11009 David Sherwood <david.sherwood@arm.com>
11011 * optabs.def (while_ult_optab): New optab.
11012 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11013 * internal-fn.def (WHILE_ULT): New internal function.
11014 * internal-fn.h (direct_internal_fn_supported_p): New override
11015 that takes two types as argument.
11016 * internal-fn.c (while_direct): New macro.
11017 (expand_while_optab_fn): New function.
11018 (convert_optab_supported_p): Likewise.
11019 (direct_while_optab_supported_p): New macro.
11020 * wide-int.h (wi::udiv_ceil): New function.
11021 * tree-vectorizer.h (rgroup_masks): New structure.
11022 (vec_loop_masks): New typedef.
11023 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11024 and fully_masked_p.
11025 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11026 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11027 (vect_max_vf): New function.
11028 (slpeel_make_loop_iterate_ntimes): Delete.
11029 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11030 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11031 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11032 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11033 internal-fn.h, stor-layout.h and optabs-query.h.
11034 (vect_set_loop_mask): New function.
11035 (add_preheader_seq): Likewise.
11036 (add_header_seq): Likewise.
11037 (interleave_supported_p): Likewise.
11038 (vect_maybe_permute_loop_masks): Likewise.
11039 (vect_set_loop_masks_directly): Likewise.
11040 (vect_set_loop_condition_masked): Likewise.
11041 (vect_set_loop_condition_unmasked): New function, split out from
11042 slpeel_make_loop_iterate_ntimes.
11043 (slpeel_make_loop_iterate_ntimes): Rename to..
11044 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11045 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11046 (vect_do_peeling): Update call accordingly.
11047 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11049 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11050 mask_compare_type, can_fully_mask_p and fully_masked_p.
11051 (release_vec_loop_masks): New function.
11052 (_loop_vec_info): Use it to free the loop masks.
11053 (can_produce_all_loop_masks_p): New function.
11054 (vect_get_max_nscalars_per_iter): Likewise.
11055 (vect_verify_full_masking): Likewise.
11056 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11057 retries, and free the mask rgroups before retrying. Check loop-wide
11058 reasons for disallowing fully-masked loops. Make the final decision
11059 about whether use a fully-masked loop or not.
11060 (vect_estimate_min_profitable_iters): Do not assume that peeling
11061 for the number of iterations will be needed for fully-masked loops.
11062 (vectorizable_reduction): Disable fully-masked loops.
11063 (vectorizable_live_operation): Likewise.
11064 (vect_halve_mask_nunits): New function.
11065 (vect_double_mask_nunits): Likewise.
11066 (vect_record_loop_mask): Likewise.
11067 (vect_get_loop_mask): Likewise.
11068 (vect_transform_loop): Handle the case in which the final loop
11069 iteration might handle a partial vector. Call vect_set_loop_condition
11070 instead of slpeel_make_loop_iterate_ntimes.
11071 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11072 (check_load_store_masking): New function.
11073 (prepare_load_store_mask): Likewise.
11074 (vectorizable_store): Handle fully-masked loops.
11075 (vectorizable_load): Likewise.
11076 (supportable_widening_operation): Use vect_halve_mask_nunits for
11078 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11079 (vect_gen_while): New function.
11080 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11081 (aarch64_uqdec<mode>): New insn.
11083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11084 Alan Hayward <alan.hayward@arm.com>
11085 David Sherwood <david.sherwood@arm.com>
11087 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11088 (reduc_xor_scal_optab): New optabs.
11089 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11090 (reduc_xor_scal_@var{m}): Document.
11091 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11092 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11093 internal functions.
11094 * fold-const-call.c (fold_const_call): Handle them.
11095 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11096 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11097 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11098 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11099 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11100 (UNSPEC_XORV): New unspecs.
11101 (optab): Add entries for them.
11102 (BITWISEV): New int iterator.
11103 (bit_reduc_op): New int attributes.
11105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11106 Alan Hayward <alan.hayward@arm.com>
11107 David Sherwood <david.sherwood@arm.com>
11109 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11110 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11111 * optabs.def (vec_shl_insert_optab): New optab.
11112 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11113 (duplicate_and_interleave): Likewise.
11114 * tree-vect-loop.c: Include internal-fn.h.
11115 (neutral_op_for_slp_reduction): New function, split out from
11116 get_initial_defs_for_reduction.
11117 (get_initial_def_for_reduction): Handle option 2 for variable-length
11118 vectors by loading the neutral value into a vector and then shifting
11119 the initial value into element 0.
11120 (get_initial_defs_for_reduction): Replace the code argument with
11121 the neutral value calculated by neutral_op_for_slp_reduction.
11122 Use gimple_build_vector for constant-length vectors.
11123 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11124 but the first group_size elements have a neutral value.
11125 Use duplicate_and_interleave otherwise.
11126 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11127 Update call to get_initial_defs_for_reduction. Handle SLP
11128 reductions for variable-length vectors by creating one vector
11129 result for each scalar result, with the elements associated
11130 with other scalar results stubbed out with the neutral value.
11131 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11132 Require IFN_VEC_SHL_INSERT for double reductions on
11133 variable-length vectors, or SLP reductions that have
11134 a neutral value. Require can_duplicate_and_interleave_p
11135 support for variable-length unchained SLP reductions if there
11136 is no neutral value, such as for MIN/MAX reductions. Also require
11137 the number of vector elements to be a multiple of the number of
11138 SLP statements when doing variable-length unchained SLP reductions.
11139 Update call to vect_create_epilog_for_reduction.
11140 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11141 and remove initial values.
11142 (duplicate_and_interleave): Make public.
11143 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11144 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11146 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11147 Alan Hayward <alan.hayward@arm.com>
11148 David Sherwood <david.sherwood@arm.com>
11150 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11151 (can_duplicate_and_interleave_p): New function.
11152 (vect_get_and_check_slp_defs): Take the vector of statements
11153 rather than just the current one. Remove excess parentheses.
11154 Restriction rejectinon of vect_constant_def and vect_external_def
11155 for variable-length vectors to boolean types, or types for which
11156 can_duplicate_and_interleave_p is false.
11157 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11158 (duplicate_and_interleave): New function.
11159 (vect_get_constant_vectors): Use gimple_build_vector for
11160 constant-length vectors and suitable variable-length constant
11161 vectors. Use duplicate_and_interleave for other variable-length
11162 vectors. Don't defer the update when inserting new statements.
11164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11165 Alan Hayward <alan.hayward@arm.com>
11166 David Sherwood <david.sherwood@arm.com>
11168 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11169 min_profitable_iters doesn't go negative.
11171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11172 Alan Hayward <alan.hayward@arm.com>
11173 David Sherwood <david.sherwood@arm.com>
11175 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11176 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11177 * optabs.def (vec_mask_load_lanes_optab): New optab.
11178 (vec_mask_store_lanes_optab): Likewise.
11179 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11180 (MASK_STORE_LANES): Likewise.
11181 * internal-fn.c (mask_load_lanes_direct): New macro.
11182 (mask_store_lanes_direct): Likewise.
11183 (expand_mask_load_optab_fn): Handle masked operations.
11184 (expand_mask_load_lanes_optab_fn): New macro.
11185 (expand_mask_store_optab_fn): Handle masked operations.
11186 (expand_mask_store_lanes_optab_fn): New macro.
11187 (direct_mask_load_lanes_optab_supported_p): Likewise.
11188 (direct_mask_store_lanes_optab_supported_p): Likewise.
11189 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11191 (vect_load_lanes_supported): Likewise.
11192 * tree-vect-data-refs.c (strip_conversion): New function.
11193 (can_group_stmts_p): Likewise.
11194 (vect_analyze_data_ref_accesses): Use it instead of checking
11195 for a pair of assignments.
11196 (vect_store_lanes_supported): Take a masked_p parameter.
11197 (vect_load_lanes_supported): Likewise.
11198 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11199 vect_store_lanes_supported and vect_load_lanes_supported.
11200 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11201 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11202 parameter. Don't allow gaps for masked accesses.
11203 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11204 and vect_load_lanes_supported.
11205 (get_load_store_type): Take a masked_p parameter and update
11206 call to get_group_load_store_type.
11207 (vectorizable_store): Update call to get_load_store_type.
11208 Handle IFN_MASK_STORE_LANES.
11209 (vectorizable_load): Update call to get_load_store_type.
11210 Handle IFN_MASK_LOAD_LANES.
11212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11213 Alan Hayward <alan.hayward@arm.com>
11214 David Sherwood <david.sherwood@arm.com>
11216 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11218 * config/aarch64/aarch64-protos.h
11219 (aarch64_sve_struct_memory_operand_p): Declare.
11220 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11221 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11222 (VPRED, vpred): Handle SVE structure modes.
11223 * config/aarch64/constraints.md (Utx): New constraint.
11224 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11225 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11226 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11227 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11228 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11229 structure modes. Split into pieces after RA.
11230 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11231 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11233 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11234 SVE structure modes.
11235 (aarch64_classify_address): Likewise.
11236 (sizetochar): Move earlier in file.
11237 (aarch64_print_operand): Handle SVE register lists.
11238 (aarch64_array_mode): New function.
11239 (aarch64_sve_struct_memory_operand_p): Likewise.
11240 (TARGET_ARRAY_MODE): Redefine.
11242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11243 Alan Hayward <alan.hayward@arm.com>
11244 David Sherwood <david.sherwood@arm.com>
11246 * target.def (array_mode): New target hook.
11247 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11248 * doc/tm.texi: Regenerate.
11249 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11250 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11251 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11252 targetm.array_mode.
11253 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11257 Alan Hayward <alan.hayward@arm.com>
11258 David Sherwood <david.sherwood@arm.com>
11260 * fold-const.c (fold_binary_loc): Check the argument types
11261 rather than the result type when testing for a vector operation.
11263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11265 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11266 * doc/tm.texi: Regenerate.
11268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11269 Alan Hayward <alan.hayward@arm.com>
11270 David Sherwood <david.sherwood@arm.com>
11272 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11273 (sve): Document new AArch64 extension.
11274 * doc/md.texi (w): Extend the description of the AArch64
11275 constraint to include SVE vectors.
11276 (Upl, Upa): Document new AArch64 predicate constraints.
11277 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11279 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11280 (msve-vector-bits=): New option.
11281 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11282 SVE when these are disabled.
11283 (sve): New extension.
11284 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11285 modes. Adjust their number of units based on aarch64_sve_vg.
11286 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11287 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11288 aarch64_addr_query_type.
11289 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11290 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11291 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11292 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11293 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11294 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11295 (aarch64_simd_imm_zero_p): Delete.
11296 (aarch64_check_zero_based_sve_index_immediate): Declare.
11297 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11298 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11299 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11300 (aarch64_sve_float_mul_immediate_p): Likewise.
11301 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11302 rather than an rtx.
11303 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11304 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11305 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11306 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11307 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11308 (aarch64_regmode_natural_size): Likewise.
11309 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11310 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11312 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11313 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11314 for VG and the SVE predicate registers.
11315 (V_ALIASES): Add a "z"-prefixed alias.
11316 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11317 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11318 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11319 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11320 (REG_CLASS_NAMES): Add entries for them.
11321 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11322 and the predicate registers.
11323 (aarch64_sve_vg): Declare.
11324 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11325 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11326 (REGMODE_NATURAL_SIZE): Define.
11327 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11329 * config/aarch64/aarch64.c: Include cfgrtl.h.
11330 (simd_immediate_info): Add a constructor for series vectors,
11331 and an associated step field.
11332 (aarch64_sve_vg): New variable.
11333 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11334 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11335 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11336 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11337 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11338 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11339 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11340 (aarch64_get_mask_mode): New functions.
11341 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11342 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11343 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11344 predicate modes and predicate registers. Explicitly restrict
11345 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11346 to store a vector mode if it is recognized by
11347 aarch64_classify_vector_mode.
11348 (aarch64_regmode_natural_size): New function.
11349 (aarch64_hard_regno_caller_save_mode): Return the original mode
11351 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11352 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11353 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11354 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11356 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11357 does not overlap dest if the function is frame-related. Handle
11359 (aarch64_split_add_offset): New function.
11360 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11361 them aarch64_add_offset.
11362 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11363 and update call to aarch64_sub_sp.
11364 (aarch64_add_cfa_expression): New function.
11365 (aarch64_expand_prologue): Pass extra temporary registers to the
11366 functions above. Handle the case in which we need to emit new
11367 DW_CFA_expressions for registers that were originally saved
11368 relative to the stack pointer, but now have to be expressed
11369 relative to the frame pointer.
11370 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11372 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11373 IP0 and IP1 values for SVE frames.
11374 (aarch64_expand_vec_series): New function.
11375 (aarch64_expand_sve_widened_duplicate): Likewise.
11376 (aarch64_expand_sve_const_vector): Likewise.
11377 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11378 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11379 into the register, rather than emitting a SET directly.
11380 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11381 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11382 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11383 (offset_9bit_signed_scaled_p): New functions.
11384 (aarch64_replicate_bitmask_imm): New function.
11385 (aarch64_bitmask_imm): Use it.
11386 (aarch64_cannot_force_const_mem): Reject expressions involving
11387 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11388 (aarch64_classify_index): Handle SVE indices, by requiring
11389 a plain register index with a scale that matches the element size.
11390 (aarch64_classify_address): Handle SVE addresses. Assert that
11391 the mode of the address is VOIDmode or an integer mode.
11392 Update call to aarch64_classify_symbol.
11393 (aarch64_classify_symbolic_expression): Update call to
11394 aarch64_classify_symbol.
11395 (aarch64_const_vec_all_in_range_p): New function.
11396 (aarch64_print_vector_float_operand): Likewise.
11397 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11398 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11399 and the FP immediates 1.0 and 0.5.
11400 (aarch64_print_address_internal): Handle SVE addresses.
11401 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11402 (aarch64_regno_regclass): Handle predicate registers.
11403 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11405 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11406 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11407 (aarch64_convert_sve_vector_bits): New function.
11408 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11409 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11410 rather than an rtx.
11411 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11412 Handle SVE vector and predicate modes. Accept VL-based constants
11413 that need only one temporary register, and VL offsets that require
11414 no temporary registers.
11415 (aarch64_conditional_register_usage): Mark the predicate registers
11416 as fixed if SVE isn't available.
11417 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11418 Return true for SVE vector and predicate modes.
11419 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11420 rather than an unsigned int. Handle SVE modes.
11421 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11423 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11425 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11426 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11427 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11428 (aarch64_sve_float_mul_immediate_p): New functions.
11429 (aarch64_sve_valid_immediate): New function.
11430 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11431 Explicitly reject structure modes. Check for INDEX constants.
11432 Handle PTRUE and PFALSE constants.
11433 (aarch64_check_zero_based_sve_index_immediate): New function.
11434 (aarch64_simd_imm_zero_p): Delete.
11435 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11436 vector modes. Accept constants in the range of CNT[BHWD].
11437 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11438 ask for an Advanced SIMD mode.
11439 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11440 (aarch64_simd_vector_alignment): Handle SVE predicates.
11441 (aarch64_vectorize_preferred_vector_alignment): New function.
11442 (aarch64_simd_vector_alignment_reachable): Use it instead of
11444 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11445 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11447 (MAX_VECT_LEN): Delete.
11448 (expand_vec_perm_d): Add a vec_flags field.
11449 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11450 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11451 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11453 (aarch64_evpc_rev): Rename to...
11454 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11455 (aarch64_evpc_rev_global): New function.
11456 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11457 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11459 (aarch64_evpc_sve_tbl): New function.
11460 (aarch64_expand_vec_perm_const_1): Update after rename of
11461 aarch64_evpc_rev. Handle SVE permutes too, trying
11462 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11463 than aarch64_evpc_tbl.
11464 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11465 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11466 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11467 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11468 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11469 (aarch64_expand_sve_vcond): New functions.
11470 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11471 of aarch64_vector_mode_p.
11472 (aarch64_dwarf_poly_indeterminate_value): New function.
11473 (aarch64_compute_pressure_classes): Likewise.
11474 (aarch64_can_change_mode_class): Likewise.
11475 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11476 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11477 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11478 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11479 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11480 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11481 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11482 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11484 (Dn, Dl, Dr): Accept const as well as const_vector.
11485 (Dz): Likewise. Compare against CONST0_RTX.
11486 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11487 of "vector" where appropriate.
11488 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11489 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11490 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11491 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11492 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11493 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11494 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11495 (v_int_equiv): Extend to SVE modes.
11496 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11498 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11499 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11500 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11501 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11502 (SVE_COND_FP_CMP): New int iterators.
11503 (perm_hilo): Handle the new unpack unspecs.
11504 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11506 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11507 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11508 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11509 (aarch64_equality_operator, aarch64_constant_vector_operand)
11510 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11511 (aarch64_sve_nonimmediate_operand): Likewise.
11512 (aarch64_sve_general_operand): Likewise.
11513 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11514 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11515 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11516 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11517 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11518 (aarch64_sve_float_arith_immediate): Likewise.
11519 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11520 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11521 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11522 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11523 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11524 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11525 (aarch64_sve_float_arith_operand): Likewise.
11526 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11527 (aarch64_sve_float_mul_operand): Likewise.
11528 (aarch64_sve_vec_perm_operand): Likewise.
11529 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11530 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11531 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11532 as well as const_vector.
11533 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11534 in file. Use CONST0_RTX and CONSTM1_RTX.
11535 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11536 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11537 Use aarch64_simd_imm_zero.
11538 * config/aarch64/aarch64-sve.md: New file.
11539 * config/aarch64/aarch64.md: Include it.
11540 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11541 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11542 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11543 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11544 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11545 (sve): New attribute.
11546 (enabled): Disable instructions with the sve attribute unless
11548 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11549 aarch64_expand_mov_immediate.
11550 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11551 CNT[BHSD] immediates.
11552 (movti): Split CONST_POLY_INT moves into two halves.
11553 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11554 Split additions that need a temporary here if the destination
11555 is the stack pointer.
11556 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11557 (*add<mode>3_poly_1): New instruction.
11558 (set_clobber_cc): New expander.
11560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11562 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11563 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11564 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11565 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11566 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11567 Change innermode from fixed_mode_size to machine_mode.
11568 (simplify_subreg): Update call accordingly. Handle a constant-sized
11569 subreg of a variable-length CONST_VECTOR.
11571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11572 Alan Hayward <alan.hayward@arm.com>
11573 David Sherwood <david.sherwood@arm.com>
11575 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11576 (add_offset_to_base): New function, split out from...
11577 (create_mem_ref): ...here. When handling a scale other than 1,
11578 check first whether the address is valid without the offset.
11579 Add it into the base if so, leaving the index and scale as-is.
11581 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11584 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11585 fold_for_warn before checking if arg2 is INTEGER_CST.
11587 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11589 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11590 (store_multiple_operation): Delete.
11591 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11592 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11593 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11594 guarded by TARGET_STRING.
11595 (rs6000_output_load_multiple): Delete.
11596 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11597 OPTION_MASK_STRING / TARGET_STRING handling.
11598 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11599 (const rs6000_opt_masks) <"string">: Change mask to 0.
11600 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11601 (MASK_STRING): Delete.
11602 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11604 (load_multiple): Delete.
11611 (store_multiple): Delete.
11618 (movmemsi_8reg): Delete.
11619 (corresponding unnamed define_insn): Delete.
11620 (movmemsi_6reg): Delete.
11621 (corresponding unnamed define_insn): Delete.
11622 (movmemsi_4reg): Delete.
11623 (corresponding unnamed define_insn): Delete.
11624 (movmemsi_2reg): Delete.
11625 (corresponding unnamed define_insn): Delete.
11626 (movmemsi_1reg): Delete.
11627 (corresponding unnamed define_insn): Delete.
11628 * config/rs6000/rs6000.opt (mno-string): New.
11629 (mstring): Replace by deprecation warning stub.
11630 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11632 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11634 * regrename.c (regrename_do_replace): If replacing the same
11635 reg multiple times, try to reuse last created gen_raw_REG.
11638 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11639 main to workaround a bug in GDB.
11641 2018-01-12 Tom de Vries <tom@codesourcery.com>
11644 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11646 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11648 PR rtl-optimization/80481
11649 * ira-color.c (get_cap_member): New function.
11650 (allocnos_conflict_by_live_ranges_p): Use it.
11651 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11652 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11654 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11657 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11658 (*saddl_se_1): Ditto.
11659 (*ssubsi_1): Ditto.
11660 (*ssubl_se_1): Ditto.
11662 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11664 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11665 rather than wi::to_widest for DR_INITs.
11666 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11667 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11668 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11670 (vect_analyze_group_access_1): Note that here.
11672 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11674 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11675 polynomial type sizes.
11677 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11679 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11680 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11681 (gimple_add_tmp_var): Likewise.
11683 2018-01-12 Martin Liska <mliska@suse.cz>
11685 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11686 (gimple_alloc_sizes): Likewise.
11687 (dump_gimple_statistics): Use PRIu64 in printf format.
11688 * gimple.h: Change uint64_t to int.
11690 2018-01-12 Martin Liska <mliska@suse.cz>
11692 * tree-core.h: Use uint64_t instead of int.
11693 * tree.c (tree_node_counts): Likewise.
11694 (tree_node_sizes): Likewise.
11695 (dump_tree_statistics): Use PRIu64 in printf format.
11697 2018-01-12 Martin Liska <mliska@suse.cz>
11699 * Makefile.in: As qsort_chk is implemented in vec.c, add
11700 vec.o to linkage of gencfn-macros.
11701 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11702 passing the info to record_node_allocation_statistics.
11703 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11705 * ggc-common.c (struct ggc_usage): Add operator== and use
11706 it in operator< and compare function.
11707 * mem-stats.h (struct mem_usage): Likewise.
11708 * vec.c (struct vec_usage): Remove operator< and compare
11709 function. Can be simply inherited.
11711 2018-01-12 Martin Jambor <mjambor@suse.cz>
11714 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11715 * tree-ssa-math-opts.c: Include domwalk.h.
11716 (convert_mult_to_fma_1): New function.
11717 (fma_transformation_info): New type.
11718 (fma_deferring_state): Likewise.
11719 (cancel_fma_deferring): New function.
11720 (result_of_phi): Likewise.
11721 (last_fma_candidate_feeds_initial_phi): Likewise.
11722 (convert_mult_to_fma): Added deferring logic, split actual
11723 transformation to convert_mult_to_fma_1.
11724 (math_opts_dom_walker): New type.
11725 (math_opts_dom_walker::after_dom_children): New method, body moved
11726 here from pass_optimize_widening_mul::execute, added deferring logic
11728 (pass_optimize_widening_mul::execute): Moved most of code to
11729 math_opts_dom_walker::after_dom_children.
11730 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11731 * config/i386/i386.c (ix86_option_override_internal): Added
11732 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11734 2018-01-12 Richard Biener <rguenther@suse.de>
11737 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11738 inline instance vars.
11740 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11743 * config/rx/rx.c (rx_is_restricted_memory_address):
11744 Handle SUBREG case.
11746 2018-01-12 Richard Biener <rguenther@suse.de>
11748 PR tree-optimization/80846
11749 * target.def (split_reduction): New target hook.
11750 * targhooks.c (default_split_reduction): New function.
11751 * targhooks.h (default_split_reduction): Declare.
11752 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11753 target requests first reduce vectors by combining low and high
11755 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11756 (get_vectype_for_scalar_type_and_size): Export.
11757 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11758 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11759 * doc/tm.texi: Regenerate.
11760 * config/i386/i386.c (ix86_split_reduction): Implement
11761 TARGET_VECTORIZE_SPLIT_REDUCTION.
11763 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11766 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11767 in PIC mode except for TARGET_VXWORKS_RTP.
11768 * config/sparc/sparc.c: Include cfgrtl.h.
11769 (TARGET_INIT_PIC_REG): Define.
11770 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11771 (sparc_pic_register_p): New predicate.
11772 (sparc_legitimate_address_p): Use it.
11773 (sparc_legitimize_pic_address): Likewise.
11774 (sparc_delegitimize_address): Likewise.
11775 (sparc_mode_dependent_address_p): Likewise.
11776 (gen_load_pcrel_sym): Remove 4th parameter.
11777 (load_got_register): Adjust call to above. Remove obsolete stuff.
11778 (sparc_expand_prologue): Do not call load_got_register here.
11779 (sparc_flat_expand_prologue): Likewise.
11780 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11781 (sparc_use_pseudo_pic_reg): New function.
11782 (sparc_init_pic_reg): Likewise.
11783 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11784 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11786 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11788 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11789 Add item for branch_cost.
11791 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11793 PR rtl-optimization/83565
11794 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11795 not extend the result to a larger mode for rotate operations.
11796 (num_sign_bit_copies1): Likewise.
11798 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11801 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11803 Use values-Xc.o for -pedantic.
11804 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11806 2018-01-12 Martin Liska <mliska@suse.cz>
11809 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11811 (possible_polymorphic_call_targets): Use it.
11812 (ipa_devirt): Likewise.
11814 2018-01-12 Martin Liska <mliska@suse.cz>
11816 * profile-count.h (enum profile_quality): Use 0 as invalid
11817 enum value of profile_quality.
11819 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11821 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11822 -mext-string options.
11824 2018-01-12 Richard Biener <rguenther@suse.de>
11826 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11827 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11828 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11830 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11832 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11834 * configure.ac (--with-long-double-format): Add support for the
11835 configuration option to change the default long double format on
11837 * config.gcc (powerpc*-linux*-*): Likewise.
11838 * configure: Regenerate.
11839 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11840 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11841 used without modification.
11843 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11845 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11846 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11847 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11848 MISC_BUILTIN_SPEC_BARRIER.
11849 (rs6000_init_builtins): Likewise.
11850 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11852 (speculation_barrier): New define_insn.
11853 * doc/extend.texi: Document __builtin_speculation_barrier.
11855 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11858 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11859 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11860 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11862 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11863 integral modes instead of "ss" and "sd".
11864 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11865 vectors with 32-bit and 64-bit elements.
11866 (vecdupssescalarmodesuffix): New mode attribute.
11867 (vec_dup<mode>): Use it.
11869 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11872 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11873 frame if argument is passed on stack.
11875 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11878 * ree.c (combine_reaching_defs): Optimize also
11879 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11880 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11882 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11884 PR middle-end/83189
11885 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11887 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11889 PR middle-end/83718
11890 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11891 after they are computed.
11893 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11895 PR tree-optimization/83695
11896 * gimple-loop-linterchange.cc
11897 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11898 reset cached scev information after interchange.
11899 (pass_linterchange::execute): Remove call to scev_reset_htab.
11901 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11903 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11904 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11905 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11906 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11907 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11908 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11909 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11910 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11911 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11912 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11913 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11914 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11915 (V_lane_reg): Likewise.
11916 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11918 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11919 (vfmal_lane_low<mode>_intrinsic,
11920 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11921 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11922 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11923 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11924 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11925 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11927 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11929 * config/arm/arm-cpus.in (fp16fml): New feature.
11930 (ALL_SIMD): Add fp16fml.
11931 (armv8.2-a): Add fp16fml as an option.
11932 (armv8.3-a): Likewise.
11933 (armv8.4-a): Add fp16fml as part of fp16.
11934 * config/arm/arm.h (TARGET_FP16FML): Define.
11935 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11937 * config/arm/arm-modes.def (V2HF): Define.
11938 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11939 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11940 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11941 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11942 vfmsl_low, vfmsl_high): New set of builtins.
11943 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11944 (vfml_op): New code attribute.
11945 (VFMLHALVES): New int iterator.
11946 (VFML, VFMLSEL): New mode attributes.
11947 (V_reg): Define mapping for V2HF.
11948 (V_hi, V_lo): New mode attributes.
11949 (VF_constraint): Likewise.
11950 (vfml_half, vfml_half_selector): New int attributes.
11951 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11953 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11954 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11956 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11957 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11958 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11959 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11961 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11962 Document new effective target and option set.
11964 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11966 * config/arm/arm-cpus.in (armv8_4): New feature.
11967 (ARMv8_4a): New fgroup.
11968 (armv8.4-a): New arch.
11969 * config/arm/arm-tables.opt: Regenerate.
11970 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11971 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11972 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11973 Add matching rules for -march=armv8.4-a and extensions.
11974 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11976 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11979 * config/rx/rx.md (BW): New mode attribute.
11980 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11982 2018-01-11 Richard Biener <rguenther@suse.de>
11984 PR tree-optimization/83435
11985 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11986 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11987 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11989 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11990 Alan Hayward <alan.hayward@arm.com>
11991 David Sherwood <david.sherwood@arm.com>
11993 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11995 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11996 (aarch64_print_address_internal): Use it to check for a zero offset.
11998 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11999 Alan Hayward <alan.hayward@arm.com>
12000 David Sherwood <david.sherwood@arm.com>
12002 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12003 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12004 Return a poly_int64 rather than a HOST_WIDE_INT.
12005 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12006 rather than a HOST_WIDE_INT.
12007 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12008 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12009 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12010 final_offset from HOST_WIDE_INT to poly_int64.
12011 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12012 to_constant when getting the number of units in an Advanced SIMD
12014 (aarch64_builtin_vectorized_function): Check for a constant number
12016 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12018 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12019 attribute instead of GET_MODE_NUNITS.
12020 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12021 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12022 GET_MODE_SIZE for fixed-size registers.
12023 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12024 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12025 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12026 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12027 (aarch64_print_operand, aarch64_print_address_internal)
12028 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12029 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12030 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12031 Handle polynomial GET_MODE_SIZE.
12032 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12033 wider than SImode without modification.
12034 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12035 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12036 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12037 passing and returning SVE modes.
12038 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12039 rather than GEN_INT.
12040 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12041 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12042 (aarch64_allocate_and_probe_stack_space): Likewise.
12043 (aarch64_layout_frame): Cope with polynomial offsets.
12044 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12045 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12046 polynomial offsets.
12047 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12048 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12049 poly_int64 rather than a HOST_WIDE_INT.
12050 (aarch64_get_separate_components, aarch64_process_components)
12051 (aarch64_expand_prologue, aarch64_expand_epilogue)
12052 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12053 (aarch64_anchor_offset): New function, split out from...
12054 (aarch64_legitimize_address): ...here.
12055 (aarch64_builtin_vectorization_cost): Handle polynomial
12056 TYPE_VECTOR_SUBPARTS.
12057 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12059 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12060 number of elements from the PARALLEL rather than the mode.
12061 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12062 rather than GET_MODE_BITSIZE.
12063 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12064 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12065 (aarch64_expand_vec_perm_const_1): Handle polynomial
12066 d->perm.length () and d->perm elements.
12067 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12068 Apply to_constant to d->perm elements.
12069 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12070 polynomial CONST_VECTOR_NUNITS.
12071 (aarch64_move_pointer): Take amount as a poly_int64 rather
12073 (aarch64_progress_pointer): Avoid temporary variable.
12074 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12075 the mode attribute instead of GET_MODE.
12077 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12078 Alan Hayward <alan.hayward@arm.com>
12079 David Sherwood <david.sherwood@arm.com>
12081 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12082 x exists before using it.
12083 (aarch64_add_constant_internal): Rename to...
12084 (aarch64_add_offset_1): ...this. Replace regnum with separate
12085 src and dest rtxes. Handle the case in which they're different,
12086 including when the offset is zero. Replace scratchreg with an rtx.
12087 Use 2 additions if there is no spare register into which we can
12088 move a 16-bit constant.
12089 (aarch64_add_constant): Delete.
12090 (aarch64_add_offset): Replace reg with separate src and dest
12091 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12092 Use aarch64_add_offset_1.
12093 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12094 an rtx rather than an int. Take the delta as a poly_int64
12095 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12096 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12097 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12098 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12099 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12100 and aarch64_add_sp.
12101 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12102 aarch64_add_constant.
12104 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12106 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12107 Use scalar_float_mode.
12109 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12111 * config/aarch64/aarch64-simd.md
12112 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12113 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12114 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12115 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12116 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12117 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12118 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12119 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12120 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12121 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12123 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12126 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12127 targ_options->x_arm_arch_string is non NULL.
12129 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12131 * config/aarch64/aarch64.h
12132 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12134 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12137 * expmed.c (emit_store_flag_force): Swap if const op0
12138 and change VOIDmode to mode of op0.
12140 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12142 PR rtl-optimization/83761
12143 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12144 than bytes to mode_for_size.
12146 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12148 PR middle-end/83189
12149 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12150 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12153 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12155 PR middle-end/83575
12156 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12157 when in layout mode.
12158 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12159 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12162 2018-01-10 Michael Collison <michael.collison@arm.com>
12164 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12165 * config/aarch64/aarch64-option-extension.def: Add
12166 AARCH64_OPT_EXTENSION of 'fp16fml'.
12167 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12168 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12169 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12170 * config/aarch64/constraints.md (Ui7): New constraint.
12171 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12172 (VFMLA_SEL_W): Ditto.
12175 (VFMLA16_LOW): New int iterator.
12176 (VFMLA16_HIGH): Ditto.
12177 (UNSPEC_FMLAL): New unspec.
12178 (UNSPEC_FMLSL): Ditto.
12179 (UNSPEC_FMLAL2): Ditto.
12180 (UNSPEC_FMLSL2): Ditto.
12181 (f16mac): New code attribute.
12182 * config/aarch64/aarch64-simd-builtins.def
12183 (aarch64_fmlal_lowv2sf): Ditto.
12184 (aarch64_fmlsl_lowv2sf): Ditto.
12185 (aarch64_fmlalq_lowv4sf): Ditto.
12186 (aarch64_fmlslq_lowv4sf): Ditto.
12187 (aarch64_fmlal_highv2sf): Ditto.
12188 (aarch64_fmlsl_highv2sf): Ditto.
12189 (aarch64_fmlalq_highv4sf): Ditto.
12190 (aarch64_fmlslq_highv4sf): Ditto.
12191 (aarch64_fmlal_lane_lowv2sf): Ditto.
12192 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12193 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12194 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12195 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12196 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12197 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12198 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12199 (aarch64_fmlal_lane_highv2sf): Ditto.
12200 (aarch64_fmlsl_lane_highv2sf): Ditto.
12201 (aarch64_fmlal_laneq_highv2sf): Ditto.
12202 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12203 (aarch64_fmlalq_lane_highv4sf): Ditto.
12204 (aarch64_fmlsl_lane_highv4sf): Ditto.
12205 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12206 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12207 * config/aarch64/aarch64-simd.md:
12208 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12209 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12210 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12211 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12212 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12213 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12214 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12215 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12216 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12217 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12218 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12219 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12220 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12221 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12222 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12223 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12224 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12225 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12226 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12227 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12228 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12229 (vfmlsl_low_u32): Ditto.
12230 (vfmlalq_low_u32): Ditto.
12231 (vfmlslq_low_u32): Ditto.
12232 (vfmlal_high_u32): Ditto.
12233 (vfmlsl_high_u32): Ditto.
12234 (vfmlalq_high_u32): Ditto.
12235 (vfmlslq_high_u32): Ditto.
12236 (vfmlal_lane_low_u32): Ditto.
12237 (vfmlsl_lane_low_u32): Ditto.
12238 (vfmlal_laneq_low_u32): Ditto.
12239 (vfmlsl_laneq_low_u32): Ditto.
12240 (vfmlalq_lane_low_u32): Ditto.
12241 (vfmlslq_lane_low_u32): Ditto.
12242 (vfmlalq_laneq_low_u32): Ditto.
12243 (vfmlslq_laneq_low_u32): Ditto.
12244 (vfmlal_lane_high_u32): Ditto.
12245 (vfmlsl_lane_high_u32): Ditto.
12246 (vfmlal_laneq_high_u32): Ditto.
12247 (vfmlsl_laneq_high_u32): Ditto.
12248 (vfmlalq_lane_high_u32): Ditto.
12249 (vfmlslq_lane_high_u32): Ditto.
12250 (vfmlalq_laneq_high_u32): Ditto.
12251 (vfmlslq_laneq_high_u32): Ditto.
12252 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12253 (AARCH64_FL_FOR_ARCH8_4): New.
12254 (AARCH64_ISA_F16FML): New ISA flag.
12255 (TARGET_F16FML): New feature flag for fp16fml.
12256 (doc/invoke.texi): Document new fp16fml option.
12258 2018-01-10 Michael Collison <michael.collison@arm.com>
12260 * config/aarch64/aarch64-builtins.c:
12261 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12262 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12263 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12264 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12265 (AARCH64_ISA_SHA3): New ISA flag.
12266 (TARGET_SHA3): New feature flag for sha3.
12267 * config/aarch64/iterators.md (sha512_op): New int attribute.
12268 (CRYPTO_SHA512): New int iterator.
12269 (UNSPEC_SHA512H): New unspec.
12270 (UNSPEC_SHA512H2): Ditto.
12271 (UNSPEC_SHA512SU0): Ditto.
12272 (UNSPEC_SHA512SU1): Ditto.
12273 * config/aarch64/aarch64-simd-builtins.def
12274 (aarch64_crypto_sha512hqv2di): New builtin.
12275 (aarch64_crypto_sha512h2qv2di): Ditto.
12276 (aarch64_crypto_sha512su0qv2di): Ditto.
12277 (aarch64_crypto_sha512su1qv2di): Ditto.
12278 (aarch64_eor3qv8hi): Ditto.
12279 (aarch64_rax1qv2di): Ditto.
12280 (aarch64_xarqv2di): Ditto.
12281 (aarch64_bcaxqv8hi): Ditto.
12282 * config/aarch64/aarch64-simd.md:
12283 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12284 (aarch64_crypto_sha512su0qv2di): Ditto.
12285 (aarch64_crypto_sha512su1qv2di): Ditto.
12286 (aarch64_eor3qv8hi): Ditto.
12287 (aarch64_rax1qv2di): Ditto.
12288 (aarch64_xarqv2di): Ditto.
12289 (aarch64_bcaxqv8hi): Ditto.
12290 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12291 (vsha512h2q_u64): Ditto.
12292 (vsha512su0q_u64): Ditto.
12293 (vsha512su1q_u64): Ditto.
12294 (veor3q_u16): Ditto.
12295 (vrax1q_u64): Ditto.
12296 (vxarq_u64): Ditto.
12297 (vbcaxq_u16): Ditto.
12298 * config/arm/types.md (crypto_sha512): New type attribute.
12299 (crypto_sha3): Ditto.
12300 (doc/invoke.texi): Document new sha3 option.
12302 2018-01-10 Michael Collison <michael.collison@arm.com>
12304 * config/aarch64/aarch64-builtins.c:
12305 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12306 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12307 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12308 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12309 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12310 (AARCH64_ISA_SM4): New ISA flag.
12311 (TARGET_SM4): New feature flag for sm4.
12312 * config/aarch64/aarch64-simd-builtins.def
12313 (aarch64_sm3ss1qv4si): Ditto.
12314 (aarch64_sm3tt1aq4si): Ditto.
12315 (aarch64_sm3tt1bq4si): Ditto.
12316 (aarch64_sm3tt2aq4si): Ditto.
12317 (aarch64_sm3tt2bq4si): Ditto.
12318 (aarch64_sm3partw1qv4si): Ditto.
12319 (aarch64_sm3partw2qv4si): Ditto.
12320 (aarch64_sm4eqv4si): Ditto.
12321 (aarch64_sm4ekeyqv4si): Ditto.
12322 * config/aarch64/aarch64-simd.md:
12323 (aarch64_sm3ss1qv4si): Ditto.
12324 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12325 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12326 (aarch64_sm4eqv4si): Ditto.
12327 (aarch64_sm4ekeyqv4si): Ditto.
12328 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12329 (sm3part_op): Ditto.
12330 (CRYPTO_SM3TT): Ditto.
12331 (CRYPTO_SM3PART): Ditto.
12332 (UNSPEC_SM3SS1): New unspec.
12333 (UNSPEC_SM3TT1A): Ditto.
12334 (UNSPEC_SM3TT1B): Ditto.
12335 (UNSPEC_SM3TT2A): Ditto.
12336 (UNSPEC_SM3TT2B): Ditto.
12337 (UNSPEC_SM3PARTW1): Ditto.
12338 (UNSPEC_SM3PARTW2): Ditto.
12339 (UNSPEC_SM4E): Ditto.
12340 (UNSPEC_SM4EKEY): Ditto.
12341 * config/aarch64/constraints.md (Ui2): New constraint.
12342 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12343 * config/arm/types.md (crypto_sm3): New type attribute.
12344 (crypto_sm4): Ditto.
12345 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12346 (vsm3tt1aq_u32): Ditto.
12347 (vsm3tt1bq_u32): Ditto.
12348 (vsm3tt2aq_u32): Ditto.
12349 (vsm3tt2bq_u32): Ditto.
12350 (vsm3partw1q_u32): Ditto.
12351 (vsm3partw2q_u32): Ditto.
12352 (vsm4eq_u32): Ditto.
12353 (vsm4ekeyq_u32): Ditto.
12354 (doc/invoke.texi): Document new sm4 option.
12356 2018-01-10 Michael Collison <michael.collison@arm.com>
12358 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12359 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12360 (AARCH64_FL_FOR_ARCH8_4): New.
12361 (AARCH64_FL_V8_4): New flag.
12362 (doc/invoke.texi): Document new armv8.4-a option.
12364 2018-01-10 Michael Collison <michael.collison@arm.com>
12366 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12367 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12368 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12369 * config/aarch64/aarch64-option-extension.def: Add
12370 AARCH64_OPT_EXTENSION of 'sha2'.
12371 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12372 (crypto): Disable sha2 and aes if crypto disabled.
12373 (crypto): Enable aes and sha2 if enabled.
12374 (simd): Disable sha2 and aes if simd disabled.
12375 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12377 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12378 (TARGET_SHA2): New feature flag for sha2.
12379 (TARGET_AES): New feature flag for aes.
12380 * config/aarch64/aarch64-simd.md:
12381 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12382 conditional on TARGET_AES.
12383 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12384 (aarch64_crypto_sha1hsi): Make pattern conditional
12386 (aarch64_crypto_sha1hv4si): Ditto.
12387 (aarch64_be_crypto_sha1hv4si): Ditto.
12388 (aarch64_crypto_sha1su1v4si): Ditto.
12389 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12390 (aarch64_crypto_sha1su0v4si): Ditto.
12391 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12392 (aarch64_crypto_sha256su0v4si): Ditto.
12393 (aarch64_crypto_sha256su1v4si): Ditto.
12394 (doc/invoke.texi): Document new aes and sha2 options.
12396 2018-01-10 Martin Sebor <msebor@redhat.com>
12398 PR tree-optimization/83781
12399 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12402 2018-01-11 Martin Sebor <msebor@gmail.com>
12403 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12405 PR tree-optimization/83501
12406 PR tree-optimization/81703
12408 * tree-ssa-strlen.c (get_string_cst): Rename...
12409 (get_string_len): ...to this. Handle global constants.
12410 (handle_char_store): Adjust.
12412 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12413 Jim Wilson <jimw@sifive.com>
12415 * config/riscv/riscv-protos.h (riscv_output_return): New.
12416 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12417 (riscv_attribute_table, riscv_output_return),
12418 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12419 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12420 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12422 (riscv_expand_prologue): Add early return for naked function.
12423 (riscv_expand_epilogue): Likewise.
12424 (riscv_function_ok_for_sibcall): Return false for naked function.
12425 (riscv_set_current_function): New.
12426 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12427 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12428 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12429 * doc/extend.texi (RISC-V Function Attributes): New.
12431 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12433 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12434 check for 128-bit long double before checking TCmode.
12435 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12436 128-bit long doubles before checking TFmode or TCmode.
12437 (FLOAT128_IBM_P): Likewise.
12439 2018-01-10 Martin Sebor <msebor@redhat.com>
12441 PR tree-optimization/83671
12442 * builtins.c (c_strlen): Unconditionally return zero for the empty
12444 Use -Warray-bounds for warnings.
12445 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12446 for non-constant array indices with COMPONENT_REF, arrays of
12447 arrays, and pointers to arrays.
12448 (gimple_fold_builtin_strlen): Determine and set length range for
12449 non-constant character arrays.
12451 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12453 PR middle-end/81897
12454 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12457 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12459 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12461 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12464 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12465 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12466 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12467 indexed_or_indirect_operand predicate.
12468 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12469 (*vsx_le_perm_load_v8hi): Likewise.
12470 (*vsx_le_perm_load_v16qi): Likewise.
12471 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12472 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12473 (*vsx_le_perm_store_v8hi): Likewise.
12474 (*vsx_le_perm_store_v16qi): Likewise.
12475 (eight unnamed splitters): Likewise.
12477 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12479 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12480 * config/rs6000/emmintrin.h: Likewise.
12481 * config/rs6000/mmintrin.h: Likewise.
12482 * config/rs6000/xmmintrin.h: Likewise.
12484 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12487 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12489 * tree.c (tree_nop_conversion): Return true for location wrapper
12491 (maybe_wrap_with_location): New function.
12492 (selftest::check_strip_nops): New function.
12493 (selftest::test_location_wrappers): New function.
12494 (selftest::tree_c_tests): Call it.
12495 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12496 (maybe_wrap_with_location): New decl.
12497 (EXPR_LOCATION_WRAPPER_P): New macro.
12498 (location_wrapper_p): New inline function.
12499 (tree_strip_any_location_wrapper): New inline function.
12501 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12504 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12505 stack_realign_offset for the largest alignment of stack slot
12507 (ix86_find_max_used_stack_alignment): New function.
12508 (ix86_finalize_stack_frame_flags): Use it. Set
12509 max_used_stack_alignment if we don't realign stack.
12510 * config/i386/i386.h (machine_function): Add
12511 max_used_stack_alignment.
12513 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12515 * config/arm/arm.opt (-mbranch-cost): New option.
12516 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12519 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12522 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12523 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12525 2018-01-10 Richard Biener <rguenther@suse.de>
12528 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12529 early out so it also covers the case where we have a non-NULL
12532 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12534 PR tree-optimization/83753
12535 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12536 for non-strided grouped accesses if the number of elements is 1.
12538 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12541 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12542 * i386.h (TARGET_USE_GATHER): Define.
12543 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12545 2018-01-10 Martin Liska <mliska@suse.cz>
12548 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12549 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12551 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12552 CLEANUP_NO_PARTITIONING is not set.
12554 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12556 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12557 for vectors, as a partial revert of r254296.
12558 * rtl.h (const_vec_p): Delete.
12559 (const_vec_duplicate_p): Don't test for vector CONSTs.
12560 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12561 * expmed.c (make_tree): Likewise.
12564 * common.md (E, F): Use CONSTANT_P instead of checking for
12566 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12567 checking for CONST_VECTOR.
12569 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12571 PR middle-end/83575
12572 * predict.c (force_edge_cold): Handle in more sane way edges
12573 with no prediction.
12575 2018-01-09 Carl Love <cel@us.ibm.com>
12577 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12579 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12580 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12581 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12582 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12583 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12584 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12585 * config/rs6000/rs6000-protos.h: Add extern defition for
12586 rs6000_generate_float2_double_code.
12587 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12589 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12590 (float2_v2df): Add define_expand.
12592 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12595 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12596 op_mode in the force_to_mode call.
12598 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12600 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12601 instead of checking each element individually.
12602 (aarch64_evpc_uzp): Likewise.
12603 (aarch64_evpc_zip): Likewise.
12604 (aarch64_evpc_ext): Likewise.
12605 (aarch64_evpc_rev): Likewise.
12606 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12607 instead of checking each element individually. Return true without
12609 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12610 whether all selected elements come from the same input, instead of
12611 checking each element individually. Remove calls to gen_rtx_REG,
12612 start_sequence and end_sequence and instead assert that no rtl is
12615 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12617 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12618 order of HIGH and CONST checks.
12620 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12622 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12623 if the destination isn't an SSA_NAME.
12625 2018-01-09 Richard Biener <rguenther@suse.de>
12627 PR tree-optimization/83668
12628 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12630 (canonicalize_loop_form): ... here, renamed from ...
12631 (canonicalize_loop_closed_ssa_form): ... this and amended to
12632 swap successor edges for loop exit blocks to make us use
12633 the RPO order we need for initial schedule generation.
12635 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12637 PR tree-optimization/64811
12638 * match.pd: When optimizing comparisons with Inf, avoid
12639 introducing or losing exceptions from comparisons with NaN.
12641 2018-01-09 Martin Liska <mliska@suse.cz>
12644 * asan.c (shadow_mem_size): Add gcc_assert.
12646 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12648 Don't save registers in main().
12651 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12652 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12653 * config/avr/avr.c (avr_set_current_function): Don't error if
12654 naked, OS_task or OS_main are specified at the same time.
12655 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12657 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12659 * common/config/avr/avr-common.c (avr_option_optimization_table):
12660 Switch on -mmain-is-OS_task for optimizing compilations.
12662 2018-01-09 Richard Biener <rguenther@suse.de>
12664 PR tree-optimization/83572
12665 * graphite.c: Include cfganal.h.
12666 (graphite_transform_loops): Connect infinite loops to exit
12667 and remove fake edges at the end.
12669 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12671 * ipa-inline.c (edge_badness): Revert accidental checkin.
12673 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12676 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12677 symbols; not inline clones.
12679 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12682 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12683 hard registers. Formatting fixes.
12685 PR preprocessor/83722
12686 * gcc.c (try_generate_repro): Pass
12687 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12688 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12691 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12692 Kito Cheng <kito.cheng@gmail.com>
12694 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12695 (riscv_leaf_function_p): Delete.
12696 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12698 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12700 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12702 (do_ifelse): New function.
12703 (do_isel): New function.
12704 (do_sub3): New function.
12705 (do_add3): New function.
12706 (do_load_mask_compare): New function.
12707 (do_overlap_load_compare): New function.
12708 (expand_compare_loop): New function.
12709 (expand_block_compare): Call expand_compare_loop() when appropriate.
12710 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12711 option description.
12712 (-mblock-compare-inline-loop-limit): New option.
12714 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12717 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12718 Reverse order of second and third operands in first alternative.
12719 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12720 of first and second elements in UNSPEC_VPERMR vector.
12721 (altivec_expand_vec_perm_le): Likewise.
12723 2018-01-08 Jeff Law <law@redhat.com>
12725 PR rtl-optimizatin/81308
12726 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12727 (process_switch): If group_case_labels makes a change, then set
12729 (pass_convert_switch::execute): If a switch is converted, then
12730 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12732 PR rtl-optimization/81308
12733 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12736 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12738 PR target/83663 - Revert r255946
12739 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12740 generation for cases where splatting a value is not useful.
12741 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12742 across a vec_duplicate and a paradoxical subreg forming a vector
12743 mode to a vec_concat.
12745 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12747 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12748 -march=armv8.3-a variants.
12749 * config/arm/t-multilib: Likewise.
12750 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12752 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12754 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12756 (cceq_ior_compare_complement): Give it a name so I can use it, and
12757 change boolean_or_operator predicate to boolean_operator so it can
12758 be used to generate a crand.
12759 (eqne): New code iterator.
12760 (bd/bd_neg): New code_attrs.
12761 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12762 a single define_insn.
12763 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12764 decrement (bdnzt/bdnzf/bdzt/bdzf).
12765 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12766 with the new names of the branch decrement patterns, and added the
12767 names of the branch decrement conditional patterns.
12769 2018-01-08 Richard Biener <rguenther@suse.de>
12771 PR tree-optimization/83563
12772 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12775 2018-01-08 Richard Biener <rguenther@suse.de>
12777 PR middle-end/83713
12778 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12780 2018-01-08 Richard Biener <rguenther@suse.de>
12782 PR tree-optimization/83685
12783 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12784 references to abnormals.
12786 2018-01-08 Richard Biener <rguenther@suse.de>
12789 * dwarf2out.c (output_indirect_strings): Handle empty
12790 skeleton_debug_str_hash.
12791 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12793 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12795 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12796 (emit_store_direct): Likewise.
12797 (arc_trampoline_adjust_address): Likewise.
12798 (arc_asm_trampoline_template): New function.
12799 (arc_initialize_trampoline): Use asm_trampoline_template.
12800 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12801 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12802 * config/arc/arc.md (flush_icache): Delete pattern.
12804 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12806 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12807 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12810 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12813 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12814 by not USED_FOR_TARGET.
12815 (make_pass_resolve_sw_modes): Likewise.
12817 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12819 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12822 2018-01-08 Richard Biener <rguenther@suse.de>
12824 PR middle-end/83580
12825 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12827 2018-01-08 Richard Biener <rguenther@suse.de>
12829 PR middle-end/83517
12830 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12832 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12834 PR middle-end/81897
12835 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12836 basic blocks with a small number of successors.
12837 (convert_control_dep_chain_into_preds): Improve handling of
12839 (dump_predicates): Split apart into...
12840 (dump_pred_chain): ...here...
12841 (dump_pred_info): ...and here.
12842 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12843 (can_chain_union_be_invalidated_p): Improve check for invalidation
12845 (uninit_uses_cannot_happen): Avoid unnecessary if
12846 convert_control_dep_chain_into_preds yielded nothing.
12848 2018-01-06 Martin Sebor <msebor@redhat.com>
12850 PR tree-optimization/83640
12851 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12852 subtracting negative offset from size.
12853 (builtin_access::overlap): Adjust offset bounds of the access to fall
12854 within the size of the object if possible.
12856 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12858 PR rtl-optimization/83699
12859 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12860 extract_bit_field_as_subreg to cases in which the extracted
12861 value is also a vector.
12863 * lra-constraints.c (process_alt_operands): Test for the equivalence
12864 substitutions when detecting a possible reload cycle.
12866 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12869 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12870 by default if flag_selective_schedling{,2}. Formatting fixes.
12872 PR rtl-optimization/83682
12873 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12874 if it has non-VECTOR_MODE element mode.
12875 (vec_duplicate_p): Likewise.
12877 PR middle-end/83694
12878 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12879 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12881 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12884 * config/i386/i386-builtin.def
12885 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12886 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12887 Require also OPTION_MASK_ISA_AVX512F in addition to
12888 OPTION_MASK_ISA_GFNI.
12889 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12890 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12891 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12892 to OPTION_MASK_ISA_GFNI.
12893 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12894 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12895 OPTION_MASK_ISA_AVX512BW.
12896 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12897 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12898 addition to OPTION_MASK_ISA_GFNI.
12899 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12900 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12901 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12902 to OPTION_MASK_ISA_GFNI.
12903 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12904 a requirement for all ISAs rather than any of them with a few
12906 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12908 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12909 bitmasks to be enabled with 3 exceptions, instead of requiring any
12910 enabled ISA with lots of exceptions.
12911 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12912 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12913 Change avx512bw in isa attribute to avx512f.
12914 * config/i386/sgxintrin.h: Add license boilerplate.
12915 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12916 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12917 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12918 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12920 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12921 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12922 temporarily sse2 rather than sse if not enabled already.
12925 * config/i386/sse.md (VI248_VLBW): Rename to ...
12926 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12927 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12928 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12929 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12930 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12931 mode iterator instead of VI248_VLBW.
12933 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12935 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12936 (record_modified): Skip clobbers; add debug output.
12937 (param_change_prob): Use sreal frequencies.
12939 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12941 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12942 punt for user-aligned variables.
12944 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12946 * tree-chrec.c (chrec_contains_symbols): Return true for
12949 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12952 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12953 of (x|y) == x for BICS pattern.
12955 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12957 PR tree-optimization/83605
12958 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12959 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12962 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12964 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12965 * config/epiphany/rtems.h: New file.
12967 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12968 Uros Bizjak <ubizjak@gmail.com>
12971 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12972 QIreg_operand instead of register_operand predicate.
12973 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12974 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12975 comments instead of -fmitigate[-_]rop.
12977 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12980 * cgraphunit.c (symbol_table::compile): Switch to text_section
12981 before calling assembly_start debug hook.
12982 * run-rtl-passes.c (run_rtl_passes): Likewise.
12985 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12987 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12988 range_int_cst_p rather than !symbolic_range_p before calling
12989 extract_range_from_multiplicative_op_1.
12991 2018-01-04 Jeff Law <law@redhat.com>
12993 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12994 redundant test in assertion.
12996 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12998 * doc/rtl.texi: Document machine_mode wrapper classes.
13000 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13002 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13003 using tree_to_uhwi.
13005 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13007 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13008 the VEC_PERM_EXPR fold to fail.
13010 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13013 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13014 to switched_sections.
13016 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13019 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13020 test for d.testing.
13022 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13025 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13026 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13028 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13031 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13032 is BLKmode and bitpos not zero or mode change is needed.
13034 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13037 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13040 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13043 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13044 instead of MULT rtx. Update all corresponding splitters.
13045 (*saddl_se): Ditto.
13046 (*ssub<modesuffix>): Ditto.
13047 (*ssubl_se): Ditto.
13048 (*cmp_sadd_di): Update split patterns.
13049 (*cmp_sadd_si): Ditto.
13050 (*cmp_sadd_sidi): Ditto.
13051 (*cmp_ssub_di): Ditto.
13052 (*cmp_ssub_si): Ditto.
13053 (*cmp_ssub_sidi): Ditto.
13054 * config/alpha/predicates.md (const23_operand): New predicate.
13055 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13056 Look for ASHIFT, not MULT inner operand.
13057 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13059 2018-01-04 Martin Liska <mliska@suse.cz>
13061 PR gcov-profile/83669
13062 * gcov.c (output_intermediate_file): Add version to intermediate
13064 * doc/gcov.texi: Document new field 'version' in intermediate
13065 file format. Fix location of '-k' option of gcov command.
13067 2018-01-04 Martin Liska <mliska@suse.cz>
13070 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13072 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13074 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13076 2018-01-03 Martin Sebor <msebor@redhat.com>
13078 PR tree-optimization/83655
13079 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13080 checking calls with invalid arguments.
13082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13084 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13085 (vectorizable_mask_load_store): Delete.
13086 (vectorizable_call): Return false for masked loads and stores.
13087 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13088 instead of gimple_assign_rhs1.
13089 (vectorizable_load): Handle IFN_MASK_LOAD.
13090 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13094 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13096 (vectorizable_mask_load_store): ...here.
13097 (vectorizable_load): ...and here.
13099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13101 * tree-vect-stmts.c (vect_build_all_ones_mask)
13102 (vect_build_zero_merge_argument): New functions, split out from...
13103 (vectorizable_load): ...here.
13105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13107 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13109 (vectorizable_mask_load_store): ...here.
13110 (vectorizable_store): ...and here.
13112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13114 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13116 (vectorizable_mask_load_store): ...here.
13118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13120 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13121 (vect_model_store_cost): Take a vec_load_store_type instead of a
13123 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13124 (vect_model_store_cost): Take a vec_load_store_type instead of a
13126 (vectorizable_mask_load_store): Update accordingly.
13127 (vectorizable_store): Likewise.
13128 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13132 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13133 IFN_MASK_LOAD calls here rather than...
13134 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13137 Alan Hayward <alan.hayward@arm.com>
13138 David Sherwood <david.sherwood@arm.com>
13140 * expmed.c (extract_bit_field_1): For vector extracts,
13141 fall back to extract_bit_field_as_subreg if vec_extract
13144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13145 Alan Hayward <alan.hayward@arm.com>
13146 David Sherwood <david.sherwood@arm.com>
13148 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13149 they are variable or constant sized.
13150 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13151 slots for constant-sized data.
13153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13154 Alan Hayward <alan.hayward@arm.com>
13155 David Sherwood <david.sherwood@arm.com>
13157 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13158 handling COND_EXPRs with boolean comparisons, try to find a better
13159 basis for the mask type than the boolean itself.
13161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13163 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13164 is calculated and how it can be overridden.
13165 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13166 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13168 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13172 Alan Hayward <alan.hayward@arm.com>
13173 David Sherwood <david.sherwood@arm.com>
13175 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13176 Remove the mode argument.
13177 (aarch64_simd_valid_immediate): Remove the mode and inverse
13179 * config/aarch64/iterators.md (bitsize): New iterator.
13180 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13181 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13182 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13183 aarch64_simd_valid_immediate.
13184 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13185 (aarch64_reg_or_bic_imm): Likewise.
13186 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13187 with an insn_type enum and msl with a modifier_type enum.
13188 Replace element_width with a scalar_mode. Change the shift
13189 to unsigned int. Add constructors for scalar_float_mode and
13190 scalar_int_mode elements.
13191 (aarch64_vect_float_const_representable_p): Delete.
13192 (aarch64_can_const_movi_rtx_p)
13193 (aarch64_simd_scalar_immediate_valid_for_move)
13194 (aarch64_simd_make_constant): Update call to
13195 aarch64_simd_valid_immediate.
13196 (aarch64_advsimd_valid_immediate_hs): New function.
13197 (aarch64_advsimd_valid_immediate): Likewise.
13198 (aarch64_simd_valid_immediate): Remove mode and inverse
13199 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13200 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13201 and aarch64_float_const_representable_p on the result.
13202 (aarch64_output_simd_mov_immediate): Remove mode argument.
13203 Update call to aarch64_simd_valid_immediate and use of
13204 simd_immediate_info.
13205 (aarch64_output_scalar_simd_mov_immediate): Update call
13208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13209 Alan Hayward <alan.hayward@arm.com>
13210 David Sherwood <david.sherwood@arm.com>
13212 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13213 (mode_nunits): Likewise CONST_MODE_NUNITS.
13214 * machmode.def (ADJUST_NUNITS): Document.
13215 * genmodes.c (mode_data::need_nunits_adj): New field.
13216 (blank_mode): Update accordingly.
13217 (adj_nunits): New variable.
13218 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13220 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13221 listed in adj_nunits.
13222 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13223 listed in adj_nunits. Don't emit case statements for such modes.
13224 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13225 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13226 nothing if adj_nunits is nonnull.
13227 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13228 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13229 (emit_mode_fbit): Update use of print_maybe_const_decl.
13230 (emit_move_size): Likewise. Treat the array as non-const
13232 (emit_mode_adjustments): Handle adj_nunits.
13234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13236 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13237 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13238 (VECTOR_MODES): Use it.
13239 (make_vector_modes): Take the prefix as an argument.
13241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13242 Alan Hayward <alan.hayward@arm.com>
13243 David Sherwood <david.sherwood@arm.com>
13245 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13246 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13247 for MODE_VECTOR_BOOL.
13248 * machmode.def (VECTOR_BOOL_MODE): Document.
13249 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13250 (make_vector_bool_mode): New function.
13251 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13253 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13254 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13256 * stor-layout.c (int_mode_for_mode): Likewise.
13257 * tree.c (build_vector_type_for_mode): Likewise.
13258 * varasm.c (output_constant_pool_2): Likewise.
13259 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13260 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13261 for MODE_VECTOR_BOOL.
13262 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13263 of mode class checks.
13264 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13265 instead of a list of mode class checks.
13266 (expand_vector_scalar_condition): Likewise.
13267 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13270 Alan Hayward <alan.hayward@arm.com>
13271 David Sherwood <david.sherwood@arm.com>
13273 * machmode.h (mode_size): Change from unsigned short to
13275 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13276 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13277 or if measurement_type is not polynomial.
13278 (fixed_size_mode::includes_p): Check for constant-sized modes.
13279 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13280 return a poly_uint16 rather than an unsigned short.
13281 (emit_mode_size): Change the type of mode_size from unsigned short
13282 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13283 (emit_mode_adjustments): Cope with polynomial vector sizes.
13284 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13286 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13288 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13289 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13290 * caller-save.c (setup_save_areas): Likewise.
13291 (replace_reg_with_saved_mem): Likewise.
13292 * calls.c (emit_library_call_value_1): Likewise.
13293 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13294 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13295 (gen_lowpart_for_combine): Likewise.
13296 * convert.c (convert_to_integer_1): Likewise.
13297 * cse.c (equiv_constant, cse_insn): Likewise.
13298 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13299 (cselib_subst_to_values): Likewise.
13300 * dce.c (word_dce_process_block): Likewise.
13301 * df-problems.c (df_word_lr_mark_ref): Likewise.
13302 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13303 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13304 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13305 (rtl_for_decl_location): Likewise.
13306 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13307 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13308 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13309 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13310 (expand_expr_real_1): Likewise.
13311 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13312 (pad_below): Likewise.
13313 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13314 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13315 * ira.c (get_subreg_tracking_sizes): Likewise.
13316 * ira-build.c (ira_create_allocno_objects): Likewise.
13317 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13318 (ira_sort_regnos_for_alter_reg): Likewise.
13319 * ira-costs.c (record_operand_costs): Likewise.
13320 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13321 (resolve_simple_move): Likewise.
13322 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13323 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13324 (lra_constraints): Likewise.
13325 (CONST_POOL_OK_P): Reject variable-sized modes.
13326 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13327 (add_pseudo_to_slot, lra_spill): Likewise.
13328 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13329 * optabs-query.c (get_best_extraction_insn): Likewise.
13330 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13331 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13332 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13333 * recog.c (offsettable_address_addr_space_p): Likewise.
13334 * regcprop.c (maybe_mode_change): Likewise.
13335 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13336 * regrename.c (build_def_use): Likewise.
13337 * regstat.c (dump_reg_info): Likewise.
13338 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13339 (find_reloads, find_reloads_subreg_address): Likewise.
13340 * reload1.c (eliminate_regs_1): Likewise.
13341 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13342 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13343 (simplify_binary_operation_1, simplify_subreg): Likewise.
13344 * targhooks.c (default_function_arg_padding): Likewise.
13345 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13346 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13347 (verify_gimple_assign_ternary): Likewise.
13348 * tree-inline.c (estimate_move_cost): Likewise.
13349 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13350 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13351 (get_address_cost_ainc): Likewise.
13352 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13353 (vect_supportable_dr_alignment): Likewise.
13354 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13355 (vectorizable_reduction): Likewise.
13356 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13357 (vectorizable_operation, vectorizable_load): Likewise.
13358 * tree.c (build_same_sized_truth_vector_type): Likewise.
13359 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13360 * var-tracking.c (emit_note_insn_var_location): Likewise.
13361 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13362 (ADDR_VEC_ALIGN): Likewise.
13364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13365 Alan Hayward <alan.hayward@arm.com>
13366 David Sherwood <david.sherwood@arm.com>
13368 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13370 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13371 or if measurement_type is polynomial.
13372 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13373 * combine.c (make_extraction): Likewise.
13374 * dse.c (find_shift_sequence): Likewise.
13375 * dwarf2out.c (mem_loc_descriptor): Likewise.
13376 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13377 (extract_bit_field, extract_low_bits): Likewise.
13378 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13379 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13380 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13381 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13382 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13383 * reload.c (find_reloads): Likewise.
13384 * reload1.c (alter_reg): Likewise.
13385 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13386 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13387 * tree-if-conv.c (predicate_mem_writes): Likewise.
13388 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13389 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13390 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13391 * valtrack.c (dead_debug_insert_temp): Likewise.
13392 * varasm.c (mergeable_constant_section): Likewise.
13393 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13396 Alan Hayward <alan.hayward@arm.com>
13397 David Sherwood <david.sherwood@arm.com>
13399 * expr.c (expand_assignment): Cope with polynomial mode sizes
13400 when assigning to a CONCAT.
13402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13403 Alan Hayward <alan.hayward@arm.com>
13404 David Sherwood <david.sherwood@arm.com>
13406 * machmode.h (mode_precision): Change from unsigned short to
13408 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13410 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13411 or if measurement_type is not polynomial.
13412 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13413 in which the mode is already known to be a scalar_int_mode.
13414 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13415 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13417 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13418 for GET_MODE_PRECISION.
13419 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13420 for GET_MODE_PRECISION.
13421 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13423 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13424 (expand_field_assignment, make_extraction): Likewise.
13425 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13426 (get_last_value): Likewise.
13427 * convert.c (convert_to_integer_1): Likewise.
13428 * cse.c (cse_insn): Likewise.
13429 * expr.c (expand_expr_real_1): Likewise.
13430 * lra-constraints.c (simplify_operand_subreg): Likewise.
13431 * optabs-query.c (can_atomic_load_p): Likewise.
13432 * optabs.c (expand_atomic_load): Likewise.
13433 (expand_atomic_store): Likewise.
13434 * ree.c (combine_reaching_defs): Likewise.
13435 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13436 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13437 * tree.h (type_has_mode_precision_p): Likewise.
13438 * ubsan.c (instrument_si_overflow): Likewise.
13440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13441 Alan Hayward <alan.hayward@arm.com>
13442 David Sherwood <david.sherwood@arm.com>
13444 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13445 polynomial numbers of units.
13446 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13447 (valid_vector_subparts_p): New function.
13448 (build_vector_type): Remove temporary shim and take the number
13449 of units as a poly_uint64 rather than an int.
13450 (build_opaque_vector_type): Take the number of units as a
13451 poly_uint64 rather than an int.
13452 * tree.c (build_vector_from_ctor): Handle polynomial
13453 TYPE_VECTOR_SUBPARTS.
13454 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13455 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13456 (build_vector_from_val): If the number of units is variable,
13457 use build_vec_duplicate_cst for constant operands and
13458 VEC_DUPLICATE_EXPR otherwise.
13459 (make_vector_type): Remove temporary is_constant ().
13460 (build_vector_type, build_opaque_vector_type): Take the number of
13461 units as a poly_uint64 rather than an int.
13462 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13464 * cfgexpand.c (expand_debug_expr): Likewise.
13465 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13466 (store_constructor, expand_expr_real_1): Likewise.
13467 (const_scalar_mask_from_tree): Likewise.
13468 * fold-const-call.c (fold_const_reduction): Likewise.
13469 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13470 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13471 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13472 (fold_relational_const): Likewise.
13473 (native_interpret_vector): Likewise. Change the size from an
13474 int to an unsigned int.
13475 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13476 TYPE_VECTOR_SUBPARTS.
13477 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13478 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13479 duplicating a non-constant operand into a variable-length vector.
13480 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13481 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13482 * ipa-icf.c (sem_variable::equals): Likewise.
13483 * match.pd: Likewise.
13484 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13485 * print-tree.c (print_node): Likewise.
13486 * stor-layout.c (layout_type): Likewise.
13487 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13488 * tree-cfg.c (verify_gimple_comparison): Likewise.
13489 (verify_gimple_assign_binary): Likewise.
13490 (verify_gimple_assign_ternary): Likewise.
13491 (verify_gimple_assign_single): Likewise.
13492 * tree-pretty-print.c (dump_generic_node): Likewise.
13493 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13494 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13495 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13496 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13497 (vect_shift_permute_load_chain): Likewise.
13498 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13499 (expand_vector_condition, optimize_vector_constructor): Likewise.
13500 (lower_vec_perm, get_compute_type): Likewise.
13501 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13502 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13503 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13504 (vect_recog_mask_conversion_pattern): Likewise.
13505 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13506 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13507 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13508 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13509 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13510 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13511 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13512 (supportable_widening_operation): Likewise.
13513 (supportable_narrowing_operation): Likewise.
13514 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13516 * varasm.c (output_constant): Likewise.
13518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13519 Alan Hayward <alan.hayward@arm.com>
13520 David Sherwood <david.sherwood@arm.com>
13522 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13523 so that both the length == 3 and length != 3 cases set up their
13524 own permute vectors. Add comments explaining why we know the
13525 number of elements is constant.
13526 (vect_permute_load_chain): Likewise.
13528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13529 Alan Hayward <alan.hayward@arm.com>
13530 David Sherwood <david.sherwood@arm.com>
13532 * machmode.h (mode_nunits): Change from unsigned char to
13534 (ONLY_FIXED_SIZE_MODES): New macro.
13535 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13536 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13537 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13539 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13540 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13541 or if measurement_type is not polynomial.
13542 * genmodes.c (ZERO_COEFFS): New macro.
13543 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13545 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13546 Use ZERO_COEFFS when emitting initializers.
13547 * data-streamer.h (bp_pack_poly_value): New function.
13548 (bp_unpack_poly_value): Likewise.
13549 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13550 for GET_MODE_NUNITS.
13551 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13552 for GET_MODE_NUNITS.
13553 * tree.c (make_vector_type): Remove temporary shim and make
13554 the real function take the number of units as a poly_uint64
13555 rather than an int.
13556 (build_vector_type_for_mode): Handle polynomial nunits.
13557 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13558 * emit-rtl.c (const_vec_series_p_1): Likewise.
13559 (gen_rtx_CONST_VECTOR): Likewise.
13560 * fold-const.c (test_vec_duplicate_folding): Likewise.
13561 * genrecog.c (validate_pattern): Likewise.
13562 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13563 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13564 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13565 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13566 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13567 * rtlanal.c (subreg_get_info): Likewise.
13568 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13569 (vect_grouped_load_supported): Likewise.
13570 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13571 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13572 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13573 (simplify_const_unary_operation, simplify_binary_operation_1)
13574 (simplify_const_binary_operation, simplify_ternary_operation)
13575 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13576 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13577 instead of CONST_VECTOR_NUNITS.
13578 * varasm.c (output_constant_pool_2): Likewise.
13579 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13580 explicit-encoded elements in the XVEC for variable-length vectors.
13582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13584 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13587 Alan Hayward <alan.hayward@arm.com>
13588 David Sherwood <david.sherwood@arm.com>
13590 * coretypes.h (fixed_size_mode): Declare.
13591 (fixed_size_mode_pod): New typedef.
13592 * builtins.h (target_builtins::x_apply_args_mode)
13593 (target_builtins::x_apply_result_mode): Change type to
13594 fixed_size_mode_pod.
13595 * builtins.c (apply_args_size, apply_result_size, result_vector)
13596 (expand_builtin_apply_args_1, expand_builtin_apply)
13597 (expand_builtin_return): Update accordingly.
13599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13601 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13602 * cselib.c (cselib_hash_rtx): Likewise.
13603 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13604 CONST_VECTOR encoding.
13606 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13607 Jeff Law <law@redhat.com>
13610 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13611 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13612 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13613 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13616 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13617 explicitly probe *sp in a noreturn function if there were any callee
13618 register saves or frame pointer is needed.
13620 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13623 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13624 BLKmode for ternary, binary or unary expressions.
13627 * var-tracking.c (delete_vta_debug_insn): New inline function.
13628 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13629 insns from get_insns () to NULL instead of each bb separately.
13630 Use delete_vta_debug_insn. No longer static.
13631 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13632 delete_vta_debug_insns callers.
13633 * rtl.h (delete_vta_debug_insns): Declare.
13634 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13635 instead of variable_tracking_main.
13637 2018-01-03 Martin Sebor <msebor@redhat.com>
13639 PR tree-optimization/83603
13640 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13641 arguments past the endof the argument list in functions declared
13642 without a prototype.
13643 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13644 Avoid checking when arguments are null.
13646 2018-01-03 Martin Sebor <msebor@redhat.com>
13649 * doc/extend.texi (attribute const): Fix a typo.
13650 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13651 issuing -Wsuggest-attribute for void functions.
13653 2018-01-03 Martin Sebor <msebor@redhat.com>
13655 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13656 offset_int::from instead of wide_int::to_shwi.
13657 (maybe_diag_overlap): Remove assertion.
13658 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13659 * gimple-ssa-sprintf.c (format_directive): Same.
13660 (parse_directive): Same.
13661 (sprintf_dom_walker::compute_format_length): Same.
13662 (try_substitute_return_value): Same.
13664 2018-01-03 Jeff Law <law@redhat.com>
13666 PR middle-end/83654
13667 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13668 non-constant residual for zero at runtime and avoid probing in
13669 that case. Reorganize code for trailing problem to mirror handling
13672 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13674 PR tree-optimization/83501
13675 * tree-ssa-strlen.c (get_string_cst): New.
13676 (handle_char_store): Call get_string_cst.
13678 2018-01-03 Martin Liska <mliska@suse.cz>
13680 PR tree-optimization/83593
13681 * tree-ssa-strlen.c: Include tree-cfg.h.
13682 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13683 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13684 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13686 (strlen_dom_walker::before_dom_children): Call
13687 gimple_purge_dead_eh_edges. Dump tranformation with details
13689 (strlen_dom_walker::before_dom_children): Update call by adding
13690 new argument cleanup_eh.
13691 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13693 2018-01-03 Martin Liska <mliska@suse.cz>
13696 * cif-code.def (VARIADIC_THUNK): New enum value.
13697 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13700 2018-01-03 Jan Beulich <jbeulich@suse.com>
13702 * sse.md (mov<mode>_internal): Tighten condition for when to use
13703 vmovdqu<ssescalarsize> for TI and OI modes.
13705 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13707 Update copyright years.
13709 2018-01-03 Martin Liska <mliska@suse.cz>
13712 * ipa-visibility.c (function_and_variable_visibility): Skip
13713 functions with noipa attribure.
13715 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13717 * gcc.c (process_command): Update copyright notice dates.
13718 * gcov-dump.c (print_version): Ditto.
13719 * gcov.c (print_version): Ditto.
13720 * gcov-tool.c (print_version): Ditto.
13721 * gengtype.c (create_file): Ditto.
13722 * doc/cpp.texi: Bump @copying's copyright year.
13723 * doc/cppinternals.texi: Ditto.
13724 * doc/gcc.texi: Ditto.
13725 * doc/gccint.texi: Ditto.
13726 * doc/gcov.texi: Ditto.
13727 * doc/install.texi: Ditto.
13728 * doc/invoke.texi: Ditto.
13730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13732 * vector-builder.h (vector_builder::m_full_nelts): Change from
13733 unsigned int to poly_uint64.
13734 (vector_builder::full_nelts): Update prototype accordingly.
13735 (vector_builder::new_vector): Likewise.
13736 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13737 (vector_builder::operator ==): Likewise.
13738 (vector_builder::finalize): Likewise.
13739 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13740 Take the number of elements as a poly_uint64 rather than an
13742 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13743 from unsigned int to poly_uint64.
13744 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13745 (vec_perm_indices::new_vector): Likewise.
13746 (vec_perm_indices::length): Likewise.
13747 (vec_perm_indices::nelts_per_input): Likewise.
13748 (vec_perm_indices::input_nelts): Likewise.
13749 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13750 number of elements per input as a poly_uint64 rather than an
13751 unsigned int. Use the original encoding for variable-length
13752 vectors, rather than clamping each individual element.
13753 For the second and subsequent elements in each pattern,
13754 clamp the step and base before clamping their sum.
13755 (vec_perm_indices::series_p): Handle polynomial element counts.
13756 (vec_perm_indices::all_in_range_p): Likewise.
13757 (vec_perm_indices_to_tree): Likewise.
13758 (vec_perm_indices_to_rtx): Likewise.
13759 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13760 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13761 (tree_vector_builder::new_binary_operation): Handle polynomial
13762 element counts. Return false if we need to know the number
13763 of elements at compile time.
13764 * fold-const.c (fold_vec_perm): Punt if the number of elements
13765 isn't known at compile time.
13767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13769 * vec-perm-indices.h (vec_perm_builder): Change element type
13770 from HOST_WIDE_INT to poly_int64.
13771 (vec_perm_indices::element_type): Update accordingly.
13772 (vec_perm_indices::clamp): Handle polynomial element_types.
13773 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13774 (vec_perm_indices::all_in_range_p): Likewise.
13775 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13777 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13778 polynomial vec_perm_indices element types.
13779 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13780 * fold-const.c (fold_vec_perm): Likewise.
13781 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13782 * tree-vect-generic.c (lower_vec_perm): Likewise.
13783 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13784 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13785 element type to HOST_WIDE_INT.
13787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13788 Alan Hayward <alan.hayward@arm.com>
13789 David Sherwood <david.sherwood@arm.com>
13791 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13792 rather than an int. Use plus_constant.
13793 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13794 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13801 a HOST_WIDE_INT to a poly_int64.
13803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13804 Alan Hayward <alan.hayward@arm.com>
13805 David Sherwood <david.sherwood@arm.com>
13807 * calls.c (load_register_parameters): Cope with polynomial
13808 mode sizes. Require a constant size for BLKmode parameters
13809 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13810 forces a parameter to be padded at the lsb end in order to
13811 fill a complete number of words, require the parameter size
13812 to be ordered wrt UNITS_PER_WORD.
13814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13815 Alan Hayward <alan.hayward@arm.com>
13816 David Sherwood <david.sherwood@arm.com>
13818 * reload1.c (spill_stack_slot_width): Change element type
13819 from unsigned int to poly_uint64_pod.
13820 (alter_reg): Treat mode sizes as polynomial.
13822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13823 Alan Hayward <alan.hayward@arm.com>
13824 David Sherwood <david.sherwood@arm.com>
13826 * reload.c (complex_word_subreg_p): New function.
13827 (reload_inner_reg_of_subreg, push_reload): Use it.
13829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13830 Alan Hayward <alan.hayward@arm.com>
13831 David Sherwood <david.sherwood@arm.com>
13833 * lra-constraints.c (process_alt_operands): Reject matched
13834 operands whose sizes aren't ordered.
13835 (match_reload): Refer to this check here.
13837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13838 Alan Hayward <alan.hayward@arm.com>
13839 David Sherwood <david.sherwood@arm.com>
13841 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13842 that the mode size is in the set {1, 2, 4, 8, 16}.
13844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13845 Alan Hayward <alan.hayward@arm.com>
13846 David Sherwood <david.sherwood@arm.com>
13848 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13849 Use plus_constant instead of gen_rtx_PLUS.
13851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13852 Alan Hayward <alan.hayward@arm.com>
13853 David Sherwood <david.sherwood@arm.com>
13855 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13856 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13857 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13858 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13859 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13860 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13861 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13862 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13863 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13864 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13866 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13867 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13868 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13869 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13870 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13871 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13872 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13873 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13874 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13875 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13877 * expr.c (emit_move_resolve_push): Treat the input and result
13878 of PUSH_ROUNDING as a poly_int64.
13879 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13880 (emit_push_insn): Likewise.
13881 * lra-eliminations.c (mark_not_eliminable): Likewise.
13882 * recog.c (push_operand): Likewise.
13883 * reload1.c (elimination_effects): Likewise.
13884 * rtlanal.c (nonzero_bits1): Likewise.
13885 * calls.c (store_one_arg): Likewise. Require the padding to be
13886 known at compile time.
13888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13889 Alan Hayward <alan.hayward@arm.com>
13890 David Sherwood <david.sherwood@arm.com>
13892 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13893 Use plus_constant instead of gen_rtx_PLUS.
13895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13896 Alan Hayward <alan.hayward@arm.com>
13897 David Sherwood <david.sherwood@arm.com>
13899 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13900 rather than an int.
13902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13903 Alan Hayward <alan.hayward@arm.com>
13904 David Sherwood <david.sherwood@arm.com>
13906 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13907 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13908 via stack temporaries. Treat the mode size as polynomial too.
13910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13911 Alan Hayward <alan.hayward@arm.com>
13912 David Sherwood <david.sherwood@arm.com>
13914 * expr.c (expand_expr_real_2): When handling conversions involving
13915 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13916 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13917 as a poly_uint64 too.
13919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13920 Alan Hayward <alan.hayward@arm.com>
13921 David Sherwood <david.sherwood@arm.com>
13923 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13926 Alan Hayward <alan.hayward@arm.com>
13927 David Sherwood <david.sherwood@arm.com>
13929 * combine.c (can_change_dest_mode): Handle polynomial
13930 REGMODE_NATURAL_SIZE.
13931 * expmed.c (store_bit_field_1): Likewise.
13932 * expr.c (store_constructor): Likewise.
13933 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13934 and polynomial REGMODE_NATURAL_SIZE.
13935 (gen_lowpart_common): Likewise.
13936 * reginfo.c (record_subregs_of_mode): Likewise.
13937 * rtlanal.c (read_modify_subreg_p): Likewise.
13939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13940 Alan Hayward <alan.hayward@arm.com>
13941 David Sherwood <david.sherwood@arm.com>
13943 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13944 numbers of elements.
13946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13947 Alan Hayward <alan.hayward@arm.com>
13948 David Sherwood <david.sherwood@arm.com>
13950 * match.pd: Cope with polynomial numbers of vector elements.
13952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13953 Alan Hayward <alan.hayward@arm.com>
13954 David Sherwood <david.sherwood@arm.com>
13956 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13957 in a POINTER_PLUS_EXPR.
13959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13960 Alan Hayward <alan.hayward@arm.com>
13961 David Sherwood <david.sherwood@arm.com>
13963 * omp-simd-clone.c (simd_clone_subparts): New function.
13964 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13965 (ipa_simd_modify_function_body): Likewise.
13967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13968 Alan Hayward <alan.hayward@arm.com>
13969 David Sherwood <david.sherwood@arm.com>
13971 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13972 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13973 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13974 (expand_vector_condition, vector_element): Likewise.
13975 (subparts_gt): New function.
13976 (get_compute_type): Use subparts_gt.
13977 (count_type_subparts): Delete.
13978 (expand_vector_operations_1): Use subparts_gt instead of
13979 count_type_subparts.
13981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13982 Alan Hayward <alan.hayward@arm.com>
13983 David Sherwood <david.sherwood@arm.com>
13985 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13986 (vect_compile_time_alias): ...this new function. Do the calculation
13987 on poly_ints rather than trees.
13988 (vect_prune_runtime_alias_test_list): Update call accordingly.
13990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13991 Alan Hayward <alan.hayward@arm.com>
13992 David Sherwood <david.sherwood@arm.com>
13994 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13996 (vect_schedule_slp_instance): Likewise.
13998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13999 Alan Hayward <alan.hayward@arm.com>
14000 David Sherwood <david.sherwood@arm.com>
14002 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14003 constant and extern definitions for variable-length vectors.
14004 (vect_get_constant_vectors): Note that the number of units
14005 is known to be constant.
14007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14008 Alan Hayward <alan.hayward@arm.com>
14009 David Sherwood <david.sherwood@arm.com>
14011 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14012 of units as polynomial. Choose between WIDE and NARROW based
14015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14016 Alan Hayward <alan.hayward@arm.com>
14017 David Sherwood <david.sherwood@arm.com>
14019 * tree-vect-stmts.c (simd_clone_subparts): New function.
14020 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14023 Alan Hayward <alan.hayward@arm.com>
14024 David Sherwood <david.sherwood@arm.com>
14026 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14027 vectors as polynomial. Use build_index_vector for
14028 IFN_GOMP_SIMD_LANE.
14030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14031 Alan Hayward <alan.hayward@arm.com>
14032 David Sherwood <david.sherwood@arm.com>
14034 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14035 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14036 for variable-length vectors.
14037 (vectorizable_mask_load_store): Treat the number of units as
14038 polynomial, asserting that it is constant if the condition has
14039 already been enforced.
14040 (vectorizable_store, vectorizable_load): Likewise.
14042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14043 Alan Hayward <alan.hayward@arm.com>
14044 David Sherwood <david.sherwood@arm.com>
14046 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14047 of units as polynomial. Punt if we can't tell at compile time
14048 which vector contains the final result.
14050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14051 Alan Hayward <alan.hayward@arm.com>
14052 David Sherwood <david.sherwood@arm.com>
14054 * tree-vect-loop.c (vectorizable_induction): Treat the number
14055 of units as polynomial. Punt on SLP inductions. Use an integer
14056 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14057 cast of such a series for variable-length floating-point
14060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14061 Alan Hayward <alan.hayward@arm.com>
14062 David Sherwood <david.sherwood@arm.com>
14064 * tree.h (build_index_vector): Declare.
14065 * tree.c (build_index_vector): New function.
14066 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14067 of units as polynomial, forcibly converting it to a constant if
14068 vectorizable_reduction has already enforced the condition.
14069 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14070 to create a {1,2,3,...} vector.
14071 (vectorizable_reduction): Treat the number of units as polynomial.
14072 Choose vectype_in based on the largest scalar element size rather
14073 than the smallest number of units. Enforce the restrictions
14076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14077 Alan Hayward <alan.hayward@arm.com>
14078 David Sherwood <david.sherwood@arm.com>
14080 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14081 number of units as polynomial.
14083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14084 Alan Hayward <alan.hayward@arm.com>
14085 David Sherwood <david.sherwood@arm.com>
14087 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14088 * target.def (autovectorize_vector_sizes): Return the vector sizes
14089 by pointer, using vector_sizes rather than a bitmask.
14090 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14091 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14092 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14094 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14095 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14096 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14097 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14098 * omp-general.c (omp_max_vf): Likewise.
14099 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14100 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14101 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14102 * tree-vect-slp.c (vect_slp_bb): Likewise.
14103 * doc/tm.texi: Regenerate.
14104 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14106 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14107 the vector size as a poly_uint64 rather than an unsigned int.
14108 (current_vector_size): Change from an unsigned int to a poly_uint64.
14109 (get_vectype_for_scalar_type): Update accordingly.
14110 * tree.h (build_truth_vector_type): Take the size and number of
14111 units as a poly_uint64 rather than an unsigned int.
14112 (build_vector_type): Add a temporary overload that takes
14113 the number of units as a poly_uint64 rather than an unsigned int.
14114 * tree.c (make_vector_type): Likewise.
14115 (build_truth_vector_type): Take the number of units as a poly_uint64
14116 rather than an unsigned int.
14118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14119 Alan Hayward <alan.hayward@arm.com>
14120 David Sherwood <david.sherwood@arm.com>
14122 * target.def (get_mask_mode): Take the number of units and length
14123 as poly_uint64s rather than unsigned ints.
14124 * targhooks.h (default_get_mask_mode): Update accordingly.
14125 * targhooks.c (default_get_mask_mode): Likewise.
14126 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14127 * doc/tm.texi: Regenerate.
14129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14130 Alan Hayward <alan.hayward@arm.com>
14131 David Sherwood <david.sherwood@arm.com>
14133 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14134 * omp-general.c (omp_max_vf): Likewise.
14135 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14136 (expand_omp_simd): Handle polynomial safelen.
14137 * omp-low.c (omplow_simd_context): Add a default constructor.
14138 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14139 (lower_rec_simd_input_clauses): Update accordingly.
14140 (lower_rec_input_clauses): Likewise.
14142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14143 Alan Hayward <alan.hayward@arm.com>
14144 David Sherwood <david.sherwood@arm.com>
14146 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14147 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14148 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14149 (vect_analyze_slp_cost): Likewise.
14150 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14151 (vect_model_load_cost): Likewise.
14153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14154 Alan Hayward <alan.hayward@arm.com>
14155 David Sherwood <david.sherwood@arm.com>
14157 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14158 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14159 from an unsigned int * to a poly_uint64_pod *.
14160 (calculate_unrolling_factor): New function.
14161 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14164 Alan Hayward <alan.hayward@arm.com>
14165 David Sherwood <david.sherwood@arm.com>
14167 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14168 from an unsigned int to a poly_uint64.
14169 (_loop_vec_info::slp_unrolling_factor): Likewise.
14170 (_loop_vec_info::vectorization_factor): Change from an int
14172 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14173 (vect_get_num_vectors): New function.
14174 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14175 (vect_get_num_copies): Use vect_get_num_vectors.
14176 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14177 to an unsigned int *.
14178 (vect_analyze_data_refs): Change min_vf from an int * to a
14180 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14181 than an unsigned HOST_WIDE_INT.
14182 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14183 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14184 to an unsigned int *.
14185 (vect_analyze_data_ref_dependences): Likewise.
14186 (vect_compute_data_ref_alignment): Handle polynomial vf.
14187 (vect_enhance_data_refs_alignment): Likewise.
14188 (vect_prune_runtime_alias_test_list): Likewise.
14189 (vect_shift_permute_load_chain): Likewise.
14190 (vect_supportable_dr_alignment): Likewise.
14191 (dependence_distance_ge_vf): Take the vectorization factor as a
14192 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14193 (vect_analyze_data_refs): Change min_vf from an int * to a
14195 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14196 vfm1 as a poly_uint64 rather than an int. Make the same change
14197 for the returned bound_scalar.
14198 (vect_gen_vector_loop_niters): Handle polynomial vf.
14199 (vect_do_peeling): Likewise. Update call to
14200 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14201 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14203 * tree-vect-loop.c (vect_determine_vectorization_factor)
14204 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14205 (vect_get_known_peeling_cost): Likewise.
14206 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14207 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14208 (vect_transform_loop): Likewise. Use the lowest possible VF when
14209 updating the upper bounds of the loop.
14210 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14211 rather than an int.
14212 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14213 polynomial unroll factors.
14214 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14215 (vect_make_slp_decision): Likewise.
14216 (vect_supported_load_permutation_p): Likewise, and polynomial
14218 (vect_analyze_slp_cost): Handle polynomial vf.
14219 (vect_slp_analyze_node_operations): Likewise.
14220 (vect_slp_analyze_bb_1): Likewise.
14221 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14222 than an unsigned HOST_WIDE_INT.
14223 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14224 (vectorizable_load): Handle polynomial vf.
14225 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14227 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14230 Alan Hayward <alan.hayward@arm.com>
14231 David Sherwood <david.sherwood@arm.com>
14233 * match.pd: Handle bit operations involving three constants
14234 and try to fold one pair.
14236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14238 * tree-vect-loop-manip.c: Include gimple-fold.h.
14239 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14240 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14241 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14242 Add a path that uses a step of VF instead of 1, but disable it
14244 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14245 and niters_no_overflow parameters. Update calls to
14246 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14247 Create a new SSA name if the latter choses to use a ste other
14248 than zero, and return it via niters_vector_mult_vf_var.
14249 * tree-vect-loop.c (vect_transform_loop): Update calls to
14250 vect_do_peeling, vect_gen_vector_loop_niters and
14251 slpeel_make_loop_iterate_ntimes.
14252 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14253 (vect_gen_vector_loop_niters): Update declarations after above changes.
14255 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14257 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14258 128-bit round to integer instructions.
14259 (ceil<mode>2): Likewise.
14260 (btrunc<mode>2): Likewise.
14261 (round<mode>2): Likewise.
14263 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14265 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14266 unaligned VSX load/store on P8/P9.
14267 (expand_block_clear): Allow the use of unaligned VSX
14268 load/store on P8/P9.
14270 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14272 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14274 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14275 swap associated with both a load and a store.
14277 2018-01-02 Andrew Waterman <andrew@sifive.com>
14279 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14280 * config/riscv/riscv.md (clear_cache): Use it.
14282 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14284 * web.c: Remove out-of-date comment.
14286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14288 * expr.c (fixup_args_size_notes): Check that any existing
14289 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14290 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14291 (emit_single_push_insn): ...here.
14293 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14295 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14296 (const_vector_encoded_nelts): New function.
14297 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14298 (const_vector_int_elt, const_vector_elt): Declare.
14299 * emit-rtl.c (const_vector_int_elt_1): New function.
14300 (const_vector_elt): Likewise.
14301 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14302 of CONST_VECTOR_ELT.
14304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14306 * expr.c: Include rtx-vector-builder.h.
14307 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14308 directly on the tree encoding.
14309 (const_vector_from_tree): Likewise.
14310 * optabs.c: Include rtx-vector-builder.h.
14311 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14312 sequence of "u" values.
14313 * vec-perm-indices.c: Include rtx-vector-builder.h.
14314 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14315 directly on the vec_perm_indices encoding.
14317 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14319 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14320 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14321 * rtx-vector-builder.h: New file.
14322 * rtx-vector-builder.c: Likewise.
14323 * rtl.h (rtx_def::u2): Add a const_vector field.
14324 (CONST_VECTOR_NPATTERNS): New macro.
14325 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14326 (CONST_VECTOR_DUPLICATE_P): Likewise.
14327 (CONST_VECTOR_STEPPED_P): Likewise.
14328 (CONST_VECTOR_ENCODED_ELT): Likewise.
14329 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14330 (unwrap_const_vec_duplicate): Likewise.
14331 (const_vec_series_p): Check for a non-duplicated vector encoding.
14332 Say that the function only returns true for integer vectors.
14333 * emit-rtl.c: Include rtx-vector-builder.h.
14334 (gen_const_vec_duplicate_1): Delete.
14335 (gen_const_vector): Call gen_const_vec_duplicate instead of
14336 gen_const_vec_duplicate_1.
14337 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14338 (gen_const_vec_duplicate): Use rtx_vector_builder.
14339 (gen_const_vec_series): Likewise.
14340 (gen_rtx_CONST_VECTOR): Likewise.
14341 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14342 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14343 Build a new vector rather than modifying a CONST_VECTOR in-place.
14344 (handle_special_swappables): Update call accordingly.
14345 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14346 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14347 Build a new vector rather than modifying a CONST_VECTOR in-place.
14348 (handle_special_swappables): Update call accordingly.
14350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14352 * simplify-rtx.c (simplify_const_binary_operation): Use
14353 CONST_VECTOR_ELT instead of XVECEXP.
14355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14357 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14358 the selector elements to be different from the data elements
14359 if the selector is a VECTOR_CST.
14360 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14361 ssizetype for the selector.
14363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14365 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14366 before testing each element individually.
14367 * tree-vect-generic.c (lower_vec_perm): Likewise.
14369 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14371 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14372 * selftest-run-tests.c (selftest::run_tests): Call it.
14373 * vector-builder.h (vector_builder::operator ==): New function.
14374 (vector_builder::operator !=): Likewise.
14375 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14376 (vec_perm_indices::all_from_input_p): New function.
14377 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14378 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14379 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14380 instead of reading the VECTOR_CST directly. Detect whether both
14381 vector inputs are the same before constructing the vec_perm_indices,
14382 and update the number of inputs argument accordingly. Use the
14383 utility functions added above. Only construct sel2 if we need to.
14385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14387 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14388 the broadcast of the low byte.
14389 (expand_mult_highpart): Use an explicit encoding for the permutes.
14390 * optabs-query.c (can_mult_highpart_p): Likewise.
14391 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14392 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14393 (vectorizable_bswap): Likewise.
14394 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14395 explicit encoding for the power-of-2 permutes.
14396 (vect_permute_store_chain): Likewise.
14397 (vect_grouped_load_supported): Likewise.
14398 (vect_permute_load_chain): Likewise.
14400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14402 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14403 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14404 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14405 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14406 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14407 (vect_gen_perm_mask_any): Likewise.
14409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14411 * int-vector-builder.h: New file.
14412 * vec-perm-indices.h: Include int-vector-builder.h.
14413 (vec_perm_indices): Redefine as an int_vector_builder.
14414 (auto_vec_perm_indices): Delete.
14415 (vec_perm_builder): Redefine as a stand-alone class.
14416 (vec_perm_indices::vec_perm_indices): New function.
14417 (vec_perm_indices::clamp): Likewise.
14418 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14419 (vec_perm_indices::new_vector): New function.
14420 (vec_perm_indices::new_expanded_vector): Update for new
14421 vec_perm_indices class.
14422 (vec_perm_indices::rotate_inputs): New function.
14423 (vec_perm_indices::all_in_range_p): Operate directly on the
14424 encoded form, without computing elided elements.
14425 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14426 encoding. Update for new vec_perm_indices class.
14427 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14428 the given vec_perm_builder.
14429 (expand_vec_perm_var): Update vec_perm_builder constructor.
14430 (expand_mult_highpart): Use vec_perm_builder instead of
14431 auto_vec_perm_indices.
14432 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14433 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14434 or double series encoding as appropriate.
14435 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14436 vec_perm_indices instead of auto_vec_perm_indices.
14437 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14438 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14439 (vect_permute_store_chain): Likewise.
14440 (vect_grouped_load_supported): Likewise.
14441 (vect_permute_load_chain): Likewise.
14442 (vect_shift_permute_load_chain): Likewise.
14443 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14444 (vect_transform_slp_perm_load): Likewise.
14445 (vect_schedule_slp_instance): Likewise.
14446 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14447 (vectorizable_mask_load_store): Likewise.
14448 (vectorizable_bswap): Likewise.
14449 (vectorizable_store): Likewise.
14450 (vectorizable_load): Likewise.
14451 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14452 vec_perm_indices instead of auto_vec_perm_indices. Use
14453 tree_to_vec_perm_builder to read the vector from a tree.
14454 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14455 vec_perm_builder instead of a vec_perm_indices.
14456 (have_whole_vector_shift): Use vec_perm_builder and
14457 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14458 truncation to calc_vec_perm_mask_for_shift.
14459 (vect_create_epilog_for_reduction): Likewise.
14460 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14461 from auto_vec_perm_indices to vec_perm_indices.
14462 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14463 instead of changing individual elements.
14464 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14465 the vector in d.perm.
14466 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14467 from auto_vec_perm_indices to vec_perm_indices.
14468 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14469 instead of changing individual elements.
14470 (arm_vectorize_vec_perm_const): Use new_vector to install
14471 the vector in d.perm.
14472 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14473 Update vec_perm_builder constructor.
14474 (rs6000_expand_interleave): Likewise.
14475 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14476 (rs6000_expand_interleave): Likewise.
14478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14480 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14481 to qimode could truncate the indices.
14482 * optabs.c (expand_vec_perm_var): Likewise.
14484 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14486 * Makefile.in (OBJS): Add vec-perm-indices.o.
14487 * vec-perm-indices.h: New file.
14488 * vec-perm-indices.c: Likewise.
14489 * target.h (vec_perm_indices): Replace with a forward class
14491 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14492 * optabs.h: Include vec-perm-indices.h.
14493 (expand_vec_perm): Delete.
14494 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14495 (expand_vec_perm_const): Declare.
14496 * target.def (vec_perm_const_ok): Replace with...
14497 (vec_perm_const): ...this new hook.
14498 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14499 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14500 * doc/tm.texi: Regenerate.
14501 * optabs.def (vec_perm_const): Delete.
14502 * doc/md.texi (vec_perm_const): Likewise.
14503 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14504 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14505 expand_vec_perm for constant permutation vectors. Assert that
14506 the mode of variable permutation vectors is the integer equivalent
14507 of the mode that is being permuted.
14508 * optabs-query.h (selector_fits_mode_p): Declare.
14509 * optabs-query.c: Include vec-perm-indices.h.
14510 (selector_fits_mode_p): New function.
14511 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14512 is defined, instead of checking whether the vec_perm_const_optab
14513 exists. Use targetm.vectorize.vec_perm_const instead of
14514 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14515 fit in the vector mode before using a variable permute.
14516 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14517 vec_perm_indices instead of an rtx.
14518 (expand_vec_perm): Replace with...
14519 (expand_vec_perm_const): ...this new function. Take the selector
14520 as a vec_perm_indices rather than an rtx. Also take the mode of
14521 the selector. Update call to shift_amt_for_vec_perm_mask.
14522 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14523 Use vec_perm_indices::new_expanded_vector to expand the original
14524 selector into bytes. Check whether the indices fit in the vector
14525 mode before using a variable permute.
14526 (expand_vec_perm_var): Make global.
14527 (expand_mult_highpart): Use expand_vec_perm_const.
14528 * fold-const.c: Includes vec-perm-indices.h.
14529 * tree-ssa-forwprop.c: Likewise.
14530 * tree-vect-data-refs.c: Likewise.
14531 * tree-vect-generic.c: Likewise.
14532 * tree-vect-loop.c: Likewise.
14533 * tree-vect-slp.c: Likewise.
14534 * tree-vect-stmts.c: Likewise.
14535 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14537 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14538 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14539 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14540 (aarch64_vectorize_vec_perm_const): ...this new function.
14541 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14542 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14543 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14544 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14545 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14546 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14547 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14549 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14550 check for NEON modes.
14551 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14552 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14553 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14554 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14556 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14557 the old VEC_PERM_CONST conditions.
14558 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14559 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14560 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14561 (ia64_vectorize_vec_perm_const_ok): Merge into...
14562 (ia64_vectorize_vec_perm_const): ...this new function.
14563 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14564 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14565 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14566 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14567 * config/mips/mips.c (mips_expand_vec_perm_const)
14568 (mips_vectorize_vec_perm_const_ok): Merge into...
14569 (mips_vectorize_vec_perm_const): ...this new function.
14570 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14571 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14572 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14573 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14574 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14575 (rs6000_expand_vec_perm_const): Delete.
14576 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14578 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14579 (altivec_expand_vec_perm_const_le): Take each operand individually.
14580 Operate on constant selectors rather than rtxes.
14581 (altivec_expand_vec_perm_const): Likewise. Update call to
14582 altivec_expand_vec_perm_const_le.
14583 (rs6000_expand_vec_perm_const): Delete.
14584 (rs6000_vectorize_vec_perm_const_ok): Delete.
14585 (rs6000_vectorize_vec_perm_const): New function.
14586 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14587 an element count and rtx array.
14588 (rs6000_expand_extract_even): Update call accordingly.
14589 (rs6000_expand_interleave): Likewise.
14590 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14591 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14592 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14593 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14594 (rs6000_expand_vec_perm_const): Delete.
14595 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14596 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14597 (altivec_expand_vec_perm_const_le): Take each operand individually.
14598 Operate on constant selectors rather than rtxes.
14599 (altivec_expand_vec_perm_const): Likewise. Update call to
14600 altivec_expand_vec_perm_const_le.
14601 (rs6000_expand_vec_perm_const): Delete.
14602 (rs6000_vectorize_vec_perm_const_ok): Delete.
14603 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14604 reference to the SPE evmerge intructions.
14605 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14606 an element count and rtx array.
14607 (rs6000_expand_extract_even): Update call accordingly.
14608 (rs6000_expand_interleave): Likewise.
14609 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14610 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14612 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14616 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14617 vector mode and that that mode matches the mode of the data
14619 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14620 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14621 directly using expand_vec_perm_1 when forcing selectors into
14623 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14627 * optabs-query.h (can_vec_perm_p): Delete.
14628 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14629 * optabs-query.c (can_vec_perm_p): Split into...
14630 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14631 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14632 particular selector is valid.
14633 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14634 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14635 (vect_grouped_load_supported): Likewise.
14636 (vect_shift_permute_load_chain): Likewise.
14637 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14638 (vect_transform_slp_perm_load): Likewise.
14639 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14640 (vectorizable_bswap): Likewise.
14641 (vect_gen_perm_mask_checked): Likewise.
14642 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14643 implementations of variable permutation vectors into account
14644 when deciding which selector to use.
14645 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14646 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14647 with a false third argument.
14648 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14649 to test whether the constant selector is valid and can_vec_perm_var_p
14650 to test whether a variable selector is valid.
14652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14654 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14655 * optabs-query.c (can_vec_perm_p): Likewise.
14656 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14657 instead of vec_perm_indices.
14658 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14659 (vect_gen_perm_mask_checked): Likewise,
14660 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14661 (vect_gen_perm_mask_checked): Likewise,
14663 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14665 * optabs-query.h (qimode_for_vec_perm): Declare.
14666 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14667 (qimode_for_vec_perm): ...this new function.
14668 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14670 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14672 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14673 does not have a conditional at the top.
14675 2018-01-02 Richard Biener <rguenther@suse.de>
14677 * ipa-inline.c (big_speedup_p): Fix expression.
14679 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14682 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14685 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14689 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14690 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14691 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14692 cond_taken_branch_cost 3->4.
14694 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14696 PR tree-optimization/83581
14697 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14698 TODO_cleanup_cfg if any changes have been made.
14700 PR middle-end/83608
14701 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14702 convert_modes if target mode has the right side, but different mode
14705 PR middle-end/83609
14706 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14707 last argument when extracting from CONCAT. If either from_real or
14708 from_imag is NULL, use expansion through memory. If result is not
14709 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14710 the parts directly to inner mode, if even that fails, use expansion
14713 PR middle-end/83623
14714 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14715 check for bswap in mode rather than HImode and use that in expand_unop
14718 Copyright (C) 2018 Free Software Foundation, Inc.
14720 Copying and distribution of this file, with or without modification,
14721 are permitted in any medium without royalty provided the copyright
14722 notice and this notice are preserved.