1 2018-06-07 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/85935
4 * graphite-scop-detection.c (find_params_in_bb): Analyze
5 condition operands with respect to the correct loop. Assert
6 the analysis doesn't fail.
8 2018-06-04 Carl Love <cel@us.ibm.com>
10 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
11 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
12 as it is slightly cheaper.
13 (first_match_or_eos_index_<mode>):
14 Calculate index using natural element order.
15 (first_match_index_<mode>):
16 Calculate index using natural element order.
17 (first_match_or_eos_index_<mode>):
18 Calculate index using natural order.
19 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
21 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
22 P9V_BUILTIN_VCLZLSBB_V16QI.
23 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
26 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
28 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
29 indentation and line wrap for many prototypes. Add missing
30 @smallexample directives around block of prototypes for vec_xl and
33 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
35 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
36 track if we pass or return IEEE 128-bit floating point.
37 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
38 C++ mangling that is compatible with GCC 8.1.
39 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
40 (init_cumulative_args): Note if we pass or return IEEE 128-bit
42 (rs6000_function_arg_advance_1): Likewise.
43 (rs6000_mangle_type): Optionally generate mangled names that match
44 what GCC 8.1 generated for IEEE 128-bit floating point types.
45 (rs6000_globalize_decl_name): If we have an external function that
46 passes or returns IEEE 128-bit types, generate a weak reference
47 from the mangled name used in GCC 8.1 to the current mangled
49 (rs6000_init_builtins): Make __ibm128 use the long double type if
50 long double is IBM extended double. Make __float128 use the long
51 double type if long double is IEEE 128-bit.
54 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
55 macro for __ibm128 built-in functions.
56 (PACK_IF): Add __ibm128 pack/unpack functions.
57 (UNPACK_IF): Likewise.
58 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
59 enable long double built-in functions if long double is IEEE
60 128-bit floating point.
61 (rs6000_invalid_builtin): Update long double built-in function
63 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
64 functions, adjust the built-in function to use the long double
65 built-in function if __ibm128 and long double are the same type.
66 * doc/extend.texi (PowerPC builtins): Update documention for
67 __builtin_{,un}pack_longdouble. Add documentation for
68 __builtin_{,un}pack_ibm128.
70 2018-06-06 Jim Wilson <jimw@sifive.com>
72 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
73 (struct machine_function): New field interrupt_mode.
74 (riscv_handle_type_attribute): New function. Add forward declaration.
75 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
76 (riscv_expand_epilogue): Check interrupt_mode field.
77 (riscv_set_current_function): Check interrupt attribute args and
78 set interrupt_mode field.
79 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
80 (riscv_sret, riscv_uret): New.
81 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
82 new arguments to interrupt attribute.
84 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
87 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
88 Don't handle -mcpu=power8 if -mpower9-vector is also used.
90 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
92 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
93 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
94 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
95 several redundant entries.
97 2018-06-06 David Malcolm <dmalcolm@redhat.com>
99 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
100 type from "rtx" to "rtx_insn *".
101 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
102 for local "call_insn", removing cast.
103 (ix86_expand_call): Likewise, introducing a "call_insn" local.
105 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
107 PR tree-optimization/86066
108 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
109 for BIT_INSERT_EXPR stores.
111 2018-06-06 Richard Biener <rguenther@suse.de>
113 PR tree-optimization/86062
114 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
116 of to be offsetted base.
118 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
120 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
121 to be static and remove check on interrupt attribute name.
123 gcc/testsuite/gcc.target/msp430/
124 * function-attributes-4.c: New test.
125 * static-interrupts.c: New test.
127 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
129 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
130 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
132 2018-06-05 Steve Ellcey <sellcey@cavium.com>
135 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
137 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
138 Remove second argument, change how error is called.
139 (aarch64_layout_arg): Remove second argument from
140 aarch64_err_no_fpadvsimd call.
141 (aarch64_init_cumulative_args): Ditto.
142 (aarch64_gimplify_va_arg_expr): Ditto.
143 * config/aarch64/aarch64.md (mov<mode>): Ditto.
145 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
147 * config/i386/i386.md (simple_return_indirect_internal): New expander.
148 (*simple_return_indirect_internal<mode>): Rename from
149 simple_return_indirect_internal. Use W mode iterator.
150 (rstorssp): New expander.
151 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
152 (clrssbsy): New expander.
153 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
155 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
157 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
159 (cmse_check_pointed_object): Likewise.
161 2018-06-05 Martin Liska <mliska@suse.cz>
163 PR gcov-profile/47618
164 * doc/invoke.texi: Document how -fprofile-dir format
167 2018-06-05 Richard Biener <rguenther@suse.de>
169 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
170 removal pretend DOM info isn't available so we do not update
171 it and only remove edges, not dominated blocks. Actually free
172 DOM info in case we removed something. Remove unreachable blocks.
173 (mfb_keep_latches): Work with either DOM info or marked backedges.
174 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
175 first. Mark backedges if DOM info isn't available.
176 (Re-)compute DOM info after cleanup_control_flow_pre.
178 2018-06-05 Richard Biener <rguenther@suse.de>
180 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
181 (locus_discrim_hasher::hash): Adjust.
182 (locus_discrim_hasher::equal): Likewise.
183 (next_discriminator_for_locus): Work on line directly.
184 (same_line_p): Pass in expanded locus1 as well.
185 (assign_discriminators): Avoid redundant location expansions.
187 2018-06-05 Richard Biener <rguenther@suse.de>
189 PR tree-optimization/86046
190 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
191 if required after clearing TREE_ADDRESSABLE.
193 2018-06-05 Richard Biener <rguenther@suse.de>
195 PR tree-optimization/86047
196 * tree-ssa-loop.c (for_each_index): Glob handling of all
197 decls and constants and really handle all of them.
199 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
202 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
203 qualifier_void_pointer and qualifier_const_void_pointer.
204 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
205 (arm_init_builtins): Handle the above.
206 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
207 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
208 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
211 2018-06-05 Martin Liska <mliska@suse.cz>
213 * auto-profile.c (read_autofdo_file): Do not use
214 gcov_ctr_summary struct.
215 (afdo_callsite_hot_enough_for_early_inline): Likewise.
216 * coverage.c (struct counts_entry): Likewise.
217 (read_counts_file): Read just single summary entry.
218 (get_coverage_counts): Use gcov_summary struct.
219 * coverage.h (get_coverage_counts): Likewise.
220 * gcov-dump.c (dump_working_sets): Likewise.
221 (tag_summary): Dump just single summary.
222 * gcov-io.c (gcov_write_summary): Write just histogram
224 (gcov_read_summary): Read just single summary.
225 (compute_working_sets): Use gcov_summary struct.
226 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
227 of GCOV_COUNTERS_SUMMABLE.
228 (GCOV_COUNTERS_SUMMABLE): Remove.
229 (GCOV_FIRST_VALUE_COUNTER): Replace with
230 GCOV_COUNTER_V_INTERVAL.
231 (struct gcov_ctr_summary): Remove.
232 (struct gcov_summary): Directly use fields of former
234 (compute_working_sets): Use gcov_summary struct.
235 * gcov.c (read_count_file): Do not use ctrs fields.
236 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
238 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
240 * profile.c: Likewise.
241 * profile.h: Likewise.
243 2018-06-05 Martin Liska <mliska@suse.cz>
245 PR gcov-profile/84846
246 * gcov.c (output_lines): Print working directory only
247 in intermediate format.
249 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
251 * config/s390/s390-builtin-types.def: Add void function type.
252 * config/s390/s390-builtins.def: Use the function type for the
255 2018-06-04 Jim Wilson <jimw@sifive.com>
257 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
259 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
260 and maybe_eh_return. Change regno to unsigned int. Use new args to
261 handle EH_RETURN_DATA_REGNO registers properly.
262 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
263 (riscv_expand_epilogue): Update comment. Change argument name and
264 type. Update code to use new name and type. Pass new args to
265 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
267 * config/riscv/riscv.md (NORMAL_RETURN): New.
268 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
269 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
270 (eh_return): Call gen_eh_return_internal and emit barrier.
271 (eh_return_internal): Call riscv_expand_epilogue.
273 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
275 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
276 bit_insertion field and declare can_be_merged_into method.
277 (merged_store_group::can_be_merged_into): New method.
278 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
279 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
280 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
282 2018-06-04 Richard Biener <rguenther@suse.de>
284 PR tree-optimization/85955
285 * builtins.c (fold_builtin_sincos): Convert pointers to
286 destination to appropriate type before dereferencing.
288 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
290 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
292 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
294 * expr.c (expand_expr_real_1): Force the operand into memory if
295 its TYPE_MODE is BLKmode and if there is no integer mode for
296 the number of bits being extracted.
298 2018-06-04 Jakub Jelinek <jakub@redhat.com>
302 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
303 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
305 2018-06-04 Richard Biener <rguenther@suse.de>
307 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
308 (cleanup_tree_cfg_noloop): ... single caller. Do
309 start_recording_case_labels later.
311 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
313 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
314 to _IMMINTRIN_H_INCLUDED.
315 * config/i386/pconfigintrin.h: Ditto.
316 * config/i386/waitpkgintrin.h: Ditto.
317 * config/i386/immintrin.h: Add includes for sgxintrin.h,
318 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
319 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
320 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
321 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
322 waitpkgintrin.h and cldemoteintrin.h.
324 2018-06-04 Richard Biener <rguenther@suse.de>
326 PR tree-optimization/86038
327 * tracer.c (find_best_successor): Check probability for
328 being initialized, bail out if not.
330 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
333 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
334 of bits to ignore when comparing architectures.
336 2018-06-04 Jakub Jelinek <jakub@redhat.com>
338 PR tree-optimization/69615
339 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
340 maximum or minimum of the type, try to merge it also as if
341 range1 is + [-, x - 1] or + [x + 1, -].
344 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
346 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
348 PR tree-optimization/86034
349 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
350 the unsigned bitfield type in a bit insertion sequence if it does not
351 have a larger precision than the bitfield size.
352 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
354 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
356 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
358 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
360 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
361 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
362 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
363 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
365 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
367 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
368 Disable -fdelete-null-pointer-checks for ELF toolchain.
370 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
371 Kito Cheng <kito.cheng@gmail.com>
373 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
374 (nds32le-*-*, nds32be-*-*): Integrate checking process.
375 (nds32*-*-*): Add glibc and uclibc conditions.
376 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
377 (TARGET_EXCEPT_UNWIND_INFO): Define.
378 * config/nds32/elf.h: New file.
379 * config/nds32/linux.h: New file.
380 * config/nds32/nds32-elf.opt: New file.
381 * config/nds32/nds32-linux.opt: New file.
382 * config/nds32/nds32-fp-as-gp.c
383 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
384 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
386 (nds32_asm_file_end): Ditto.
387 (nds32_print_operand): Ditto.
388 (nds32_insert_attributes): Ditto.
389 (nds32_init_libfuncs): New function.
390 (TARGET_HAVE_TLS): Define.
391 (TARGET_INIT_LIBFUNCS): Define.
392 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
394 (TARGET_ELF): Apply different mcmodel setting.
395 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
396 been migrated into elf.h and linux.h files.
397 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
398 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
399 (mcmodel): The content has been migrated into nds32-elf.opt and
400 nds32-linux.opt files.
401 * config/nds32/t-elf: New file.
402 * config/nds32/t-linux: New file.
404 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
405 Shiva Chen <shiva0217@gmail.com>
407 * config/nds32/constants.md (unspec_volatile_element): Add
408 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
409 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
411 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
412 (make_pass_nds32_fp_as_gp): Declare.
413 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
415 (nds32_asm_function_end_prologue): Remove unused asm output.
416 (nds32_asm_function_begin_epilogue): Remove unused asm output.
417 (nds32_asm_file_start): Output necessary fp_as_gp information.
418 (nds32_option_override): Adjust register usage.
419 (nds32_expand_prologue): Consider fp_as_gp situation.
420 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
421 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
424 (simple_return): Ditto.
425 (omit_fp_begin): Output special directive for fp_as_gp.
426 (omit_fp_end): Output special directive for fp_as_gp.
427 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
428 mforbid-fp-as-gp): New options.
430 2018-06-01 Mark Wielaard <mark@klomp.org>
432 * dwarf2out.c (dwarf2out_finish): Remove generation of
435 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
437 * gimple-ssa-store-merging.c: Include gimple-fold.h.
438 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
439 (struct merged_store_group): Add bit_insertion field.
440 (dump_char_array): Use standard hexadecimal format.
441 (merged_store_group::merged_store_group): Set bit_insertion to false.
442 (merged_store_group::apply_stores): Use optimal buffer size. Deal
443 with BIT_INSERT_EXPR stores. Move up code updating the mask and
444 also print the mask in the dump file.
445 (pass_store_merging::gate): Minor tweak.
446 (imm_store_chain_info::coalesce_immediate): Fix wrong association
447 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
448 stores with INTEGER_CST stores.
449 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
450 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
451 and use it throughout. Generate bit insertion sequences if need be.
452 (pass_store_merging::process_store): Remove redundant condition.
453 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
455 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
457 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
458 the 128-bit floating point types. Fix function comment.
460 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
462 * config/aarch64/aarch64-simd.md
463 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
465 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
468 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
470 PR tree-optimization/85989
471 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
473 (backprop::intersect_uses): Check it when deciding whether this
474 is a backedge reference.
475 (backprop::process_block): Add each phi to m_visited_phis
476 after visiting it, then clear it at the end.
478 2018-06-01 Richard Biener <rguenther@suse.de>
480 * tree-vectorizer.h (vect_dr_stmt): New function.
481 (vect_get_load_cost): Adjust.
482 (vect_get_store_cost): Likewise.
483 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
484 Use vect_dr_stmt instead of DR_SMTT.
485 (vect_record_base_alignments): Likewise.
486 (vect_calculate_target_alignment): Likewise.
487 (vect_compute_data_ref_alignment): Likewise and make static.
488 (vect_update_misalignment_for_peel): Likewise.
489 (vect_verify_datarefs_alignment): Likewise.
490 (vector_alignment_reachable_p): Likewise.
491 (vect_get_data_access_cost): Likewise. Pass down
492 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
493 (vect_get_peeling_costs_all_drs): Likewise.
494 (vect_peeling_hash_get_lowest_cost): Likewise.
495 (vect_enhance_data_refs_alignment): Likewise.
496 (vect_find_same_alignment_drs): Likewise.
497 (vect_analyze_data_refs_alignment): Likewise.
498 (vect_analyze_group_access_1): Likewise.
499 (vect_analyze_group_access): Likewise.
500 (vect_analyze_data_ref_access): Likewise.
501 (vect_analyze_data_ref_accesses): Likewise.
502 (vect_vfa_segment_size): Likewise.
503 (vect_small_gap_p): Likewise.
504 (vectorizable_with_step_bound_p): Likewise.
505 (vect_prune_runtime_alias_test_list): Likewise.
506 (vect_analyze_data_refs): Likewise.
507 (vect_supportable_dr_alignment): Likewise.
508 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
509 (vect_gen_prolog_loop_niters): Likewise.
510 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
511 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
513 (vect_recog_mask_conversion_pattern): Likewise.
514 (vect_try_gather_scatter_pattern): Likewise.
515 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
516 to vect_get_store_cost.
517 (vect_get_store_cost): Get stmt_info instead of DR.
518 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
519 (vect_get_load_cost): Get stmt_info instead of DR.
521 2018-06-01 Richard Biener <rguenther@suse.de>
524 * gimple-fold.c (var_decl_component_p): Also allow offsetted
525 vars wrapped in MEM_REFs.
527 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
529 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
530 Fix subreg tests so that we only return a choice between
531 GENERAL_REGS and FP_REGS if the original classes included both.
533 2018-06-01 Richard Biener <rguenther@suse.de>
536 * tree-ssa-structalias.c (get_function_part_constraint):
537 Handle NULL fi->decl.
538 (find_func_aliases_for_call): Properly handle indirect
540 (find_func_clobbers): Likewise.
541 (ipa_pta_execute): Likewise.
542 (create_variable_info_for): For functions that are ifunc_resolver
543 resolve to a varinfo that contains the result of the resolver call.
544 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
547 2018-05-31 Michael Collison <michael.collison@arm.com>
549 * config/aarch64/aarch64.md:
550 (*fix_to_zero_extenddfdi2): New pattern.
551 * gcc.target/aarch64/fix_extend1.c: New testcase.
553 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
557 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
558 and BUILT_IN_STRNCMP_EQ.
559 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
561 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
562 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
563 (gimple_fold_builtin): Likewise.
564 * tree-ssa-strlen.c (compute_string_length): New function.
565 (determine_min_obsize): New function.
566 (handle_builtin_string_cmp): New function to handle calls to
567 string compare functions.
568 (strlen_optimize_stmt): Add handling to builtin string compare
570 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
571 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
572 * tree.c (build_common_builtin_nodes): Add new defines of
573 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
575 2018-05-31 Jakub Jelinek <jakub@redhat.com>
578 * bb-reorder.c (pass_partition_blocks::gate): Return false for
579 functions with naked attribute.
581 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
583 * config/i386/sse.md (avx_vec_concat<mode>):
584 Substitute concat_tg_mode mode attribute with xtg_mode.
585 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
586 (concat_tg_mode): Remove mode attribute.
588 2018-05-31 Martin Sebor <msebor@redhat.com>
591 * calls.c (alloc_max_size): Correct a logic error/typo.
592 Treat excessive arguments as infinite. Warn for invalid arguments.
593 * doc/invoke.texi (-Walloc-size-larger-than): Update.
595 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
598 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
599 and movx for Haswell.
601 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
602 Cesar Philippidis <cesar@codesourcery.com>
605 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
606 when emitting error on private/firstprivate reductions.
607 * omp-low.c (lower_omp_target): Avoid reference-type processing
608 on pointers for firstprivate clause.
610 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
612 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
615 * config/aarch64/aarch64-simd.md
616 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
617 (aarch64_ld1_x3_<mode>): Likewise
618 (aarch64_st1x2<VALLDIF:mode>): Likewise
619 (aarch64_st1_x2_<mode>): Likewise
620 (aarch64_st1x3<VALLDIF:mode>): Likewise
621 (aarch64_st1_x3_<mode>): Likewise
622 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
623 (vld1_s8_x3): Likewise.
624 (vld1_u16_x3): Likewise.
625 (vld1_s16_x3): Likewise.
626 (vld1_u32_x3): Likewise.
627 (vld1_s32_x3): Likewise.
628 (vld1_u64_x3): Likewise.
629 (vld1_s64_x3): Likewise.
630 (vld1_f16_x3): Likewise.
631 (vld1_f32_x3): Likewise.
632 (vld1_f64_x3): Likewise.
633 (vld1_p8_x3): Likewise.
634 (vld1_p16_x3): Likewise.
635 (vld1_p64_x3): Likewise.
636 (vld1q_u8_x3): Likewise.
637 (vld1q_s8_x3): Likewise.
638 (vld1q_u16_x3): Likewise.
639 (vld1q_s16_x3): Likewise.
640 (vld1q_u32_x3): Likewise.
641 (vld1q_s32_x3): Likewise.
642 (vld1q_u64_x3): Likewise.
643 (vld1q_s64_x3): Likewise.
644 (vld1q_f16_x3): Likewise.
645 (vld1q_f32_x3): Likewise.
646 (vld1q_f64_x3): Likewise.
647 (vld1q_p8_x3): Likewise.
648 (vld1q_p16_x3): Likewise.
649 (vld1q_p64_x3): Likewise.
650 (vst1_s64_x2): Likewise.
651 (vst1_u64_x2): Likewise.
652 (vst1_f64_x2): Likewise.
653 (vst1_s8_x2): Likewise.
654 (vst1_p8_x2): Likewise.
655 (vst1_s16_x2): Likewise.
656 (vst1_p16_x2): Likewise.
657 (vst1_s32_x2): Likewise.
658 (vst1_u8_x2): Likewise.
659 (vst1_u16_x2): Likewise.
660 (vst1_u32_x2): Likewise.
661 (vst1_f16_x2): Likewise.
662 (vst1_f32_x2): Likewise.
663 (vst1_p64_x2): Likewise.
664 (vst1q_s8_x2): Likewise.
665 (vst1q_p8_x2): Likewise.
666 (vst1q_s16_x2): Likewise.
667 (vst1q_p16_x2): Likewise.
668 (vst1q_s32_x2): Likewise.
669 (vst1q_s64_x2): Likewise.
670 (vst1q_u8_x2): Likewise.
671 (vst1q_u16_x2): Likewise.
672 (vst1q_u32_x2): Likewise.
673 (vst1q_u64_x2): Likewise.
674 (vst1q_f16_x2): Likewise.
675 (vst1q_f32_x2): Likewise.
676 (vst1q_f64_x2): Likewise.
677 (vst1q_p64_x2): Likewise.
678 (vst1_s64_x3): Likewise.
679 (vst1_u64_x3): Likewise.
680 (vst1_f64_x3): Likewise.
681 (vst1_s8_x3): Likewise.
682 (vst1_p8_x3): Likewise.
683 (vst1_s16_x3): Likewise.
684 (vst1_p16_x3): Likewise.
685 (vst1_s32_x3): Likewise.
686 (vst1_u8_x3): Likewise.
687 (vst1_u16_x3): Likewise.
688 (vst1_u32_x3): Likewise.
689 (vst1_f16_x3): Likewise.
690 (vst1_f32_x3): Likewise.
691 (vst1_p64_x3): Likewise.
692 (vst1q_s8_x3): Likewise.
693 (vst1q_p8_x3): Likewise.
694 (vst1q_s16_x3): Likewise.
695 (vst1q_p16_x3): Likewise.
696 (vst1q_s32_x3): Likewise.
697 (vst1q_s64_x3): Likewise.
698 (vst1q_u8_x3): Likewise.
699 (vst1q_u16_x3): Likewise.
700 (vst1q_u32_x3): Likewise.
701 (vst1q_u64_x3): Likewise.
702 (vst1q_f16_x3): Likewise.
703 (vst1q_f32_x3): Likewise.
704 (vst1q_f64_x3): Likewise.
705 (vst1q_p64_x3): Likewise.
707 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
709 * config/msp430/msp430.c (msp430_output_labelref): Prepend
710 user_label_prefix to name.
712 * tree-core.h: Update comment about the format of NAME string
713 passed to handler in attribute_spec.
715 * config/msp430/msp430.md: Remove erroneous subreg expression from
716 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
717 zero_extend{q,h}isi2.
719 2018-05-30 Borislav Petkov <bp@suse.de>
721 * doc/extend.texi: Document some architecture specific
722 constraints and sort entries.
724 2018-05-30 Martin Sebor <msebor@redhat.com>
727 * builtins.c (expand_builtin_stpcpy_1): New function.
728 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
729 only if the former succeeds.
731 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
733 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
736 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
738 * doc/invoke.texi (-flinker-output): Document
740 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
742 * passes.c (ipa_write_summaries): Only modify statements if body
744 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
745 incrementally linking.
746 (ipa_passes): Likewise.
747 * lto-cgraph.c (lto_output_node): When incrementally linking do not
748 pass down resolution info.
749 * common.opt (flag_incremental_link): Update info.
750 * gcc.c (plugin specs): Turn flinker-output=* to
751 -plugin-opt=-linker-output-known
752 * toplev.c (compile_file): Also cut compilation when doing incremental
754 * flag-types. (enum lto_partition_model): Add
755 LTO_LINKER_OUTPUT_NOLTOREL.
756 (invoke.texi): Add -flinker-output docs.
757 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
758 link same way as WPA; do not stream in dead initializers.
760 * dwarf2out.c (dwarf2out_die_ref_for_decl,
761 darf2out_register_external_decl): Support incremental link.
763 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
765 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
767 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
769 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
770 it down to simple_object_copy_lto_debug_sections.
771 (run_gcc): Determine incremental LTO link time and configure
772 lto1 into non-wpa mode, disable renaming of debug sections.
774 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
776 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
777 descriptions of various incorrectly documented functions.
779 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
782 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
783 address check not strict.
785 2018-05-30 Richard Biener <rguenther@suse.de>
787 PR tree-optimization/85964
788 * tracer.c (better_p): Drop initialized count check, we only
789 call the function with initialized counts now.
790 (find_best_successor): Do find a best edge if one
791 has uninitialized count.
792 (find_best_predecessor): Likewise. Do BB frequency check only
793 if count is initialized.
795 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
797 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
798 (aarch64_ldrstr_offset_compare): New.
799 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
800 load/store orderings.
801 (aarch64_gen_adjusted_ldpstp): Likewise.
803 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
805 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
806 Check for subset of GENERAL_REGS and FP_REGS.
807 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
810 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
812 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
813 and wi::to_poly_offset. Add the current offset and then check
814 whether the sum fits, rather than using an unchecked addition of
815 a checked term. Check for a shwi rather than a uhwi.
816 * expr.c (get_bit_range): Use tree_to_poly_uint64.
817 (store_constructor): Use poly_int_tree_p.
818 (expand_expr_real_1): Likewise.
819 * function.c (assign_temp): Likewise.
820 * fold-const.c (const_binop): Use poly_int_tree_p and
822 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
824 * ipa-icf-gimple.c (func_checker::compare_operand): Use
825 to_poly_offset for MEM offsets.
826 * ipa-icf.c (sem_variable::equals): Likewise.
827 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
828 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
829 wi::to_poly_offset for BIT_FIELD_REF offsets.
830 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
832 * var-tracking.c (emit_note_insn_var_location): Use
835 2018-05-29 Jim Wilson <jimw@sifive.com>
837 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
839 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
842 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
843 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
845 (sse4_1_round<mode>2): Use nonimmediate_operand
846 for operand 1 predicate.
848 2018-05-29 Martin Sebor <msebor@redhat.com>
849 Richard Biener <rguenther@suse.de>
852 * calls.c (get_size_range): Call determine_value_range instead
854 * tree-vrp.h (determine_value_range): Declared new function.
855 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
857 2018-05-29 Richard Biener <rguenther@suse.de>
859 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
860 sure to use non-pattern stmts for get_earlier_stmt arguments.
861 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
862 called on pattern stmts.
863 (get_later_stmt): Likewise.
865 2018-05-29 Martin Liska <mliska@suse.cz>
867 PR gcov-profile/85759
868 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
871 2018-05-29 Jakub Jelinek <jakub@redhat.com>
873 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
875 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
879 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
880 VEC_PACK_FLOAT_EXPR): New tree codes.
881 * tree-pretty-print.c (op_code_prio): Handle
882 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
883 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
884 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
885 * tree-inline.c (estimate_operator_cost): Likewise.
886 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
887 * fold-const.c (const_binop): Likewise.
888 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
889 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
890 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
891 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
892 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
893 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
894 * expr.c (expand_expr_real_2): Likewise.
895 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
896 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
897 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
899 * optabs.c (expand_widen_pattern_expr): For
900 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
901 sign from result type rather than operand's type.
902 (expand_binop_directly): For vec_packu_float_optab and
903 vec_packs_float_optab allow result type to be different from operand's
905 * optabs-tree.c (optab_for_tree_code): Handle
906 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
907 VEC_PACK_FLOAT_EXPR. Formatting fixes.
908 * tree-vect-generic.c (expand_vector_operations_1): Handle
909 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
911 * tree-vect-stmts.c (supportable_widening_operation): Handle
913 (supportable_narrowing_operation): Handle FLOAT_EXPR.
914 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
915 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
916 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
917 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
919 (vec_pack<floatprefix>_float_<mode>): New expander.
920 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
922 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
923 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
924 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
925 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
926 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
928 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
929 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
930 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
931 VEC_PACK_FLOAT_EXPR): Document.
933 2018-05-29 Richard Biener <rguenther@suse.de>
935 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
937 (stmt_vec_info_vec): Make pointer.
938 (init_stmt_vec_info_vec): Remove.
939 (free_stmt_vec_info_vec): Likewise.
940 (set_stmt_vec_info_vec): New function.
941 (free_stmt_vec_infos): Likewise.
942 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
943 (set_vinfo_for_stmt): Likewise.
944 (get_earlier_stmt): Likewise.
945 (get_later_stmt): Likewise.
946 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
947 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
948 (vec_info::~vec_info): Free stmt_vec_infos.
949 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
950 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
951 (pass_slp_vectorize::execute): Likewise.
952 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
953 (free_stmt_vec_info_vec): Likewise.
954 (set_stmt_vec_info_vec): New function.
955 (free_stmt_vec_infos): Likewise.
956 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
957 the global stmt_vec_info_vec.
958 * tree-parloops.c (gather_scalar_reductions): Use
959 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
962 2018-05-29 Richard Biener <rguenther@suse.de>
964 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
966 2018-05-29 Martin Liska <mliska@suse.cz>
967 David Malcolm <dmalcolm@redhat.com>
969 * vec.c (test_reverse): New.
970 (vec_c_tests): Add new test.
971 * vec.h (vl_ptr>::reverse): New function.
973 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
975 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
977 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
980 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
982 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
984 2018-05-28 Richard Biener <rguenther@suse.de>
986 PR tree-optimization/85933
987 * tree-vect-data-refs.c (vect_record_base_alignments): Only
988 look at stmts marked as vectorizable.
990 2018-05-28 Richard Biener <rguenther@suse.de>
992 PR tree-optimization/85934
993 * tree-vect-generic.c (expand_vector_operations_1): Hoist
994 vector boolean check before scalar optimization.
996 2018-05-28 Jakub Jelinek <jakub@redhat.com>
998 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
1001 2018-05-28 Mark Wielaard <mark@klomp.org>
1003 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
1004 if it is an expression containing a minus sign.
1006 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
1008 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1010 2018-05-27 Paul Koning <ni1d@arrl.net>
1012 * config/pdp11/pdp11.md (truncsihi2): Remove.
1014 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
1015 Chung-Ju Wu <jasonwucj@gmail.com>
1017 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
1019 (unaligned_store_dw): Ditto.
1020 * config/nds32/nds32-memory-manipulation.c
1021 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
1022 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
1023 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
1024 (emit_setmem_word_loop): Rename to ...
1025 (emit_setmem_doubleword_loop): ... this.
1026 (nds32_gen_dup_4_byte_to_word_value): New function.
1027 (nds32_gen_dup_8_byte_to_double_word_value): New function.
1028 (nds32_expand_setmem_loop): Refine implementation.
1029 (nds32_expand_setmem_loop_v3m): Ditto.
1030 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
1033 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1035 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
1037 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
1039 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
1040 (nds32_init_machine_status): Initialize machine->attr_naked_p and
1041 machine->attr_no_prologue_p.
1042 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
1043 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
1044 (nds32_expand_epilogue): Consider attr_naked_p.
1045 (nds32_expand_epilogue_v3pop): Likewise.
1046 (nds32_can_use_return_insn): Likewise.
1047 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
1048 attr_no_prologue_p fields.
1049 * config/nds32/nds32.opt (mret-in-naked-func): New option.
1051 2018-05-27 Jakub Jelinek <jakub@redhat.com>
1054 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
1056 * config/i386/sse.md
1057 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
1059 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
1061 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
1063 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
1065 (*<floatsuffix>floatv2div2sf2): Rename to ...
1066 (*float<floatunssuffix>v2div2sf2): ... this.
1067 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
1068 (float<floatunssuffix>v2div2sf2_mask): ... this.
1069 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
1070 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
1071 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
1073 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
1075 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1077 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
1079 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1081 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
1083 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
1084 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
1085 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
1086 gen_ufix_truncv8dfv8si2.
1087 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
1088 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
1089 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
1090 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
1091 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
1092 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
1093 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
1094 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
1096 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1100 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
1102 2018-05-25 Jim Wilson <jimw@sifive.com>
1104 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
1105 * config/riscv/riscv.c (struct machine_function): Add
1106 interrupt_handler_p and attribute_checked_p fields.
1107 (riscv_attribute_table): Add interrupt.
1108 (riscv_interrupt_type_p): New.
1109 (riscv_save_reg_p): Save extra regs for interrupt handler.
1110 (riscv_use_save_libcall): Return false for interrupt handler.
1111 (riscv_first_stack_step): Add forward declaration.
1112 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
1113 for interrupt handler with large frame. Use it for saved reg list.
1114 (riscv_expand_prologue): Move flag_stack_usage_info support to
1115 eliminate duplication.
1116 (riscv_expand_epilogue): Generate mret for interrupt handler.
1117 (riscv_epilogue_uses): New.
1118 (riscv_can_use_return_insn): Return false for interrupt handler.
1119 (riscv_function_ok_for_sibcall): Likewise.
1120 (riscv_set_current_function): Add interrupt handler support.
1121 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
1122 * config/riscv/riscv.md (UNSPECV_MRET): New.
1124 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1126 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1128 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1130 PR tree-optimization/85712
1131 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1132 this candidate has already been replaced in-situ by a copy.
1134 2018-05-25 Jason Merrill <jason@redhat.com>
1136 PR c++/80485 - inline function non-zero address.
1137 * symtab.c (nonzero_address): Check DECL_COMDAT.
1139 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1142 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1143 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1144 extension of SImode operation. Use const123_operand predicate.
1145 (*saddsi_1): Remove.
1146 (*saddl_se_1): Ditto.
1148 (*ssubl_se_1): Ditto.
1149 * config/alpha/predicates.md (const123_operand): New predicate.
1150 * config/alpha/constraints.md (P): Use IN_RANGE.
1152 2018-05-25 Richard Biener <rguenther@suse.de>
1154 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1156 (ref_maybe_used_by_stmt_p): Likewise.
1157 (stmt_may_clobber_ref_p): Likewise.
1158 (stmt_may_clobber_ref_p_1): Likewise.
1159 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1161 (ref_maybe_used_by_stmt_p): Likewise.
1162 (stmt_may_clobber_ref_p): Likewise.
1163 (stmt_may_clobber_ref_p_1): Likewise.
1164 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1165 the alias oracle to disambiguate DRs with stmts DR analysis
1167 (vect_analyze_data_refs): Do not give up on not analyzable
1168 DRs for BB vectorization. Remove code truncating the dataref
1171 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1174 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1175 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1176 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1178 2018-05-25 Richard Biener <rguenther@suse.de>
1180 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1181 function, combining stmt data ref gathering and fatal analysis
1183 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1184 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1185 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1186 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1187 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1189 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1191 PR tree-optimization/85720
1192 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1193 SCC if all partitions are builtins.
1194 (version_loop_by_alias_check): New parameter. Generate cancelable
1195 runtime alias check if all partitions are builtins.
1196 (distribute_loop): Update call to above function.
1198 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1200 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1201 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1202 (parm_default_def_partition_arg): Ditto.
1203 (set_parm_default_def_partition): Ditto.
1204 (get_parm_default_def_partitions): Ditto and make it static.
1205 (get_undefined_value_partitions): Ditto and make it static.
1206 (remove_ssa_form): Refactor call to init_var_map here.
1207 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1208 computation for loop region.
1209 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1210 (register_default_def): Delete.
1211 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1212 (parm_default_def_partition_arg): Ditto.
1213 (set_parm_default_def_partition): Ditto.
1214 (get_parm_default_def_partitions): Ditto and make it static.
1215 (get_undefined_value_partitions): Ditto and make it static.
1216 (coalesce_with_default, coalesce_with_default): Update comment.
1217 (create_coalesce_list_for_region): New func factored out from
1218 create_outofssa_var_map.
1219 (populate_coalesce_list_for_outofssa): New func factored out from
1220 create_outofssa_var_map and coalesce_ssa_name.
1221 (create_outofssa_var_map): Delete.
1222 (coalesce_ssa_name): Refactor to support live range computation.
1223 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1224 (get_parm_default_def_partitions): Delete.
1225 (get_undefined_value_partitions): Ditto.
1226 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1227 computation for loop region.
1228 (new_tree_live_info, loe_visit_block): Ditto.
1229 (live_worklist, set_var_live_on_entry): Ditto.
1230 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1231 * tree-ssa-live.h (struct _var_map): New fields.
1232 (init_var_map): Change decl.
1233 (region_contains_p): New.
1235 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1237 * tree-ssa-live.h (live_merge_and_clear): Delete.
1239 2018-05-25 Richard Biener <rguenther@suse.de>
1242 * tree-dump.c (dequeue_and_dump): Remove access to removed
1243 operand 2 of a SWITCH_EXPR.
1245 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1247 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1248 multiplication and division.
1249 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1250 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1251 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1252 (cond_udiv_optab, cond_umod_optab): New optabs.
1253 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1254 (IFN_COND_RDIV): New internal functions.
1255 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1256 TRUNC_MOD_EXPR and RDIV_EXPR.
1257 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1258 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1260 (SVE_INT_BINARY): Include mult.
1261 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1262 (optab, sve_int_op): Handle mult.
1263 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1265 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1266 for SVE_INT_BINARY_SD.
1268 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1270 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1271 (optab, sve_int_op): Handle div and udiv.
1272 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1273 for SVE_INT_BINARY_SD.
1274 (*<optab><mode>3): New insn for the same.
1276 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1278 * tree-vect-patterns.c: Include predict.h.
1279 (vect_recog_divmod_pattern): Restrict check for division support
1280 to when optimizing for size.
1282 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1284 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1285 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1286 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1287 (gimple_match_op::set_op): Likewise.
1288 (gimple_resimplify4): Declare.
1289 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1290 (expr::gen_transform): Likewise.
1291 (decision_tree::gen): Generate a simplification routine for 4 operands.
1292 * gimple-match-head.c (gimple_simplify): Add an overload for
1293 4 operands. In the top-level function, handle up to 4 call
1294 arguments and call gimple_resimplify4.
1295 (gimple_resimplify4): New function.
1296 (build_call_internal): Pass a fourth operand.
1297 (maybe_push_to_seq): Likewise.
1298 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1299 Fold VEC_COND_EXPRs of an operation and a default value into
1300 an IFN_COND_* function if possible.
1301 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1303 (SVE_COND_FP_BINARY): Include them.
1304 (optab, sve_fp_op): Handle them.
1305 (SVE_INT_BINARY_REV): New code iterator.
1306 (SVE_COND_FP_BINARY_REV): New int iterator.
1307 (commutative): New int attribute.
1308 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1310 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1312 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1313 (*cond_<optab><mode>): New patterns for reversed operands.
1315 2018-05-25 Richard Biener <rguenther@suse.de>
1317 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1318 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1319 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1320 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1321 * tree-vect-data-refs.c (everywhere): Adjust users.
1322 * tree-vect-loop.c (everywhere): Likewise.
1323 * tree-vect-slp.c (everywhere): Likewise.
1324 * tree-vect-stmts.c (everywhere): Likewise.
1325 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1327 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1329 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1331 (gcc_cv_as_section_exclude): ... this.
1332 Try Solaris as #exclude syntax.
1333 * configure: Regenerate.
1334 * config.in: Regenerate.
1335 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1337 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1338 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1340 * varasm.c (default_elf_asm_named_section): Don't check if
1341 HAVE_GAS_SECTION_EXCLUDE is defined.
1343 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1345 * doc/md.texi: Update the documentation of the cond_* optabs
1346 to mention the new final operand. Fix GET_MODE_NUNITS call.
1347 Describe the scalar case too.
1348 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1349 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1351 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1352 (get_conditional_internal_fn): Update comment.
1353 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1354 accumulator value as a final argument to conditional functions.
1355 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1356 a define_expand and add an "else" operand. Assert for now that
1357 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1358 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1359 (*cond_<optab><mode>): New patterns.
1360 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1361 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1362 (UNSPEC_COND_EOR): Delete.
1363 (optab): Remove associated mappings.
1364 (SVE_INT_BINARY): New code iterator.
1365 (sve_int_op): Remove int attribute and add "minus" to the code
1367 (SVE_COND_INT_OP): Delete.
1368 (SVE_COND_FP_OP): Rename to...
1369 (SVE_COND_FP_BINARY): ...this.
1371 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1373 * optabs.c (can_reuse_operands_p): New function.
1374 (maybe_legitimize_operands): Try to reuse the results for
1377 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1379 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1380 Add {q} suffix to insn mnemonic.
1382 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1384 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1385 (msp430_warn_func_return): New.
1387 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1389 * fold-const.c (tree_nonzero_bits): New function.
1390 * fold-const.h (tree_nonzero_bits): Likewise.
1391 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1392 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1394 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1398 * varasm.c (assemble_alias): Check ifunc_resolver only on
1401 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1404 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1405 when memory input operand is handled.
1407 2018-05-24 Luis Machado <luis.machado@linaro.org>
1409 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1411 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1413 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1415 * match.pd: Delay FMA folds until after vectorization.
1417 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1420 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1421 address check not strict.
1423 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1425 * gimple-match.h (gimple_match_op): New class.
1426 (mprts_hook): Replace parameters with a gimple_match_op *.
1427 (maybe_build_generic_op): Likewise.
1428 (gimple_simplified_result_is_gimple_val): Replace parameters with
1429 a const gimple_match_op *.
1430 (gimple_simplify): Replace code_helper * and tree * parameters with
1431 a gimple_match_op * parameter.
1432 (gimple_resimplify1): Replace code_helper *, tree and tree *
1433 parameters with a gimple_match_op * parameter.
1434 (gimple_resimplify2): Likewise.
1435 (gimple_resimplify3): Likewise.
1436 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1437 parameters with a gimple_match_op * parameter.
1438 * gimple-match-head.c (gimple_simplify): Change prototypes of
1439 auto-generated functions to take a gimple_match_op * instead of
1440 separate code_helper * and tree * parameters. Make the same
1441 change in the top-level overload and update calls to the
1442 gimple_resimplify routines. Update calls to the auto-generated
1443 functions and to maybe_push_res_to_seq in the publicly-facing
1444 operation-specific gimple_simplify overloads.
1445 (gimple_match_op::MAX_NUM_OPS): Define.
1446 (gimple_resimplify1): Replace rcode and ops with a single res_op
1447 parameter. Update call to gimple_simplify.
1448 (gimple_resimplify2): Likewise.
1449 (gimple_resimplify3): Likewise.
1450 (mprts_hook): Replace parameters with a gimple_match_op *.
1451 (maybe_build_generic_op): Likewise.
1452 (build_call_internal): Replace type, nargs and ops with
1453 a gimple_match_op *.
1454 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1455 with a single gimple_match_op *. Update calls to mprts_hook,
1456 build_call_internal and gimple_simplified_result_is_gimple_val.
1457 Factor out code that is common to the tree_code and combined_fn cases.
1458 * genmatch.c (expr::gen_transform): Replace tem_code and
1459 tem_ops with a gimple_match_op called tem_op. Update calls
1460 to the gimple_resimplify functions and maybe_push_res_to_seq.
1461 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1462 res_ops. Update call to the gimple_resimplify functions.
1463 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1464 (decision_tree::gen): Make the functions take a gimple_match_op *
1465 called res_op instead of separate res_code and res_ops parameters.
1466 Update call accordingly.
1467 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1468 and ops with a single res_op parameter. Update calls to
1469 maybe_build_generic_op and maybe_push_res_to_seq.
1470 (fold_stmt_1): Update calls to gimple_simplify and
1471 replace_stmt_with_simplification.
1472 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1473 and gimple_simplified_result_is_gimple_val.
1474 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1476 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1477 with a gimple_match_op *.
1478 (vn_nary_build_or_lookup): Likewise. Update call to
1479 vn_nary_build_or_lookup_1.
1480 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1481 gimple_match_op *. Update calls to the gimple_resimplify routines
1482 and to gimple_simplified_result_is_gimple_val.
1483 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1484 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1485 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1486 (visit_nary_op): Likewise.
1487 (visit_reference_op_load): Likewise.
1489 2018-05-23 Luis Machado <luis.machado@linaro.org>
1491 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1492 modifier for printing the step amount.
1494 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1497 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1500 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1502 * doc/sourcebuild.texi (Endianness): New subsubsection.
1504 2018-05-23 Luis Machado <luis.machado@linaro.org>
1506 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1507 <prefetch_dynamic_strides>: New const bool field.
1508 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1509 prefetch_dynamic_strides.
1510 (exynosm1_prefetch_tune): Likewise.
1511 (thunderxt88_prefetch_tune): Likewise.
1512 (thunderx_prefetch_tune): Likewise.
1513 (thunderx2t99_prefetch_tune): Likewise.
1514 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1516 (aarch64_override_options_internal): Update to set
1517 PARAM_PREFETCH_DYNAMIC_STRIDES.
1518 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1519 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1520 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1521 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1522 prefetch-dynamic-strides setting.
1524 2018-05-23 Luis Machado <luis.machado@linaro.org>
1526 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1527 <minimum_stride>: New const int field.
1528 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1529 minimum_stride field defaulting to -1.
1530 (exynosm1_prefetch_tune): Likewise.
1531 (thunderxt88_prefetch_tune): Likewise.
1532 (thunderx_prefetch_tune): Likewise.
1533 (thunderx2t99_prefetch_tune): Likewise.
1534 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1535 <default_opt_level>: Set to 3.
1536 (aarch64_override_options_internal): Update to set
1537 PARAM_PREFETCH_MINIMUM_STRIDE.
1538 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1539 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1540 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1541 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1542 stride is constant and is below the minimum stride threshold.
1544 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1546 * config/arm/arm-cpus.in (mode26): Delete.
1547 (armv4): Delete mode26 reference.
1548 * config/arm/arm.c (arm_configure_build_target): Delete use of
1551 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1553 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1555 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1556 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1558 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1559 (floatunsdidf2): Ditto.
1561 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1563 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1564 (fixuns_trunc<mode>si2_avx512f): Ditto.
1565 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1566 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1567 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1569 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1571 PR rtl-optimization/79985
1572 * df-scan.c (df_insn_refs_collect): Remove special case for
1573 global registers and asm statements.
1575 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1577 * extend.texi (Global Register Variables): Rewrite the bullet list.
1578 Note that the register is available for allocation. Note that access
1579 via inline asm must use constraints. Add note about async-signal
1580 handlers. Remove paragraph about automagic register selection.
1582 2018-05-23 Richard Biener <rguenther@suse.de>
1584 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1585 of fixed offset from memset VN.
1587 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1589 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1591 (alloc_cand_and_find_basis): Initialize first_interp field.
1592 (slsr_process_mul): Modify first_interp field.
1593 (slsr_process_add): Likewise.
1594 (slsr_process_cast): Modify first_interp field for each new
1596 (slsr_process_copy): Likewise.
1597 (dump_candidate): Dump first_interp field.
1598 (replace_mult_candidate): Process all interpretations, not just
1600 (replace_rhs_if_not_dup): Likewise.
1601 (replace_one_candidate): Likewise.
1603 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1605 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1607 (aarch64_needs_frame_chain): New function.
1608 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1610 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1613 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1614 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1615 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1616 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1617 as true for strict-align.
1618 (aarch64_can_inline_p): Perform checks even when callee has no
1619 attributes to check for strict alignment.
1620 * doc/extend.texi (AArch64 Function Attributes): Document
1622 * doc/invoke.texi: (AArch64 Options): Likewise.
1624 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1626 PR tree-optimization/85853
1627 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1628 the handling of the root of the node to...
1629 (vect_slp_analyze_node_operations_1): ...this new function,
1630 and run the whole thing with the child nodes' def types
1631 set according to their SLP node's def type.
1633 2018-05-23 Richard Biener <rguenther@suse.de>
1636 * tree-data-ref.c (create_runtime_alias_checks): Defer
1637 and ignore overflow warnings.
1639 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1641 PR tree-optimization/85822
1642 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1645 2018-05-23 Richard Biener <rguenther@suse.de>
1647 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1648 memset constants via native_interpret_expr.
1650 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1653 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1655 (cgraph_node::create_alias): Likewise.
1656 (cgraph_node::get_availability): Check ifunc_resolver instead
1657 of looking up ifunc attribute.
1658 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1659 * varasm.c (do_assemble_alias): Likewise.
1660 (assemble_alias): Likewise.
1661 (default_binds_local_p_3): Likewise.
1662 * cgraph.h (cgraph_node): Add ifunc_resolver.
1663 (cgraph_node::only_called_directly_or_aliased_p): Return false
1665 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1667 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1668 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1669 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1670 instead of looking up ifunc attribute.
1672 2018-05-22 Luis Machado <luis.machado@linaro.org>
1674 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1676 2018-05-22 Martin Sebor <msebor@redhat.com>
1679 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1680 only when expasion succeeds.
1681 (expand_builtin_strcmp): Same.
1682 (expand_builtin_strncmp): Same.
1684 2018-05-22 Martin Sebor <msebor@redhat.com>
1686 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1688 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1689 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1691 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1692 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1694 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1695 Add check for legitimate_address.
1696 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1697 (aarch64_swap_ldrstr_operands): New.
1698 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1701 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1702 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1704 * config/aarch64/aarch64.md: New patterns to generate stp
1706 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1707 single words and double words.
1708 (load_pair_sw, load_pair_dw): Likewise.
1709 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1711 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1713 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1714 for different mode ldpstp and add peephole for merged zero stores.
1716 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1718 (aarch64_gen_store_pair): Rename calls to match new patterns.
1719 (aarch64_gen_load_pair): Rename calls to match new patterns.
1720 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1721 (load_pair<DREG:mode><DREG2:mode>): ... This.
1722 (store_pair<mode>): Rename to...
1723 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1724 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1726 (V_INT_EQUIV): Handle SImode.
1727 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1730 2018-05-22 Martin Sebor <msebor@redhat.com>
1733 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1734 or ajust the presumed bound on an operation to avoid unnecessary
1737 2018-05-22 Martin Sebor <msebor@redhat.com>
1739 PR tree-optimization/85826
1740 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1741 assuming that a DECL necesarily has a constant size.
1743 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1746 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1748 2018-05-22 Richard Biener <rguenther@suse.de>
1750 PR tree-optimization/85834
1751 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1752 non-constant and non-zero memset arguments.
1754 2018-05-22 Martin Liska <mliska@suse.cz>
1757 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1759 2018-05-22 Richard Biener <rguenther@suse.de>
1761 PR tree-optimization/85863
1762 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1763 comparisons when vectype is specified.
1764 (vectorizable_condition): Do not specify vectype for
1765 vect_is_simple_cond when SLP vectorizing.
1767 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1770 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1771 define __ibm128 as long double.
1772 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1773 as a distinct type when IEEE 128-bit support is enabled.
1774 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1775 128-bit types to use the correct functions.
1776 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1777 convert between 128-bit floating point types that have different
1778 modes but the same representation, instead of using gen_lowpart to
1780 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1782 (IFKF_reg): New attributes to give the register constraints for
1784 (extend<mode>tf2_internal): New insns to mark an explicit
1785 conversion between 128-bit floating point types that have a
1786 different mode but share the same representation.
1788 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1790 PR tree-optimization/85814
1791 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1792 a null return from get_strinfo when unsharing the next
1793 strinfo in the chain.
1795 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1798 * varasm.c (weak_finish): Clean up weak_decls.
1800 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1802 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1803 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1804 UNSPEC_UADALP values.
1805 * config/aarch64/iterators.md (ABAL): New int iterator.
1808 (sur): Add mappings for the above.
1809 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1811 (aarch64_<sur>abal<mode>_4): Likewise.
1812 (aarch64_<sur>adalp<mode>_3): Likewise.
1813 (<sur>sadv16qi): New define_expand.
1815 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1817 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1818 (*movdf_internal): Ditto.
1819 (*rcpsf2_sse): Ditto.
1820 (*rsqrtsf2_sse): Ditto.
1821 (*sqrt<mode>2_sse): Ditto.
1823 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1825 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1827 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1828 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1829 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1830 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1832 * config/aarch64/arm_neon.h: Likewise.
1833 * config/aarch64/iterators.md (VQ_I): New.
1835 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1837 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1839 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1842 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1844 * config/nds32/constraints.md (S): New constraint.
1845 * config/nds32/nds32.md (call_internal): Use constraint S.
1846 (call_value_internal): Likewise.
1847 (sibcall_internal): Likewise.
1848 (sibcall_value_internal): Likewise.
1850 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1851 Chung-Ju Wu <jasonwucj@gmail.com>
1853 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1856 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1857 Chung-Ju Wu <jasonwucj@gmail.com>
1859 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1860 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1861 (nds32_rtx_costs_impl): Simplify.
1862 (nds32_address_cost_impl): Simplify.
1863 (nds32_init_rtx_costs): New function.
1864 (nds32_rtx_costs_speed_prefer): Likewise.
1865 (nds32_rtx_costs_size_prefer): Likewise.
1866 (nds32_address_cost_speed_prefer): Likewise.
1867 (nds32_address_cost_speed_fwprop): Likewise.
1868 (nds32_address_cost_size_prefer): Likewise.
1869 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1870 * config/nds32/nds32.c (nds32_option_override): Use
1871 nds32_init_rtx_costs function.
1873 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1875 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1876 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1877 (TARGET_PIPELINE_N8): Likewise.
1878 (TARGET_PIPELINE_N10): Likewise.
1879 (TARGET_PIPELINE_N13): Likewise.
1880 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1882 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1884 * config/nds32/nds32-fpu.md: Update copyright year.
1886 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1888 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1890 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1892 * config/nds32/nds32.c
1893 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1894 * config/nds32/nds32.opt (minline-asm-r15): New option.
1896 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1898 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1900 * config/nds32/nds32.md (abssi2): New pattern.
1902 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1904 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1905 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1906 Merge insn pattern from sse_cvtsi2ss<round_name> and
1907 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1908 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1909 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1910 using SWI48 mode iterator.
1911 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1912 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1913 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1914 pattern from sse_cvttss2si<round_saeonly_name>
1915 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1916 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1917 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1918 using SWI48 mode iterator.
1919 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1920 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1921 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1922 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1923 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1924 using SWI48 mode iterator.
1925 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1926 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1927 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1928 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1929 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1930 SWI48 mode iterator.
1931 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1932 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1933 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1934 pattern from sse_cvttsd2si<round_saeonly_name>
1935 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1937 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1939 * config/nds32/nds32-md-auxiliary.c
1940 (nds32_valid_smw_lwm_base_p): Refine.
1941 (nds32_output_smw_single_word): Refine.
1942 (nds32_output_smw_double_word): New.
1943 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1945 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1947 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1948 (nds32_output_stack_pop): Refine.
1949 (nds32_expand_unaligned_load): Refine.
1950 (nds32_expand_unaligned_store): Refine.
1952 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1953 Chung-Ju Wu <jasonwucj@gmail.com>
1955 * config/nds32/constants.md: Add TP_REGNUM constant.
1956 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1957 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1959 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1960 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1961 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1962 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1963 and PIC code generation.
1964 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1966 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1968 * config/nds32/nds32.md: Support TLS and PIC.
1969 * config/nds32/nds32.c: Support TLS and PIC.
1970 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1971 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1974 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1976 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1977 mode with E_ prefix.
1979 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1980 Chung-Ju Wu <jasonwucj@gmail.com>
1982 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1983 * config/nds32/nds32-md-auxiliary.c
1984 (symbolic_reference_mentioned_p): New.
1985 (nds32_legitimize_ict_address): New.
1986 (nds32_expand_ict_move): New.
1987 (nds32_indirect_call_referenced_p): New.
1988 (nds32_symbol_binds_local_p): Delete.
1989 (nds32_long_call_p): Modify.
1990 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1991 * config/nds32/nds32-protos.h
1992 (symbolic_reference_mentioned_p): Declare.
1993 (nds32_legitimize_ict_address): Declare.
1994 (nds32_expand_ict_move): Declare.
1995 (nds32_indirect_call_referenced_p): Declare.
1996 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1997 (nds32_relax_group): Use nds32_ict_const_p as condition.
1998 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1999 (nds32_asm_file_start): Output ict_model directive in asm code.
2000 (nds32_legitimate_address_p): Consider indirect call.
2001 (nds32_print_operand): Consider indirect call.
2002 (nds32_print_operand_address): Consider indirect call.
2003 (nds32_insert_attributes): Handle "indirect_call" attribute.
2004 (TARGET_LEGITIMATE_ADDRESS_P): Define.
2005 (TARGET_LEGITIMATE_CONSTANT_P): Define.
2006 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
2007 (TARGET_DELEGITIMIZE_ADDRESS): Define.
2008 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
2009 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
2010 (TARGET_ICT_MODEL_SMALL): Define.
2011 (TARGET_ICT_MODEL_LARGE): Define.
2012 * config/nds32/nds32.md (movsi): Consider ict model.
2013 (call, call_value): Consider ict model.
2014 (sibcall, sibcall_value): Consider ict model.
2015 * config/nds32/nds32.opt (mict-model): New option.
2016 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
2019 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
2020 Monk Chiang <sh.chiang04@gmail.com>
2021 Jim Wilson <jimw@sifive.com>
2023 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
2024 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
2025 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
2026 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
2027 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
2028 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
2029 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
2030 compute save_libcall_adjustment properly.
2031 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
2032 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
2033 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
2034 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
2035 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
2036 (ABI_SPEC): Handle mabi=ilp32e.
2037 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
2038 (RVE): Add RVE mask.
2039 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
2040 <-march>: Add rv32e as an example.
2042 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
2045 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
2046 (intra_create_variable_infos): Handle C++ constructors.
2048 2018-05-18 Martin Liska <mliska@suse.cz>
2050 * passes.def: Remove a redundant pass.
2052 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
2055 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
2057 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2059 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
2061 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
2063 (armv2, armv2a, armv3, armv3m): Delete architectures.
2064 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
2065 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
2066 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
2068 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
2069 (*mulsidi3adddi): Likewise.
2070 (mulsidi3): Likewise.
2071 (*mulsidi3_nov6): Likewise.
2072 (umulsidi3): Likewise.
2073 (umulsidi3_nov6): Likewise.
2074 (umaddsidi4): Likewise.
2075 (*umulsidi3adddi): Likewise.
2076 (smulsi3_highpart): Likewise.
2077 (*smulsi3_highpart_nov6): Likewise.
2078 (umulsi3_highpart): Likewise.
2079 (*umulsi3_highpart_nov6): Likewise.
2080 * config/arm/arm.h (arm_arch3m): Delete.
2081 * config/arm/arm.c (arm_arch3m): Delete.
2082 (arm_option_override_internal): Update armv3-related comment.
2083 (arm_configure_build_target): Delete use of isa_bit_mode32.
2084 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
2085 (arm_rtx_costs_internal): Delete check of arm_arch3m.
2086 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
2088 (mulusa3): Likewise.
2089 * config/arm/arm-protos.h (arm_arch3m): Delete.
2090 * config/arm/arm-tables.opt: Regenerate.
2091 * config/arm/arm-tune.md: Likewise.
2092 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
2093 deleted architectures.
2095 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2097 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
2098 (armv5t, armv5te): New features.
2099 (ARMv5, ARMv5e): Delete fgroups.
2100 (ARMv5t, ARMv5te): Adjust for above changes.
2102 (armv5, armv5e): Delete arches.
2103 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
2105 (*call_reg_arm): Likewise.
2106 (*call_value_reg_armv5): Likewise.
2107 (*call_value_reg_arm): Likewise.
2108 (*call_symbol): Likewise.
2109 (*call_value_symbol): Likewise.
2110 (*sibcall_insn): Likewise.
2111 (*sibcall_value_insn): Likewise.
2113 (prefetch): Likewise.
2114 (define_split and define_peephole2 dependent on arm_arch5):
2116 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
2118 (TARGET_ARM_QBIT): Likewise.
2119 (TARGET_DSP_MULTIPLY): Likewise.
2120 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
2121 (arm_arch5, arm_arch5e): Delete.
2122 (arm_arch5t, arm_arch5te): Declare.
2123 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
2124 (arm_arch5t): Declare.
2125 (arm_option_reconfigure_globals): Update for the above.
2126 (arm_options_perform_arch_sanity_checks): Update comment, replace
2127 use of arm_arch5 with arm_arch5t.
2128 (use_return_insn): Likewise.
2129 (arm_emit_call_insn): Likewise.
2130 (output_return_instruction): Likewise.
2131 (arm_final_prescan_insn): Likewise.
2132 (arm_coproc_builtin_available): Likewise.
2133 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2134 arm_arch5e with arm_arch5t and arm_arch5te.
2135 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2136 (arm_arch5t, arm_arch5te): Declare.
2137 * config/arm/arm-tables.opt: Regenerate.
2138 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2139 * config/arm/t-multilib: Likewise.
2140 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2141 instead of arm_arch5.
2142 (*call_reg_thumb1): Likewise.
2143 (*call_value_reg_thumb1_v5): Likewise.
2144 (*call_value_reg_thumb1): Likewise.
2145 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2147 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2149 2018-05-18 Martin Liska <mliska@suse.cz>
2151 PR gcov-profile/84846
2152 * doc/gcov.texi: Document -t option of gcov tool.
2154 2018-05-18 Martin Liska <mliska@suse.cz>
2156 PR gcov-profile/84846
2157 * gcov.c (print_usage): Add new -t option.
2158 (process_args): Handle the option.
2159 (generate_results): Use stdout as output when requested by
2162 2018-05-18 Martin Liska <mliska@suse.cz>
2164 PR gcov-profile/84846
2165 * coverage.c (coverage_init): Write PWD to .gcno file.
2166 * doc/gcov.texi: Document how working directory is printed.
2167 * gcov-dump.c (dump_gcov_file): Print PWD.
2168 * gcov.c (output_intermediate_file): Likewise.
2169 (read_graph_file): Read PWD string.
2170 (output_lines): Print PWD.
2172 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2175 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2176 for retval and return false if all args to phi are zero.
2178 2018-05-18 Richard Biener <rguenther@suse.de>
2180 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2182 (evrp_dom_walker::before_dom_children): Call it.
2184 2018-05-18 Richard Biener <rguenther@suse.de>
2186 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2187 results when processing array refs with variable index.
2189 2018-05-18 Toon Moene <toon@moene.org>
2191 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2192 directly after that of -floop-interchange. Indicate that both
2193 options are enabled by default when specifying -O3.
2195 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2197 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2198 iterator. Delete separate integer-mode vec_set<mode> expander.
2199 (aarch64_simd_vec_setv2di): Delete.
2200 (vec_setv2di): Delete.
2201 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2202 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2203 the "w, r" alternative.
2205 2018-05-18 Martin Liska <mliska@suse.cz>
2207 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2208 * tree-pass.h (make_pass_lower_switch_O0): New function.
2209 * tree-switch-conversion.c (node_has_low_bound): Remove.
2210 (node_has_high_bound): Likewise.
2211 (node_is_bounded): Likewise.
2212 (class pass_lower_switch): Make it a template type and create
2214 (pass_lower_switch::execute): Add template argument.
2215 (make_pass_lower_switch): New function.
2216 (make_pass_lower_switch_O0): New function.
2217 (do_jump_if_equal): Remove.
2218 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2219 all the hard work to tree optimization passes.
2221 2018-05-18 Martin Liska <mliska@suse.cz>
2223 * dbgcnt.c (limit_low): Renamed from limit.
2224 (limit_high): New variable.
2225 (dbg_cnt_is_enabled): Check for upper limit.
2226 (dbg_cnt): Adjust dumping.
2227 (dbg_cnt_set_limit_by_index): Add new argument for high
2229 (dbg_cnt_set_limit_by_name): Likewise.
2230 (dbg_cnt_process_single_pair): Parse new format.
2231 (dbg_cnt_process_opt): Use strtok.
2232 (dbg_cnt_list_all_counters): Remove 'value' and add
2234 * doc/invoke.texi: Document changes.
2236 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2238 * doc/sourcebuild.texi (scalar_all_fma): Document.
2239 * tree.def (FMA_EXPR): Delete.
2240 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2241 * internal-fn.c (ternary_direct): New macro.
2242 (expand_ternary_optab_fn): Likewise.
2243 (direct_ternary_optab_supported_p): Likewise.
2244 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2245 * builtins.c (fold_builtin_fma): Delete.
2246 (fold_builtin_3): Don't call it.
2247 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2248 * expr.c (expand_expr_real_2): Likewise.
2249 * fold-const.c (operand_equal_p): Likewise.
2250 (fold_ternary_loc): Likewise.
2251 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2252 * gimple.c (DEFTREECODE): Likewise.
2253 * gimplify.c (gimplify_expr): Likewise.
2254 * optabs-tree.c (optab_for_tree_code): Likewise.
2255 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2256 * tree-eh.c (operation_could_trap_p): Likewise.
2257 (stmt_could_throw_1_p): Likewise.
2258 * tree-inline.c (estimate_operator_cost): Likewise.
2259 * tree-pretty-print.c (dump_generic_node): Likewise.
2260 (op_code_prio): Likewise.
2261 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2262 * tree-ssa-operands.c (get_expr_operands): Likewise.
2263 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2264 * fold-const-call.h (fold_fma): Delete.
2265 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2266 CFN_FNMA and CFN_FNMS.
2268 * genmatch.c (combined_fn): New enum.
2269 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2270 (commutative_op): New function.
2271 (commutate): Use it. Handle more than 2 operands.
2272 (dt_operand::gen_gimple_expr): Use commutative_op.
2273 (parser::parse_expr): Allow :c to be used with non-binary
2274 operators if the commutative operand is known.
2275 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2276 CFN_FMS, CFN_FNMA and CFN_FNMS.
2277 (backprop::process_assign_use): Remove FMA_EXPR handling.
2278 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2279 (gen_hsa_fma): New function.
2280 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2281 IFN_FNMA and IFN_FNMS.
2282 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2283 * gimple-fold.h (follow_all_ssa_edges): Declare.
2284 * gimple-fold.c (follow_all_ssa_edges): New function.
2285 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2286 gimple_build interface and use follow_all_ssa_edges to fold the result.
2287 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2288 instead of checking for optabs directly.
2289 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2290 rather than FMA_EXPRs.
2291 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2292 call to IFN_FMA instead of an FMA_EXPR.
2294 2018-05-17 Jim Wilson <jimw@sifive.com>
2296 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2297 sign extended promoted subreg, and the range does not have the sign bit
2298 set, then do a sign extend.
2300 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2301 test, check for sign extended subreg and/or constant operands, and
2302 do a sign extend in that case.
2304 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2306 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2307 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2309 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2310 Change logics_shift_reg to logics_shift_imm.
2311 (thunderx2t99_fp_loadpair_basic): Delete.
2312 (thunderx2t99_fp_storepair_basic): Delete.
2313 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2314 (thunderx2t99_asimd_polynomial): Delete.
2315 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2316 and neon_fp_mul_d_scalar_q.
2317 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2318 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2319 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2320 (thunderx2t99_asimd_lut): Add missing tbl types.
2321 (thunderx2t99_asimd_ext): Delete.
2322 (thunderx2t99_asimd_load1_1_mult): Delete.
2323 (thunderx2t99_asimd_load1_2_mult): Delete.
2324 (thunderx2t99_asimd_load1_ldp): New.
2325 (thunderx2t99_asimd_load1): New.
2326 (thunderx2t99_asimd_load2): Add missing *load2* types.
2327 (thunderx2t99_asimd_load3): New.
2328 (thunderx2t99_asimd_load4): New.
2329 (thunderx2t99_asimd_store1_1_mult): Delete.
2330 (thunderx2t99_asimd_store1_2_mult): Delete.
2331 (thunderx2t99_asimd_store2_mult): Delete.
2332 (thunderx2t99_asimd_store2_onelane): Delete.
2333 (thunderx2t99_asimd_store_stp): New.
2334 (thunderx2t99_asimd_store1): New.
2335 (thunderx2t99_asimd_store2): New.
2336 (thunderx2t99_asimd_store3): New.
2337 (thunderx2t99_asimd_store4): New.
2339 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2341 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2342 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2344 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2345 Segher Boessenkool <segher@kernel.crashing.org>
2348 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2351 2018-05-17 Richard Biener <rguenther@suse.de>
2353 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2354 for pruning loop and prune defs feeding only already visited PHIs.
2356 2018-05-17 Richard Biener <rguenther@suse.de>
2358 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2360 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2361 Richard Biener <rguenther@suse.de>
2363 PR tree-optimization/85793
2364 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2365 for VMAT_ELEMENTWISE.
2367 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2369 * internal-fn.h (lookup_internal_fn): Declare
2370 * internal-fn.c (lookup_internal_fn): New function.
2371 * gimple.c (gimple_build_call_from_tree): Handle calls to
2373 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2374 internal function names.
2375 * tree-pretty-print.c (dump_generic_node): Likewise.
2376 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2378 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2380 * gimple-fold.h (gimple_build): Make the function forms take
2381 combined_fn rather than built_in_function.
2382 (gimple_simplify): Likewise.
2383 * gimple-match-head.c (gimple_simplify): Likewise.
2384 * gimple-fold.c (gimple_build): Likewise.
2385 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2386 rather than gimple_build_call_internal.
2387 (get_initial_defs_for_reduction): Likewise.
2388 (vect_create_epilog_for_reduction): Likewise.
2389 (vectorizable_live_operation): Likewise.
2391 2018-05-17 Martin Liska <mliska@suse.cz>
2393 * gimple-ssa-sprintf.c (format_directive): Do not use
2394 space in between 'G_' and '('.
2396 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2399 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2400 even if the mask is not all ones.
2403 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2405 (ix86_gimple_fold_builtin): Likewise.
2408 * config/i386/i386.c: Include tree-vector-builder.h.
2409 (ix86_vector_shift_count): New function.
2410 (ix86_fold_builtin): Fold shift builtins by scalar count.
2411 (ix86_gimple_fold_builtin): Likewise.
2413 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2414 _mm512_setzero): New intrinsics.
2416 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2417 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2419 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2420 code generation for cases where splatting a value is not useful.
2421 * simplify-rtx.c (simplify_ternary_operation): Simplify
2422 vec_merge across a vec_duplicate and a paradoxical subreg forming
2423 a vector mode to a vec_concat.
2425 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2427 * config.gcc: Support "goldmont-plus".
2428 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2430 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2431 PROCESSOR_GOLDMONT_PLUS.
2432 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2433 (processor_target_table): Add "goldmont-plus".
2434 (PTA_GOLDMONT_PLUS): Define.
2435 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2436 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2437 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2438 (fold_builtin_cpu): Add "goldmont-plus".
2439 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2440 (ix86_option_override_internal): Add "goldmont-plus".
2441 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2442 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2443 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2444 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2446 2018-05-17 Richard Biener <rguenther@suse.de>
2448 PR tree-optimization/85757
2449 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2450 remove defs that only feed that PHI from further processing.
2452 2018-05-16 Jim Wilson <jimw@sifive.com>
2454 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2456 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2458 2018-05-16 Mark Wielaard <mark@klomp.org>
2460 * dwarf2out.c (count_index_strings): New function.
2461 (output_indirect_strings): Call count_index_strings and generate
2462 header for dwarf_version >= 5.
2464 2018-05-16 Mark Wielaard <mark@klomp.org>
2466 * dwarf2out.c (dwarf_FORM): New function.
2467 (set_indirect_string): Use dwarf_FORM.
2468 (reset_indirect_string): Likewise.
2469 (size_of_die): Likewise.
2470 (value_format): Likewise.
2471 (output_die): Likewise.
2472 (add_skeleton_AT_string): Likewise.
2473 (output_macinfo_op): Likewise.
2474 (index_string): Likewise.
2475 (output_index_string_offset): Likewise.
2476 (output_index_string): Likewise.
2477 (count_index_strings): Likewise.
2479 2018-05-16 Carl Love <cel@us.ibm.com>
2481 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2482 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2484 2018-05-16 Martin Jambor <mjambor@suse.cz>
2486 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2487 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2489 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2491 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2492 (fnma<mode>4): Likewise.
2493 (fms<mode>4): Likewise.
2494 (fnms<mode>4): Likewise.
2495 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2496 (aarch64_fnma<mode>4): Likewise.
2497 (aarch64_fms<mode>4): Likewise.
2498 (aarch64_fnms<mode>4): Likewise.
2499 (aarch64_fnmadd<mode>4): Likewise.
2501 2018-05-16 Jason Merrill <jason@redhat.com>
2503 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2505 2018-05-16 Richard Biener <rguenther@suse.de>
2507 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2508 (dump_stmt_cost): Declare.
2509 (add_stmt_cost): Dump cost we add.
2510 (add_stmt_costs): New function.
2511 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2513 (vect_analyze_stmt): Adjust prototype.
2514 (vectorizable_condition): Likewise.
2515 (vectorizable_live_operation): Likewise.
2516 (vectorizable_reduction): Likewise.
2517 (vectorizable_induction): Likewise.
2518 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2519 cost vector to pass to vectorizable_ and record afterwards.
2520 (vect_model_reduction_cost): Take cost vector argument and adjust.
2521 (vect_model_induction_cost): Likewise.
2522 (vectorizable_reduction): Likewise.
2523 (vectorizable_induction): Likewise.
2524 (vectorizable_live_operation): Likewise.
2525 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2526 SLP_TREE_NUMBER_OF_VEC_STMTS.
2527 (vect_analyze_slp_cost_1): Remove.
2528 (vect_analyze_slp_cost): Likewise.
2529 (vect_slp_analyze_node_operations): Take visited args and
2530 a target cost vector. Avoid processing already visited stmt sets.
2531 (vect_slp_analyze_operations): Use a local cost vector to gather
2532 costs and register those of non-discarded instances.
2533 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2534 (vect_schedule_slp_instance): Remove copying of
2535 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2537 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2538 adding cost. Record cost entry location.
2539 (vect_prologue_cost_for_slp_op): Function to compute cost of
2540 a constant or invariant generated for SLP vect in the prologue,
2541 split out from vect_analyze_slp_cost_1.
2542 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2543 (vect_model_promotion_demotion_cost): Likewise.
2544 (vect_model_store_cost): Likewise, make static.
2545 (vect_model_load_cost): Likewise.
2546 (vectorizable_bswap): Add cost vector arg and adjust.
2547 (vectorizable_call): Likewise.
2548 (vectorizable_simd_clone_call): Likewise.
2549 (vectorizable_conversion): Likewise.
2550 (vectorizable_assignment): Likewise.
2551 (vectorizable_shift): Likewise.
2552 (vectorizable_operation): Likewise.
2553 (vectorizable_store): Likewise.
2554 (vectorizable_load): Likewise.
2555 (vectorizable_condition): Likewise.
2556 (vectorizable_comparison): Likewise.
2557 (can_vectorize_live_stmts): Likewise.
2558 (vect_analyze_stmt): Likewise.
2559 (vect_transform_stmt): Adjust calls to vectorizable_*.
2560 * tree-vectorizer.c: Include gimple-pretty-print.h.
2561 (dump_stmt_cost): New function.
2563 2018-05-16 Richard Biener <rguenther@suse.de>
2565 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2566 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2567 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2568 (check_name): New callback.
2569 (dse_classify_store): Track cycles via a visited bitmap of PHI
2570 defs and simplify handling of in-loop and across loop dead stores
2571 and properly fail for loop-variant refs. Handle byte-tracking with
2572 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2575 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2577 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2578 (vect_get_mask_type_for_stmt): Likewise.
2579 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2581 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2582 to determine the statement's vector type and the vector type that
2583 should be used for calculating nunits. Deal with cases in which
2584 the type has to be deferred.
2585 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2586 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2587 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2588 (vect_determine_vf_for_stmt): New functions, split out from...
2589 (vect_determine_vectorization_factor): ...here.
2590 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2591 (vect_get_mask_type_for_stmt): New functions, split out from
2592 vect_determine_vectorization_factor.
2594 2018-05-16 Richard Biener <rguenther@suse.de>
2596 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2597 verify the [VEC_]COND_EXPR embedded comparison.
2599 2018-05-15 Martin Sebor <msebor@redhat.com>
2601 PR tree-optimization/85753
2602 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2603 RECORD_TYPE in addition to ARRAY_TYPE.
2605 2018-05-15 Martin Sebor <msebor@redhat.com>
2608 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2610 2018-05-15 Richard Biener <rguenther@suse.de>
2612 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2613 add by_clobber_p one. Change algorithm to collect all defs
2614 representing uses we need to walk and try reducing them to
2615 a single one before failing.
2616 (dse_dom_walker::dse_optimize_stmt): Adjust.
2618 2018-05-13 Mark Wielaard <mark@klomp.org>
2620 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2621 (size_of_loc_descr): Likewise.
2622 (output_loc_operands): Likewise.
2623 (output_loc_operands_raw): Likewise.
2624 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2625 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2626 (hash_loc_operands): Likewise.
2627 (compare_loc_operands): Likewise.
2629 2018-05-14 Mark Wielaard <mark@klomp.org>
2631 * dwarf2out.c (count_index_addrs): New function.
2632 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2634 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2636 PR tree-optimization/83648
2637 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2638 return value as malloc candidate.
2640 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2643 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2644 param as true in call to suggest_attribute.
2646 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2648 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2649 -mreadonly-in-sdata.
2651 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2653 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2655 (aarch64_crypto_aesd_fused): Likewise.
2657 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2659 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2660 (movsi_aarch64): Likewise.
2661 (load_pairsi): Likewise.
2662 (load_pairdi): Likewise.
2663 (store_pairsi): Likewise.
2664 (store_pairdi): Likewise.
2665 (load_pairsf): Likewise.
2666 (load_pairdf): Likewise.
2667 (store_pairsf): Likewise.
2668 (store_pairdf): Likewise.
2669 (zero_extend): Likewise.
2670 (trunc): Swap alternatives.
2671 (fcvt_target): Add '?' to prefer w over r.
2673 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2676 * config/i386/i386.md: Disallow non-commutative arithmetics in
2677 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2678 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2679 in the peephole2 before it.
2681 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2683 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2684 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2685 (ix86_handle_option): Handle -mcldemote.
2686 * config.gcc: New header.
2687 * config/i386/cldemoteintrin.h: New file.
2688 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2689 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2691 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2692 OPTION_MASK_ISA_CLDEMOTE.
2693 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2694 (ix86_valid_target_attribute_inner_p): Ditto.
2695 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2696 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2697 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2698 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2699 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2701 * config/i386/i386.opt: Add -mcldemote.
2702 * config/i386/x86intrin.h: New header.
2703 * doc/invoke.texi: Add -mcldemote.
2705 2018-05-14 Richard Biener <rguenther@suse.de>
2707 * doc/match-and-simplify.texi: Adjust :s documentation.
2709 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2711 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2712 intended memcpy size.
2713 (REORDER_45): Likewise.
2715 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2717 * sort.cc: New file.
2718 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2719 * vec.c (qsort_chk): Use gcc_qsort.
2720 * Makefile.in (OBJS-libcommon): Add sort.o.
2721 (build/sort.o): New target. Use it...
2722 (BUILD_RTL): ... here, and...
2723 (build/gencfn-macros): ... here, and...
2724 (build/genmatch): ... here.
2726 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2727 Chung-Ju Wu <jasonwucj@gmail.com>
2729 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2730 * config/nds32/nds32-graywolf.md: New file.
2731 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2732 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2734 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2735 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2736 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2737 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2738 * config/nds32/pipelines.md: Include n15 settings.
2740 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2741 Chung-Ju Wu <jasonwucj@gmail.com>
2743 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2744 * config/nds32/nds32-n13.md: New file.
2745 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2746 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2748 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2749 * config/nds32/nds32.md (pipeline_model): Add n13.
2750 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2751 * config/nds32/pipelines.md: Include n13 settings.
2753 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2754 Chung-Ju Wu <jasonwucj@gmail.com>
2756 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2757 * config/nds32/nds32-n10.md: New file.
2758 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2759 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2761 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2762 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2763 * config/nds32/nds32.md (pipeline_model): Add n10.
2764 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2765 * config/nds32/pipelines.md: Include n10 settings.
2767 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2768 Kito Cheng <kito.cheng@gmail.com>
2769 Chung-Ju Wu <jasonwucj@gmail.com>
2771 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2772 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2773 Add enum values for DSP extension instructions.
2774 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2776 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2777 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2779 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2780 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2781 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2782 * config/nds32/nds32-intrinsic.md: Likewise.
2783 * config/nds32/nds32_intrinsic.h: Likewise.
2784 * config/nds32/nds32-md-auxiliary.c: Likewise.
2785 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2786 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2787 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2788 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2789 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2790 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2791 TYPE_DMAC in switch statement.
2792 * config/nds32/nds32.c: New checking and implementation for DSP
2793 extension instructions.
2794 * config/nds32/nds32.h: Likewise.
2795 * config/nds32/nds32.md: Likewise.
2796 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2797 * config/nds32/predicates.md: Implement new predicates for DSP
2800 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2802 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2803 Reformat alternatives and attributes so it is easier to identify
2804 which constraints/attributes go with which instruction.
2805 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2806 (mov<mode>_softfloat32, FMOVE64): Likewise.
2807 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2808 (mov<mode>_softfloat64, FMOVE64): Likewise.
2810 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2812 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2814 (Basic PowerPC Built-in Functions): The new name of the
2815 subsection previously known as "PowerPC Built-in Functions".
2816 (Basic PowerPC Built-in Functions Available on all Configurations):
2818 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2819 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2820 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2821 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2823 2018-05-11 Martin Jambor <mjambor@suse.cz>
2826 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2829 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2832 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2834 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2836 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2837 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2838 (ix86_handle_option): Handle -mwaitpkg.
2839 * config.gcc: New header.
2840 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2841 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2842 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2844 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2845 OPTION_MASK_ISA_WAITPKG.
2846 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2847 (ix86_option_override_internal): Add PTA_WAITPKG.
2848 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2849 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2850 IX86_BUILTIN_TPAUSE.
2851 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2852 __builtin_ia32_umwait and __builtin_ia32_tpause.
2853 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2854 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2855 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2856 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2857 UNSPECV_TPAUSE): New.
2858 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2859 * config/i386/i386.opt: Add -mwaitpkg.
2860 * config/i386/waitpkgintrin.h: New file.
2861 * config/i386/x86intrin.h: New header.
2862 * doc/invoke.texi: Add -mwaitpkg.
2864 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2867 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2869 (cortex-m0): Use armv6s-m isa.
2870 (cortex-m0plus): Likewise.
2871 (cortex-m1): Likewise.
2872 (cortex-m0.small-multiply): Likewise.
2873 (cortex-m0plus.small-multiply): Likewise.
2874 (cortex-m1.small-multiply): Likewise.
2876 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2877 Jakub Jelinek <jakub@redhat.com>
2879 PR tree-optimization/85692
2880 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2881 source permute as well.
2883 2018-05-11 Martin Liska <mliska@suse.cz>
2886 * doc/extend.texi: Document LLVM style format for no_sanitize
2889 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2891 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2892 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2893 (mode_supports_vsx_dform_quad): Likewise.
2894 (mode_supports_vmx_dform): Move these functions to be next to the
2895 other mode_supports functions.
2896 (mode_supports_dq_form): Likewise.
2897 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2898 mode_supports_dq_form.
2899 (reg_offset_addressing_ok_p): Likewise.
2900 (offsettable_ok_by_alignment): Likewise.
2901 (rs6000_legitimate_offset_address_p): Likewise.
2902 (legitimate_lo_sum_address_p): Likewise.
2903 (rs6000_legitimize_address): Likewise.
2904 (rs6000_legitimize_reload_address): Likewise.
2905 (rs6000_secondary_reload_inner): Likewise.
2906 (rs6000_preferred_reload_class): Likewise.
2907 (rs6000_output_move_128bit): Likewise.
2909 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2911 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2912 Generate SImode target register for null target.
2913 <case IX86_BUILTIN_XGETBV>: Ditto.
2914 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2915 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2917 2018-05-10 Carl Love <cel@us.ibm.com>
2919 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2920 dcbtt and dcbtstt if operands[2] is 0.
2922 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2925 * config/i386/sse.md (usadv64qi): New expander.
2927 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2929 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2930 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2931 -maltivec=be support.
2932 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2933 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2934 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2935 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2936 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2937 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2938 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2939 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2940 altivec_vsumsws): Adjust.
2941 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2942 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2943 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2945 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2946 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2947 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2948 (altivec_lve<VI_char>x): Delete expand.
2949 (*altivec_lve<VI_char>x_internal): Rename to...
2950 (altivec_lve<VI_char>x): ... this.
2951 (altivec_lvxl_<mode>): Delete expand.
2952 (*altivec_lvxl_<mode>_internal): Rename to ...
2953 (altivec_lvxl_<mode>): ... this.
2954 (altivec_stvxl_<mode>): Delete expand.
2955 (*altivec_stvxl_<mode>_internal): Rename to ...
2956 (altivec_stvxl_<mode>): ... this.
2957 (altivec_stve<VI_char>x): Delete expand.
2958 (*altivec_stve<VI_char>x_internal): Rename to ...
2959 (altivec_stve<VI_char>x): ... this.
2960 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2961 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2962 reduc_plus_scal_<mode>): Adjust.
2963 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2965 (rs6000_cpu_cpp_builtins): Adjust.
2966 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2967 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2968 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2969 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2970 -maltivec=be support.
2971 (rs6000_split_vec_extract_var): Adjust.
2972 (rs6000_split_v4si_init): Adjust.
2973 (swap_selector_for_mode): Delete.
2974 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2975 altivec_expand_stvex_be): Delete.
2976 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2977 -maltivec=be support.
2978 (rs6000_gimple_fold_builtin): Ditto.
2979 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2981 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2982 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2983 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2984 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2985 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2986 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2987 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2988 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2989 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2990 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2991 anonymous split): Adjust.
2992 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2993 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2995 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2997 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2998 when --with-gxx-include-dir is also specified.
2999 * configure: Regenerate.
3001 2018-05-09 Jim Wilson <jimw@sifive.com>
3004 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
3005 * config/riscv/t-withmultilib: New.
3006 * config/riscv/withmultilib.h: New.
3007 * doc/install.texi: Document RISC-V --with-multilib-list support.
3009 2018-05-09 Richard Biener <rguenther@suse.de>
3011 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
3013 (vect_bb_vectorization_profitable_p): Adjust. Compute
3014 actual scalar cost using the cost vector and the add_stmt_cost
3017 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3019 PR rtl-optimization/85645
3020 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
3021 in the REG_CFA_REGISTER note for LR, don't leave it empty.
3023 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3025 PR rtl-optimization/85645
3026 * shrink-wrap.c (spread_components): Return a boolean saying if
3027 anything was changed.
3028 (try_shrink_wrapping_separate): Iterate spread_components until
3029 nothing changes anymore.
3031 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3033 PR rtl-optimization/85645
3034 * regrename.c (build_def_use): Also kill the chains that include the
3035 destination of a REG_CFA_REGISTER note.
3037 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
3039 PR rtl-optimization/85645
3040 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3041 insn that has a REG_CFA_REGISTER note.
3043 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
3045 * cfgexpand.c (expand_clobber): New function.
3046 (expand_gimple_stmt_1): Use it.
3047 * tree-vect-stmts.c (vect_clobber_variable): New function,
3049 (vectorizable_simd_clone_call): ...here.
3050 (vectorizable_store): Emit a clobber either side of an
3051 IFN_STORE_LANES sequence.
3052 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
3054 2018-05-09 Tom de Vries <tom@codesourcery.com>
3057 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
3058 (define_insn "trap_if_false"): Add exit after trap.
3060 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3062 PR rtl-optimization/85638
3063 * bb-reorder.c: Include common/common-target.h.
3064 (create_forwarder_block): New function extracted from...
3065 (fix_up_crossing_landing_pad): ...here. Rename into...
3066 (dw2_fix_up_crossing_landing_pad): ...this.
3067 (sjlj_fix_up_crossing_landing_pad): New function.
3068 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3069 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3070 from both partitions and exit the loop after one iteration.
3072 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3075 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3077 (Basic PowerPC Built-in Functions): The new name of the
3078 subsection previously known as "PowerPC Built-in Functions".
3079 (Basic PowerPC Built-in Functions Available on all Configurations):
3081 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3083 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3085 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3087 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3090 2018-05-08 Jim Wilson <jimw@sifive.com>
3092 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
3093 (LD_EMUL_SUFFIX): New.
3094 (LINK_SPEC): Use it.
3096 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
3098 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3100 (Basic PowerPC Built-in Functions): The new name of the
3101 subsection previously known as "PowerPC Built-in Functions".
3102 (Basic PowerPC Built-in Functions Available on all Configurations):
3104 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
3106 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
3108 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
3110 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
3113 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3116 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3117 after cmpelim optimization.
3119 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
3121 * config.gcc: Support "goldmont".
3122 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
3123 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3125 * config/i386/i386.c (m_GOLDMONT): Define.
3126 (processor_target_table): Add "goldmont".
3127 (PTA_GOLDMONT): Define.
3128 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3129 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3130 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3131 (fold_builtin_cpu): Add "goldmont".
3132 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3133 (ix86_option_override_internal): Add "goldmont".
3134 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3135 (processor_type): Add PROCESSOR_GOLDMONT.
3136 * config/i386/i386.md: Add CPU "glm".
3137 * config/i386/glm.md: New file.
3138 * config/i386/x86-tune.def: Add m_GOLDMONT.
3139 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3141 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3144 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3146 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3147 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3148 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3151 * config/i386/i386.c (ix86_fold_builtin): Handle
3152 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3155 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3156 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3158 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3161 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3162 (check_arch): Likewise.
3163 (check_fpu): Return the result rather than printing it.
3164 (end arch): Fix operator precedence.
3165 (end cpu): Likewise.
3166 (END): Print the result from check_fpu.
3168 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3169 Alan Hayward <alan.hayward@arm.com>
3170 David Sherwood <david.sherwood@arm.com>
3172 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3173 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3174 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3175 (*fcmuo<mode>_and): New patterns.
3177 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3179 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3180 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3181 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3182 (cmp_op, sve_imm_con): New code attributes.
3183 (SVE_COND_INT_CMP, imm_con): Delete.
3184 (cmp_op): Remove above unspecs from int attribute.
3185 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3187 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3188 comparison-specific unspecs.
3189 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3190 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3191 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3192 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3193 (*vec_fcm<cmp_op><mode>): Rename to...
3194 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3195 (*vec_fcmuo<mode>): Rename to...
3196 (*fcmuo<mode>): ...this and adjust likewise.
3197 (*pred_fcm<cmp_op><mode>): New pattern.
3198 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3199 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3201 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3203 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3204 (aarch64_emit_sve_predicated_cond): New function.
3205 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3206 (aarch64_emit_unspec_cond_or): Replace with...
3207 (aarch64_emit_sve_or_conds): ...this new function. Use
3208 aarch64_emit_sve_ptrue_op for the individual comparisons and
3209 aarch64_emit_binop to OR them together.
3210 (aarch64_emit_inverted_unspec_cond): Replace with...
3211 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3212 aarch64_emit_sve_ptrue_op for the comparison and
3213 aarch64_emit_unop to invert the result.
3214 (aarch64_expand_sve_vec_cmp_float): Update after the above
3215 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3217 2018-05-07 Nathan Sidwell <nathan@acm.org>
3219 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3220 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3221 (Backwards Compatibility): Likewise.
3223 2018-05-07 Luis Machado <luis.machado@linaro.org>
3227 2018-05-07 Luis Machado <luis.machado@linaro.org>
3229 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3230 <prefetch_dynamic_strides>: New const bool field.
3231 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3232 prefetch_dynamic_strides.
3233 (exynosm1_prefetch_tune): Likewise.
3234 (thunderxt88_prefetch_tune): Likewise.
3235 (thunderx_prefetch_tune): Likewise.
3236 (thunderx2t99_prefetch_tune): Likewise.
3237 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3239 (aarch64_override_options_internal): Update to set
3240 PARAM_PREFETCH_DYNAMIC_STRIDES.
3241 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3242 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3243 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3244 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3245 prefetch-dynamic-strides setting.
3247 2018-05-07 Luis Machado <luis.machado@linaro.org>
3249 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3250 <minimum_stride>: New const int field.
3251 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3252 minimum_stride field.
3253 (exynosm1_prefetch_tune): Likewise.
3254 (thunderxt88_prefetch_tune): Likewise.
3255 (thunderx_prefetch_tune): Likewise.
3256 (thunderx2t99_prefetch_tune): Likewise.
3257 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3258 (aarch64_override_options_internal): Update to set
3259 PARAM_PREFETCH_MINIMUM_STRIDE.
3260 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3261 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3262 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3263 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3264 stride is constant and is below the minimum stride threshold.
3266 2018-05-07 Luis Machado <luis.machado@linaro.org>
3268 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3271 2018-05-07 Luis Machado <luis.machado@linaro.org>
3273 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3274 <prefetch_dynamic_strides>: New const bool field.
3275 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3276 prefetch_dynamic_strides.
3277 (exynosm1_prefetch_tune): Likewise.
3278 (thunderxt88_prefetch_tune): Likewise.
3279 (thunderx_prefetch_tune): Likewise.
3280 (thunderx2t99_prefetch_tune): Likewise.
3281 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3283 (aarch64_override_options_internal): Update to set
3284 PARAM_PREFETCH_DYNAMIC_STRIDES.
3285 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3286 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3287 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3288 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3289 prefetch-dynamic-strides setting.
3291 2018-05-07 Luis Machado <luis.machado@linaro.org>
3293 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3294 <minimum_stride>: New const int field.
3295 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3296 minimum_stride field.
3297 (exynosm1_prefetch_tune): Likewise.
3298 (thunderxt88_prefetch_tune): Likewise.
3299 (thunderx_prefetch_tune): Likewise.
3300 (thunderx2t99_prefetch_tune): Likewise.
3301 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3302 (aarch64_override_options_internal): Update to set
3303 PARAM_PREFETCH_MINIMUM_STRIDE.
3304 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3305 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3306 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3307 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3308 stride is constant and is below the minimum stride threshold.
3310 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3313 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3314 the type is addressable. Don't force op into register if it has
3317 2018-05-05 Roland McGrath <mcgrathr@google.com>
3320 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3321 any section for which we don't know a specific type it should have,
3322 regardless of name. Previously this was done only for the exact
3323 names ".init_array", ".fini_array", and ".preinit_array".
3324 (default_elf_asm_named_section): Add comment about
3325 relationship with default_section_type_flags and SECTION_NOTYPE.
3326 (get_section): Don't consider it a type conflict if one side has
3327 SECTION_NOTYPE and the other doesn't, as long as neither has the
3328 SECTION_BSS et al used in the default_section_type_flags logic.
3330 2018-05-05 Tom de Vries <tom@codesourcery.com>
3333 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3334 (workaround_barsyncs): New function.
3335 (nvptx_reorg): Use workaround_barsyncs.
3336 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3337 (define_expand "nvptx_membar_cta"): New define_expand.
3338 (define_insn "*nvptx_membar_cta"): New insn.
3340 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3342 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3343 To improve optimization opportunities.
3344 * builtin-types.def: The new needed builtin types for the above.
3346 2018-05-04 Richard Biener <rguenther@suse.de>
3348 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3349 * gimple-ssa-store-merging.c
3350 (imm_store_chain_info::output_merged_store): Remove redundant create,
3351 release split_store vector contents on failure.
3352 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3353 scalar stmt vector on cache hit.
3355 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3357 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3359 * config.gcc (powerpc-xilinx-eabi*): Remove.
3360 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3362 (fusion_addis_mem_combo_load): Ditto.
3363 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3365 (rs6000_cpu_cpp_builtins): Ditto.
3366 * config/rs6000/rs6000-linux.c
3367 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3368 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3369 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3371 (rs6000_setup_reg_addr_masks): Ditto.
3372 (rs6000_init_hard_regno_mode_ok): Ditto.
3373 (rs6000_option_override_internal): Ditto.
3374 (legitimate_lo_sum_address_p): Ditto.
3375 (rs6000_legitimize_address): Ditto.
3376 (rs6000_legitimize_reload_address): Ditto.
3377 (rs6000_legitimate_address_p): Ditto.
3378 (abi_v4_pass_in_fpr): Ditto.
3379 (setup_incoming_varargs): Ditto.
3380 (rs6000_gimplify_va_arg): Ditto.
3381 (rs6000_split_multireg_move): Ditto.
3382 (rs6000_savres_strategy): Ditto.
3383 (rs6000_emit_prologue_components): Ditto.
3384 (rs6000_emit_epilogue_components): Ditto.
3385 (rs6000_emit_prologue): Ditto.
3386 (rs6000_emit_epilogue): Ditto.
3387 (rs6000_elf_file_end): Ditto.
3388 (rs6000_function_value): Ditto.
3389 (rs6000_libcall_value): Ditto.
3390 * config/rs6000/rs6000.h: Ditto.
3391 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3392 (TARGET_MINMAX): ... this. New.
3393 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3394 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3395 (*movsi_internal1_single): Delete.
3396 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3397 mfpu=, mxilinx-fpu): Delete.
3398 * config/rs6000/singlefp.h: Delete.
3399 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3400 * config/rs6000/t-rs6000: Ditto.
3401 * config/rs6000/t-xilinx: Delete.
3402 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3403 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3404 (VStype_simple): Delete.
3405 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3406 * config/rs6000/xfpu.h: Delete.
3407 * config/rs6000/xfpu.md: Delete.
3408 * config/rs6000/xilinx.h: Delete.
3409 * config/rs6000/xilinx.opt: Delete.
3410 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3411 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3413 2018-05-04 Tom de Vries <tom@codesourcery.com>
3416 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3419 2018-05-04 Richard Biener <rguenther@suse.de>
3422 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3423 (expand_complex_div_wide): Likewise.
3424 (expand_complex_operations_1): Likewise.
3425 (expand_complex_libcall): Preserve EH info of the original stmt.
3426 (tree_lower_complex): Handle removed blocks.
3427 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3428 on complex multiplication and division libcall builtins.
3430 2018-05-04 Richard Biener <rguenther@suse.de>
3433 * fold-const.c (negate_expr_p): Restrict negation of operand
3434 zero of a division to when we know that can happen without
3436 (fold_negate_expr_1): Likewise.
3438 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3441 * real.h (real_nextafter): Declare.
3442 * real.c (real_nextafter): New function.
3443 * fold-const-call.c (fold_const_nextafter): New function.
3444 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3445 CASE_CFN_NEXTTOWARD.
3446 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3447 even when arg1_mode is different from arg0_mode.
3449 2018-05-03 Nathan Sidwell <nathan@acm.org>
3451 * doc/extend.texi (Deprecated Features): Remove
3453 (Backwards Compatibility): Likewise.
3454 * doc/invoke.texi (C++ Language Options): Likewise.
3455 (C++ Dialect Options): Likewise.
3457 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3460 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3461 _mm512_mask_mullox_epi64): New intrinsics.
3463 2018-05-03 Tom de Vries <tom@codesourcery.com>
3466 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3467 dump files): Add offload-tree.
3469 2018-05-03 Richard Biener <rguenther@suse.de>
3471 PR tree-optimization/85615
3472 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3473 to loops not nested in BBs loop father to avoid creating multi-entry
3476 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3478 PR tree-optimization/70291
3479 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3480 arguments. Change return type to tree. Emit libcall as a new
3481 statement rather than replacing existing one when inplace_p is true.
3482 (expand_complex_multiplication_components): New function.
3483 (expand_complex_multiplication): Expand floating-point complex
3484 multiplication using the above.
3485 (expand_complex_division): Rename inner_type parameter to type.
3486 Update expand_complex_libcall call-site.
3487 (expand_complex_operations_1): Update expand_complex_multiplication
3488 and expand_complex_division call-sites.
3490 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3493 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3494 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3495 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3496 the highest significant bit of the shift count mask is clear. In
3497 check whether and[sq]i3 is needed verify that all significant bits
3498 of the shift count other than the highest are set.
3500 2018-05-02 Tom de Vries <tom@codesourcery.com>
3503 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3504 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3505 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3506 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3507 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3508 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3509 __builtin_goacc_parlevel_size.
3511 2018-05-02 Richard Biener <rguenther@suse.de>
3513 PR tree-optimization/85597
3514 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3515 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3518 2018-05-02 Tom de Vries <tom@codesourcery.com>
3521 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3522 dump files): Add ltrans-tree.
3524 2018-05-02 Tom de Vries <tom@codesourcery.com>
3527 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3528 dump files): Add wpa-ipa.
3530 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3532 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3533 powerpc*-*-linux*paired* target.
3534 * config/rs6000/750cl.h: Delete.
3535 * config/rs6000/paired.h: Delete.
3536 * config/rs6000/paired.md: Delete.
3537 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3539 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3540 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3541 comment. Remove paired float support.
3542 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3543 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3545 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3546 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3548 * config/rs6000/rs6000.c: Remove paired float support.
3549 (paired_expand_vector_init, paired_expand_vector_move,
3550 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3551 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3552 paired_expand_builtin, paired_expand_predicate_builtin,
3553 paired_init_builtins): Delete.
3554 * config/rs6000/rs6000.h: Remove paired float support.
3555 * config/rs6000/rs6000.md: Remove paired float support.
3556 (move_from_CR_ov_bit): Delete.
3557 * config/rs6000/rs6000.opt (mpaired): Delete.
3558 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3559 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3561 2018-05-02 Richard Biener <rguenther@suse.de>
3564 * gimplify.c (gimplify_save_expr): When in SSA form allow
3565 SAVE_EXPRs to compute to SSA vars.
3567 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3570 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3571 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3572 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3573 clobber operands[2], instead use a new pseudo. Formatting fixes.
3575 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3577 PR tree-optimization/85586
3578 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3579 exit early for statements in the same group if the accesses are
3582 2018-05-02 Tom de Vries <tom@codesourcery.com>
3585 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3588 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3590 PR tree-optimization/85143
3591 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3593 2018-05-01 Tom de Vries <tom@codesourcery.com>
3596 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3597 not found" error message.
3599 2018-05-01 Tom de Vries <tom@codesourcery.com>
3602 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3603 * vec.c (test_ordered_remove_if): New function.
3604 (vec_c_tests): Call test_ordered_remove_if.
3605 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3606 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3607 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3608 VEC_ORDERED_REMOVE_IF.
3610 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3612 PR tree-optimization/82665
3613 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3614 pointer subtraction where arguments come from a memchr call.
3616 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3618 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3619 --push-state --as-needed and --pop-state instead of --as-needed and
3620 --no-as-needed if ld supports it.
3621 * configure: Regenerated.
3624 * doc/install.texi2html: Replace _002d with - and _002a with * in
3625 generated html files using sed.
3627 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3630 * gcc-rich-location.c (blank_line_before_p): New function.
3631 (use_new_line): New function.
3632 (gcc_rich_location::add_fixit_insert_formatted): New function.
3633 * gcc-rich-location.h
3634 (gcc_rich_location::add_fixit_insert_formatted): New function.
3636 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3638 * selftest.c (assert_streq): Rename "expected" and "actual" to
3639 "val1" and "val2". Extend NULL-handling to cover both inputs
3640 symmetrically, while still requiring both to be non-NULL for a pass.
3641 * selftest.h (assert_streq): Rename "expected" and "actual" to
3643 (ASSERT_EQ): Likewise.
3644 (ASSERT_EQ_AT): Likewise.
3645 (ASSERT_KNOWN_EQ): Likewise.
3646 (ASSERT_KNOWN_EQ_AT): Likewise.
3647 (ASSERT_NE): Likewise.
3648 (ASSERT_MAYBE_NE): Likewise.
3649 (ASSERT_MAYBE_NE_AT): Likewise.
3650 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3651 the assertion to pass.
3652 (ASSERT_STREQ_AT): Likewise.
3654 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3656 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3657 interaction with -pie.
3659 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3661 * selftest.h: Fix alphabetization of per-source-file selftest
3664 2018-04-30 Jason Merrill <jason@redhat.com>
3666 PR c++/61982 - dead stores to destroyed objects.
3667 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3670 2018-04-30 Jason Merrill <jason@redhat.com>
3672 * tree.c (build_clobber): New.
3673 * tree.h: Declare it.
3674 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3676 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3678 * diagnostic-show-locus.c (layout::layout): Update for
3679 location_get_source_line returning a char_span.
3680 (struct char_span): Move to input.h.
3681 (struct correction): Update for fields in char_span becoming
3683 (struct source_line): Update for location_get_source_line
3684 returning a char_span.
3685 (layout::print_line): Likewise.
3686 * edit-context.c (edited_file::print_content): Likewise.
3687 (edited_file::print_diff_hunk): Likewise.
3688 (edited_file::print_run_of_changed_lines): Likewise.
3689 (edited_file::get_num_lines): Likewise.
3690 (edited_line::edited_line): Likewise.
3691 * final.c (asm_show_source): Likewise.
3692 * input.c (location_get_source_line): Convert return type
3693 from const char * to char_span, losing the final "line_len"
3695 (dump_location_info): Update for the above.
3696 (get_substring_ranges_for_loc): Likewise. Use a char_span
3697 when handling the literal within the line.
3698 (test_reading_source_line): Update for location_get_source_line
3699 returning a char_span.
3700 * input.h (class char_span): Move here from
3701 diagnostic-show-locus.c, converting from a struct to a class.
3702 Make data members private.
3703 (char_span::operator bool): New.
3704 (char_span::length): New.
3705 (char_span::get_buffer): New.
3706 (char_span::operator[]): New.
3707 (char_span::subspan): Make const.
3708 (char_span::xstrdup): New.
3709 (location_get_source_line): Convert return type from const char *
3710 to char_span, losing the final "line_size" param.
3712 2018-04-30 Jan Hubicka <jh@suse.cz>
3714 * lto-wrapper.c (ltrans_priorities): New static var.
3715 (cmp_priority): New.
3716 (run_gcc): Read priorities and if doing parallel build order
3717 the Makefile by them.
3719 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3721 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3723 2018-04-30 Richard Biener <rguenther@suse.de>
3725 * tree-cfg.c (verify_address): Remove base argument, add
3726 flag whether to check TREE_ADDRESSABLE and do that.
3727 (verify_expr): Remove.
3728 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3729 (verify_gimple_assign_single): Likewise.
3730 (verify_gimple_switch): Likewise.
3731 (verify_expr_location_1): Dereference tp once. Add (disabled)
3732 piece from verify_expr.
3733 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3735 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3737 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3739 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3741 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3742 (small_data_pattern): Likewise.
3743 (arc_rewrite_small_data): Likewise.
3744 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3745 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3746 (get_symbol_alignment): New function.
3747 (legitimate_small_data_address_p): Likewise.
3748 (legitimate_scaled_address): Update, call
3749 legitimate_small_data_address_p.
3750 (output_sdata): New static variable.
3751 (arc_print_operand): Update how we handle small data operands.
3752 (arc_print_operand_address): Likewise.
3753 (arc_legitimate_address_p): Update, use
3754 legitimate_small_data_address_p.
3755 (arc_rewrite_small_data_p): Remove.
3756 (arc_rewrite_small_data_1): Likewise.
3757 (arc_rewrite_small_data): Likewise.
3758 (small_data_pattern): Likewise.
3759 (compact_sda_memory_operand): Update to use
3760 legitimate_small_data_address_p and get_symbol_alignment.
3761 (prepare_move_operands): Don't rewite sdata pattern.
3762 (prepare_extend_operands): Remove.
3763 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3765 (zero_extendqisi2): Likewise.
3766 (zero_extendhisi2): Likewise.
3767 (extendqihi2): Likewise.
3768 (extendqisi2): Likewise.
3769 (extendhisi2): Likewise.
3773 * config/arc/constraints.md (Usd): Change it to memory constraint.
3775 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3777 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3778 as source of std instructions.
3779 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3780 allow 6-bit constants as source for store instructions.
3781 (movdi_insn): Update instruction pattern to allow 6-bit constants
3782 as source for store instructions.
3784 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3786 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3788 2018-04-30 Nathan Sidwell <nathan@acm.org>
3789 Sandra Loosemore <sandra@codesourcery.com>
3791 * dumpfile.c (dump_open): Allow '-' for stdout.
3792 * doc/invoke.texi (Developer Options): Document dump filename
3793 determination early. Document stdin/stdout selection.
3795 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3797 Microblaze Target: PIC data text relative
3799 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3800 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3802 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3803 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3804 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3805 New addressing mode for data-text relative position indepenedent code.
3806 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3807 'ADDRESS_SYMBOLIC_TXT_REL'.
3808 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3809 (microblaze_legitimate_pic_operand): Exclude function calls from
3810 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3811 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3813 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3814 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3815 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3816 for 'address + offset'.
3817 (microblaze_expand_prologue): Add new function prologue call for
3819 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3820 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3821 table in case of TARGET_PIC_DATA_TEXT_REL.
3822 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3823 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3824 Add new macros 'UNSPEC_TEXT',
3825 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3826 + exclude function calls from 'UNSPEC_PLT' in case of data text
3828 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3829 new target hook for generating address diff vector tables in case of
3831 * doc/tm.texi : Regenerate.
3832 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3833 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3834 of addr diff vector generation.
3835 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3836 target hook definition.
3837 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3838 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3839 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3840 option for data text relative.
3842 2018-04-30 Richard Biener <rguenther@suse.de>
3844 * tree-chrec.h (evolution_function_is_constant_p): Remove
3846 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3848 2018-04-30 Richard Biener <rguenther@suse.de>
3851 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3853 2018-04-30 Richard Biener <rguenther@suse.de>
3855 PR tree-optimization/28364
3856 PR tree-optimization/85275
3857 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3858 copying first exit test.
3860 2018-04-28 Mark Wielaard <mark@klomp.org>
3862 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3864 (dwarf_AT): Handle DW_AT_addr_base.
3865 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3867 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3870 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3871 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3872 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3873 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3875 2018-04-28 Richard Biener <rguenther@suse.de>
3877 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3878 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3879 to reflect use. Only add interesting stmts.
3881 2018-04-27 Martin Jambor <mjambor@suse.cz>
3884 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3885 the jump function allows for passing through aggregate values.
3887 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3889 * input.h (in_system_header_at): Convert from macro to inline
3891 (from_macro_expansion_at): Likewise.
3892 (from_macro_definition_at): Likewise.
3894 2018-04-27 Jeff Law <law@redhat.com>
3896 * config.gcc: Mark tile* targets as deprecated/obsolete.
3898 2018-04-27 Richard Biener <rguenther@suse.de>
3900 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3903 2018-04-27 Richard Biener <rguenther@suse.de>
3905 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3907 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3909 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3910 with Yd constraint. Set "preferred_for_speed" attribute from
3911 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3913 (*movdi_internal): Ditto.
3914 (movti_interunit splitters): Remove
3915 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3916 (movdi_interunit splitters): Ditto.
3917 * config/i386/constraints.md (Ye): Remove.
3918 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3920 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3923 * config/aarch64/constraints.md (Usg): Limit to 31.
3926 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3928 PR tree-optimization/85529
3929 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3930 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3931 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3932 zero extension or masking of the MSB bit.
3933 (optimize_range_tests): Add FIRST_BB argument, pass it through
3934 to optimize_range_tests_var_bound.
3935 (maybe_optimize_range_tests, reassociate_bb): Adjust
3936 optimize_range_tests callers.
3938 2018-04-26 Richard Biener <rguenther@suse.de>
3939 Jakub Jelinek <jakub@redhat.com>
3941 * cgraph.h (symbol_table): Just declare debug method here.
3942 * symtab.c (symbol_table::debug): Define.
3944 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3946 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3948 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3950 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3951 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3952 (*movdi_internal): Substitute Yi and Yj constraint with x
3953 and Ym and Yn constraint with y constraint. Update "isa"
3954 attribute and set "preferred_for_speed" attribute from
3955 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3956 (*movsi_internal): Ditto.
3957 (*movdf_internal): Ditto.
3958 (*movsf_internal): Ditto.
3959 (*zero_extendsidi2): Ditto.
3960 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3961 (sse2_loadld): Ditto.
3962 (*vec_extract<ssevecmodelower>_0): Ditto.
3963 (*vec_extractv4si_0_zext_sse4): Ditto.
3964 (vec_concatv2di): Ditto.
3965 (*vec_dup<mode>): Ditto.
3966 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3967 * config/i386/constraints.md (Yi): Remove.
3972 2018-04-26 Nathan Sidwell <nathan@acm.org>
3974 * dumpfile.c (dump_open): New.
3975 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3976 (dump_finish): Detect stdio/stderr by value not name.
3978 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3980 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3982 2018-04-26 Tom de Vries <tom@codesourcery.com>
3985 * config/nvptx/nvptx.c (verify_neutering_jumps)
3986 (verify_neutering_labels): New function
3987 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3989 2018-04-26 Tom de Vries <tom@codesourcery.com>
3992 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3993 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3996 2018-04-26 Richard Biener <rguenther@suse.de>
3997 Tom de Vries <tom@codesourcery.com>
4000 * lto-streamer-out.c (output_function): Fixup loops if required to match
4001 discovery done in the reader.
4003 2018-04-26 Richard Biener <rguenther@suse.de>
4005 PR tree-optimization/85116
4006 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
4007 have a loop exit from the single latch predecessor. Remove
4008 case of header with just condition.
4009 (ch_base::copy_headers): Exclude infinite loops from any
4011 (pass_ch::execute): Record exits.
4013 2018-04-26 Richard Biener <rguenther@suse.de>
4015 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
4016 prologue cost vector and pass it to vect_get_load_cost.
4017 (vect_get_peeling_costs_all_drs): Likewise.
4018 (vect_peeling_hash_get_lowest_cost): Likewise.
4019 (vect_enhance_data_refs_alignment): Likewise.
4021 2018-04-26 Richard Biener <rguenther@suse.de>
4024 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
4025 checking of integer<->pointer conversions.
4026 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
4027 sign-/zero-extending pointer types.
4028 (expand_omp_for_static_chunk): Likewise.
4030 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
4031 Jean Lee <xiaoyur347@gmail.com>
4033 * config/mips/mips.c (mips_asan_shadow_offset): New function.
4034 (TARGET_ASAN_SHADOW_OFFSET): Define.
4035 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
4036 true for -fsanitize=address.
4038 2018-04-25 Mark Wielaard <mark@klomp.org>
4040 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
4043 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4045 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
4046 than "alu", remove explicit "memory" and "imm_disp" attributes.
4047 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
4050 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
4051 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
4052 gen_lowpart_no_emit.
4054 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
4057 * config/i386/i386.c (ix86_expand_builtin): Change memory
4058 operand to XI, extend p0 to Pmode.
4059 * config/i386/i386.md: Change unspec volatile and operand
4060 1 mode to XI, change operand 0 mode to P.
4062 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4064 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
4065 GET_MODE_MASK before any checking.
4066 (nds32_can_use_bset_p): Likewise.
4067 (nds32_can_use_btgl_p): Likewise.
4069 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4071 * config/nds32/nds32-doubleword.md: New define_split pattern for
4072 illegal register number.
4074 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4076 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
4078 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
4080 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
4082 2018-04-25 Richard Biener <rguenther@suse.de>
4084 * lto-streamer.h (LTO_major_version): Bump to 8.
4086 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4088 * BASE-VER: Set to 9.0.0.
4090 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4092 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4093 in __abskf2 and __powikf2.
4095 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4098 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4099 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4100 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4101 Use the above on operand 2. Reindent.
4102 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4104 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4107 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4109 * config/i386/i386.opt (mcet): Removed.
4110 * doc/install.texi: Remove -mcet documentation.
4111 * doc/invoke.texi: Likewise.
4113 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4116 * doc/install.texi: Remove -mcet from bootstrap-cet.
4118 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4121 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4122 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4126 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4127 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4128 containing a CONST_VECTOR.
4130 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4132 * doc/install.texi: Update newlib dependency for nvptx.
4134 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4137 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4138 instead of INTVAL when shifting x left.
4140 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4142 PR tree-optimization/85478
4143 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4144 vect_grouped_store_supported for single element vectors.
4146 2018-04-24 Richard Biener <rguenther@suse.de>
4149 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4150 load cost increase to the case of non-constant step.
4152 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4155 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4156 destination if any_malformed_asm.
4158 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4161 * expr.c (store_field): In the bitfield case, if the value comes from
4162 a function call and is returned in registers by means of a PARALLEL,
4163 do not change the mode of the temporary unless BLKmode and VOIDmode.
4165 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4167 PR rtl-optimization/85423
4168 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4169 dependencies to debug insns when the previous insn is non-debug.
4171 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4173 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4174 enums into a single definition.
4175 (fls): Fix predicates and printing.
4178 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4180 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4181 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4182 and short u6 immediate.
4183 (check_if_valid_sleep_operand): Remove.
4184 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4186 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4188 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4189 flag_always_save_lp condition.
4190 * config/nds32/nds32.opt (malways-save-lp): New option.
4192 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4194 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4195 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4196 * config/nds32/nds32.h
4197 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4198 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4200 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4202 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4203 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4205 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4206 Chung-Ju Wu <jasonwucj@gmail.com>
4208 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4210 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4211 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4213 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4215 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4217 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4219 * config/nds32/nds32-protos.h (nds32_data_alignment,
4220 nds32_local_alignment): Declare.
4221 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4222 nds32_local_alignment): New functions.
4223 (TARGET_CONSTANT_ALIGNMENT): Define.
4224 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4226 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4228 * config/nds32/nds32.c
4229 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4230 (TARGET_MODES_TIEABLE_P): Likewise.
4232 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4234 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4237 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4238 Chung-Ju Wu <jasonwucj@gmail.com>
4240 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4241 for unaligned access.
4242 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4244 * config/nds32/nds32-intrinsic.md: Likewise.
4245 * config/nds32/nds32_intrinsic.h: Likewise.
4246 * config/nds32/nds32.h (nds32_builtins): Likewise.
4247 * config/nds32/nds32.opt (munaligned-access): New option.
4248 * config/nds32/nds32.c (nds32_asm_file_start): Display
4249 flag_unaligned_access status.
4251 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4253 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4254 -mno-relax is present.
4255 * config/riscv/linux.h (LINK_SPEC): Ditto.
4257 2018-04-20 Martin Sebor <msebor@redhat.com>
4260 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4262 (gimple_fold_builtin_stxcpy_chk): Same.
4263 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4265 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4268 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4269 __powikf2 when long double is IEEE 128-bit.
4271 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4273 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4274 step to make sure stack always aligned.
4276 2018-04-20 Carl Love <cel@us.ibm.com>
4279 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4280 size check for arg0.
4282 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4283 Tom de Vries <tom@codesourcery.com>
4286 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4287 Emit insns for calls too.
4288 (nvptx_find_par): Always look for worker-level predecessor insn.
4289 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4291 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4292 (nvptx_process_pars): Propagate frames for calls.
4294 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4297 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4299 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4300 (ix86_handle_option): Don't handle OPT_mibt.
4301 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4303 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4305 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4306 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4307 (ix86_target_macros): Define __CET__ with flag_cf_protection
4308 for -fcf-protection.
4309 * config/i386/i386.c (isa2_opts): Remove -mibt.
4310 * config/i386/i386.h (TARGET_IBT): Removed.
4311 (TARGET_IBT_P): Likewise.
4312 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4313 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4314 * config/i386/i386.opt (mcet): Update help message.
4317 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4318 -mcet as an alias for -mshstk.
4320 2018-04-20 Richard Biener <rguenther@suse.de>
4323 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4324 complexity by forcing a single use of the multiply operand.
4326 2018-04-20 Martin Jambor <mjambor@suse.cz>
4329 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4330 recursion dependency to only apply to non-clones.
4332 2018-04-20 Martin Jambor <mjambor@suse.cz>
4335 * ipa-cp.c (create_specialized_node): Check that clones of
4336 self-recursive edges exist during IPA-CP.
4338 2018-04-19 Toon Moene <toon@moene.org>
4340 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4343 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4345 PR tree-optimization/85467
4346 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4347 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4348 VECTOR_CST element to type.
4350 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4353 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4354 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4355 (builtin_longjmp): Likewise.
4356 (save_stack_nonlocal): New pattern.
4357 (restore_stack_nonlocal): Likewise.
4359 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4362 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4363 Replace ASM_OUTPUT_LABEL with fprintf.
4365 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4368 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4369 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4370 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4371 define __IBT__ and __SHSTK__ for -fcf-protection.
4372 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4374 (ix86_trampoline_init): Likewise.
4375 (x86_output_mi_thunk): Likewise.
4376 (ix86_notrack_prefixed_insn_p): Likewise.
4377 (ix86_option_override_internal): Don't disallow -fcf-protection.
4378 * config/i386/i386.md (rdssp<mode>): Also enable for
4380 (incssp<mode>): Likewise.
4381 (nop_endbr): Likewise.
4382 * config/i386/i386.opt (mcet): Change help message to built-in
4386 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4387 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4388 enable CET built-in functions.
4390 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4392 * common/config/i386/i386-common.c
4393 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4394 OPTION_MASK_ISA_MOVDIRI_UNSET,
4395 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4396 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4397 * config.gcc (movdirintrin.h): New header.
4398 * config/i386/cpuid.h (bit_MOVDIRI,
4399 bit_MOVDIR64B): New bits.
4400 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4402 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4403 (VOID, PVOID, PCVOID)): New function types.
4404 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4405 __builtin_ia32_directstoreu_u64,
4406 __builtin_ia32_movdir64b): New builtins.
4407 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4408 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4410 (ix86_valid_target_attribute_inner_p): Ditto.
4411 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4412 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4413 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4414 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4415 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4416 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4417 (movdiri<mode>, movdir64b_<mode>): New.
4418 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4419 * config/i386/immintrin.h: Include movdirintrin.h.
4420 * config/i386/movdirintrin.h: New file.
4421 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4423 2018-04-19 Richard Biener <rguenther@suse.de>
4426 * cfg.c (clear_bb_flags): When loop state says we have
4427 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4429 2018-04-19 Richard Biener <rguenther@suse.de>
4431 PR tree-optimization/84737
4432 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4433 copying restrict info.
4434 (vect_setup_realignment): Use it.
4435 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4436 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4437 the first DR to all generated stores.
4438 (vectorizable_load): Likewise for loads.
4440 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4442 PR tree-optimization/85446
4443 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4444 the integral and pointer types to have the same precision.
4446 * doc/install.texi: Document --disable-cet being the default and
4449 2018-04-18 Martin Liska <mliska@suse.cz>
4451 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4454 2018-04-18 Martin Liska <mliska@suse.cz>
4457 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4460 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4461 arguments if they are comparable.
4463 2018-04-18 Martin Liska <mliska@suse.cz>
4466 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4469 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4472 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4475 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4476 ENDBR after calling __morestack.
4478 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4481 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4482 by using gcc_base_ver to generate a gcc_driver_version, and use
4483 it when generating GCC_DRIVER_NAME.
4484 * configure: Regenerate.
4486 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4489 * config.gcc: Obsolete powerpc*-*-*spe*.
4491 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4494 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4495 (stabstr_D): Change type of unum from unsigned int to
4496 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4499 2018-04-17 Jim Wilson <jimw@sifive.com>
4502 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4503 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4504 Set arg_pointer_offset after using pretend_args_size.
4506 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4508 PR rtl-optimization/85431
4509 * dse.c (record_store): Ignore zero width stores.
4512 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4513 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4514 __builtin_stack_restore rather than after it.
4515 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4516 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4517 argument instead of virtual_dynamic_stack_rtx.
4519 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4521 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4523 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4524 Add note to error message to explain internal mapping of overloaded
4525 built-in function name to non-overloaded built-in function name.
4526 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4529 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4532 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4533 where the inputs overlap with the output.
4535 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4537 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4538 (=v, v) alternative and explicit "memory" attribute.
4539 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4540 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4542 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4543 "sselog1" type instead of "sselog".
4544 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4545 "sselog". Remove explicit "memory" attribute.
4546 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4547 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4549 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4550 "sselog1" type instead of "sselog". Remove explicit "memory"
4552 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4553 use "sselog1" type instead of "sselog". Remove explicit "memory"
4555 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4556 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4558 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4559 "sselog1" type instead of "sselog". Remove explicit "memory"
4561 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4562 use "sselog1" type instead of "sselog". Remove explicit "memory"
4566 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4569 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4572 2018-04-17 Martin Jambor <mjambor@suse.cz>
4575 * ipa-cp.c (create_specialized_node): Call
4576 expand_all_artificial_thunks if necessary.
4578 2018-04-17 Martin Liska <mliska@suse.cz>
4581 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4582 in message, remote space in between '_G' and '('.
4584 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4587 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4588 avx512f_vmcmp<mode>3<round_saeonly_name>,
4589 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4590 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4591 avx512f_rndscale<mode><round_saeonly_name>,
4592 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4593 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4594 Use %<iptr>2 instead of %2 for -masm=intel.
4595 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4596 avx512f_vcvttss2usi<round_saeonly_name>,
4597 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4599 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4600 avx512f_vcvttsd2usi<round_saeonly_name>,
4601 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4602 Use %q1 instead of %1 for -masm=intel.
4603 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4604 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4605 of %3 for -masm=intel.
4606 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4608 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4610 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4612 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4613 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4615 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4617 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4618 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4619 %g1 and one with %0 and %1.
4620 (avx512er_vmrcp28<mode><round_saeonly_name>,
4621 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4623 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4624 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4625 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4626 of %0 and %{%4%} for -masm=intel.
4627 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4628 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4629 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4630 order of %0 and %{%5%}%{z%} for -masm=intel.
4632 2018-04-17 Jan Hubicka <jh@suse.cz>
4635 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4637 2018-04-17 Martin Liska <mliska@suse.cz>
4640 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4641 for target_clone error message. Make default implementation
4642 clone to be a local declaration.
4643 (separate_attrs): Add new argument and check for an empty
4645 (expand_target_clones): Handle it.
4646 (ipa_target_clone): Make redirection just for target_clones
4649 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4650 Tom de Vries <tom@codesourcery.com>
4653 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4654 tiled basic blocks without omp continue statements.
4656 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4659 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4660 vec_extract expression as having side effects to make sure it gets
4663 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4666 * config/i386/i386.c (get_builtin_code_for_version): Check
4669 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4672 * gcc/config.gcc: Support "skylake".
4673 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4675 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4676 (processor_target_table): Add "skylake".
4677 (ix86_option_override_internal): Add "skylake".
4678 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4679 PROCESSOR_CANNONLAKE.
4680 (get_builtin_code_for_version): Fix priority for
4681 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4682 PROCESSOR_SKYLAKE-AVX512.
4683 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4684 (processor_type): Add PROCESSOR_SKYLAKE.
4686 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4687 Jason Merrill <jason@redhat.com>
4690 * convert.c (convert_to_integer_1): Use direct recursion for
4691 enumeral types and types with a precision less than the number
4692 of bits in their mode.
4694 2018-04-16 Julia Koval <julia.koval@intel.com>
4697 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4698 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4700 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4703 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4704 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4705 and -mno-direct-move.
4707 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4710 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4711 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4712 Ensure negative shifts result in {0}.
4714 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4716 PR rtl-optimization/79916
4717 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4718 regs (if any) to define how to gnerate SD moves when LRA is in
4721 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4723 PR rtl-optimization/85393
4724 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4725 * except.c (expand_dw2_landing_pad_for_region): Make static.
4726 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4727 a label and unconditional jump to old_bb, rather than
4728 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4731 PR rtl-optimization/85376
4732 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4733 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4734 instead of a specific value.
4736 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4737 Bin Cheng <bin.cheng@arm.com>
4739 PR tree-optimization/82965
4740 PR tree-optimization/83991
4741 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4742 by_profile_only parameter.
4743 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4744 information if the loop was predicted to iterate too many times.
4745 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4747 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4750 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4753 2018-04-13 Martin Liska <mliska@suse.cz>
4754 Jakub Jelinek <jakub@redhat.com>
4757 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4758 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4759 * builtins.c (expand_builtin_memory_copy_args): Use
4760 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4761 handle dest_addr == pc_rtx.
4763 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4766 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4767 asked to not generate direct moves.
4768 (fix_trunc<mode>si2_stfiwx): Similar.
4769 (fix_trunc<mode>si2_internal): Similar.
4771 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4774 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4775 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4776 lookup if dest in some wider mode is known to be const0_rtx and
4777 if so, record permanent equivalence for it to be ZERO_EXTEND of
4778 the narrower mode destination.
4780 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4782 * lto-streamer-out.c (output_function): Revert 259346.
4783 * omp-expand.c (expand_oacc_for): Likewise.
4785 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4787 PR rtl-optimization/85354
4788 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4789 * sel-sched.c (sel_global_init): ... here.
4791 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4794 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4795 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4796 mode for PE-COFF targets.
4797 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4798 (i386_pe_asm_lto_end): Likewise.
4799 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4800 (TARGET_ASM_LTO_END): Likewise.
4801 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4802 (i386_pe_asm_lto_start): New function.
4803 (i386_pe_asm_lto_end): Likewise.
4805 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4806 Richard Biener <rguenther@suse.de>
4809 * lto-streamer-out.c (output_function): Fix CFG loop state before
4811 * omp-expand.c (expand_oacc_for): Handle calls to internal
4812 functions like regular functions.
4814 2018-04-12 Richard Biener <rguenther@suse.de>
4817 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4818 for the early LTO debug to properly generate references to it
4819 during DIE emission. Do not re-use that for the skeleton for
4821 (dwarf2out_early_finish): Likewise.
4823 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4826 * config/i386/sse.md
4827 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4828 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4829 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4830 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4831 and output is a reg, avoid creating invalid lowpart subreg, but
4832 instead split into a 512-bit move. Don't split if not AVX512VL,
4833 input is xmm16+ reg and output is a mem.
4834 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4835 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4836 xmm16+ reg and output is a mem.
4838 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4840 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4841 also for flag_dwarf2_cfi_asm.
4843 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4845 PR rtl-optimization/85342
4846 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4847 a bool scalar var inside of the loop instead. Don't try to update
4848 recog_data.operand after failed apply_change_group.
4850 2018-04-12 Tom de Vries <tom@codesourcery.com>
4853 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4854 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4855 array with flexible array member as array without given dimension.
4856 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4857 argument for undefined param to true.
4859 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4862 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4863 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4864 from PowerPC section.
4865 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4866 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4867 help text that is too long.
4868 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4869 help text that is too long.
4870 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4871 help text that is too long.
4873 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4875 * config/alpha/alpha.md (stack_probe_internal): Rename
4876 from "probe_stack". Update all callers.
4878 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4880 PR rtl-optimization/84566
4881 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4882 sched_macro_fuse_insns.
4884 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4887 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4888 (compute_block_dependences): ... from here.
4890 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4892 PR tree-optimization/85331
4893 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4894 from int to HOST_WIDE_INT.
4896 2018-04-11 Martin Jambor <mjambor@suse.cz>
4899 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4900 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4901 not the same as the source val.
4902 (cgraph_edge_brings_value_p): New parameter.
4903 (gather_edges_for_value): Pass destination value to
4904 cgraph_edge_brings_value_p.
4905 (perhaps_add_new_callers): Likewise.
4906 (get_info_about_necessary_edges): Likewise and exclude values brought
4907 only by self-recursive edges.
4908 (create_specialized_node): Redirect only clones of self-calling edges.
4909 (+self_recursive_pass_through_p): New function.
4910 (find_more_scalar_values_for_callers_subset): Use it.
4911 (find_aggregate_values_for_callers_subset): Likewise.
4912 (known_aggs_to_agg_replacement_list): Removed.
4913 (decide_whether_version_node): Re-calculate known constants for all
4914 remaining context clones.
4916 2018-04-11 Richard Biener <rguenther@suse.de>
4919 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4920 from early DWARF output.
4921 (dwarf2out_early_finish): Output line info unconditionally into
4922 early DWARF and add reference to it.
4924 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4927 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4928 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4929 other than V2DFmode using iptr mode attribute.
4930 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4932 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4934 PR rtl-optimization/84659
4935 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4937 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4940 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4942 (output_loc_list): Pass address of a dummy size variable even in the
4943 locview handling loop.
4944 (index_location_lists): Add comment on why skip_loc_list_entry can't
4947 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4950 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4953 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4956 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4957 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4958 and -mstring-compare-inline-limit.
4960 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4963 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4964 for stack clash protection in a register whenever we need it to be in
4967 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4969 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4970 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4972 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4975 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4977 (mlong-double-): Ditto.
4978 * config/rs6000/sysv4.opt (msdata=): Ditto.
4979 (mtls-size=): Ditto.
4981 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4983 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4984 erroneous entries for
4985 "vector int vec_ldl (int, long int *)", and
4986 "vector unsigned int vec_ldl (int, unsigned long int *)".
4987 Add comments and entries for
4988 "vector bool char vec_ldl (int, bool char *)",
4989 "vector bool short vec_ldl (int, bool short *)",
4990 "vector bool int vec_ldl (int, bool int *)",
4991 "vector bool long long vec_ldl (int, bool long long *)",
4992 "vector pixel vec_ldl (int, pixel *)",
4993 "vector long long vec_ldl (int, long long *)",
4994 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4995 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4996 type tree bool_long_long_type_node and correct definition of
4997 bool_V2DI_type_node to make reference to this new type tree.
4998 (rs6000_mangle_type): Replace erroneous reference to
4999 bool_long_type_node with bool_long_long_type_node.
5000 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
5001 comments to emphasize sign distinctions for char and int types and
5002 replace RS6000_BTI_bool_long constant with
5003 RS6000_BTI_bool_long_long constant. Also add comment to restrict
5004 use of RS6000_BTI_pixel.
5005 (bool_long_type_node): Remove this macro definition.
5006 (bool_long_long_type_node): New macro definition
5008 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5010 PR rtl-optimization/85300
5011 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
5012 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
5013 simplify_unary_operation fails.
5015 2018-04-10 Martin Liska <mliska@suse.cz>
5017 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
5018 cgraph_edge and ipa_ref.
5020 2018-04-10 Jakub Jelinek <jakub@redhat.com>
5024 * config/i386/sse.md
5025 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
5026 computation of the VEC_MERGE selector from mask.
5027 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
5028 Fix decoding of the VEC_MERGE selector into mask.
5030 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5032 PR tree-optimization/85286
5033 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
5035 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
5037 * final.c (final_1): Set insn_last_address as well as
5038 insn_current_address.
5040 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5043 * explow.c (emit_stack_probe): Call validize_mem on memory location
5044 before passing it to gen_probe_stack. Create address operand and
5045 legitimize it for the probe_stack_address case.
5047 2018-04-09 Jan Hubicka <jh@suse.cz>
5050 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
5051 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
5052 * tree.c (free_lang_data_in_type): Fix handling of binfos;
5054 (free_lang_data): Rebuild type inheritance graph.
5056 2018-04-09 Martin Sebor <msebor@redhat.com>
5058 * invoke.texi (-finline-small-functions): Mention other optimization
5060 (-findirect-inlining, -fpartial-inlining): Same.
5061 (-finline-functions-called-once): Same.
5062 (-freorder-blocks-and-partition): Same.
5064 2018-04-09 Jan Hubicka <jh@suse.cz>
5067 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
5068 jumps; choose last target that matches the criteria (i.e.
5069 no partition changes for non-crossing jumps).
5070 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
5071 support for redirecting crossing jumps to non-crossing.
5073 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
5075 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
5076 also for naked functions.
5078 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
5080 * config/arc/arc.md (add_shift): New pattern.
5081 (add_shift2): Likewise.
5082 (sub_shift): Likewise.
5083 (sub_shift_cmp0_noout): Likewise.
5084 (compare_si_ashiftsi): Likewise.
5085 (xbfu_cmp0_noout): New combine pattern.
5086 (xbfu_cmp0"): Likewise.
5087 (movsi_set_cc_insn): Place the predicable variant first.
5088 (commutative_binary_cmp0_noout): Remove clobber.
5089 (commutative_binary_cmp0): New pattern.
5090 (noncommutative_binary_cmp0): Likewise.
5091 (noncommutative_binary_cmp0_noout): Likewise.
5092 (noncommutative_binary_comparison_result_used): Removed.
5093 (rsub_cmp0): New pattern.
5094 (rsub_cmp0_noout): Likewise.
5095 (extzvsi): Changed, keep only meaningful variants.
5096 (SQH, SEZ): New iterators.
5097 (SQH_postfix): New mode attribute.
5098 (SEZ_prefix): New code attribute.
5099 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5100 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5101 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5103 (noncommutative_operator): Check the availability of barrel
5106 2018-04-09 Richard Biener <rguenther@suse.de>
5108 PR tree-optimization/85284
5109 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5110 Only use the niter constraining form of simple_iv when the exit
5113 2018-04-09 Tom de Vries <tom@codesourcery.com>
5116 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5117 (define_expand "*memory_barrier"): New define_expand.
5118 (define_insn "memory_barrier"): New insn.
5120 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5122 PR rtl-optimization/80463
5123 PR rtl-optimization/83972
5124 PR rtl-optimization/83480
5126 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5127 correct producer for the insn.
5128 (tidy_control_flow): Fixup seqnos in case of debug insns.
5130 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5132 PR rtl-optimization/83913
5134 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5135 different sched-times when merging exprs.
5137 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5139 PR rtl-optimization/83962
5141 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5142 tidy_fallthru_edge and tidy_control_flow.
5144 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5146 PR rtl-optimization/83530
5148 * sel-sched.c (force_next_insn): New global variable.
5149 (remove_insn_for_debug): When force_next_insn is true, also leave only
5150 next insn in the ready list.
5151 (sel_sched_region): When the region wasn't scheduled, make another pass
5152 over it with force_next_insn set to 1.
5154 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5156 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5158 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5159 for interrupt control.
5160 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5161 functions for interrupt control.
5162 * config/nds32/nds32-intrinsic.md: Likewise.
5163 * config/nds32/nds32_intrinsic.h: Likewise.
5164 * config/nds32/nds32.h (nds32_builtins): Likewise.
5166 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5168 * config/nds32/nds32.c (nds32_init_machine_status,
5169 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5170 strict_aligned_p field.
5171 (nds32_expand_to_rtl_hook): New function.
5172 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5173 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5175 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5176 Chung-Ju Wu <jasonwucj@gmail.com>
5178 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5179 * config/nds32/nds32-n7.md: New file.
5180 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5181 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5183 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5184 * config/nds32/nds32.md (pipeline_model): Add n7.
5185 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5186 * config/nds32/pipelines.md: Include n7 settings.
5188 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5189 Chung-Ju Wu <jasonwucj@gmail.com>
5191 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5192 * config/nds32/nds32-e8.md: New file.
5193 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5194 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5196 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5197 * config/nds32/nds32.md (pipeline_model): Add e8.
5198 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5199 * config/nds32/pipelines.md: Include e8 settings.
5201 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5202 Chung-Ju Wu <jasonwucj@gmail.com>
5204 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5205 * config/nds32/nds32-n8.md: New file.
5206 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5207 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5209 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5210 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5211 * config/nds32/nds32.md (pipeline_model): Add n8.
5212 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5213 * config/nds32/pipelines.md: Include n8 settings.
5215 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5216 Chung-Ju Wu <jasonwucj@gmail.com>
5218 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5219 * config/nds32/nds32-n9-2r1w.md: New file.
5220 * config/nds32/nds32-n9-3r2w.md: New file.
5221 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5222 nds32_register_ports): New or modify for cpu n9.
5223 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5225 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5226 * config/nds32/nds32-utils.c: New file.
5227 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5228 TARGET_MUL_SLOW): Define.
5229 * config/nds32/nds32.md (pipeline_model): New attribute.
5230 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5231 New options that support cpu n9.
5232 * config/nds32/pipelines.md: Include n9 settings.
5233 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5235 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5237 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5238 information if necessary.
5239 (output_cond_branch_compare_zero): Likewise.
5240 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5241 (nds32_target_alignment): Refine for alignment.
5242 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5243 (FUNCTION_BOUNDARY): Modify.
5244 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5246 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5248 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5250 * config/nds32/constants.md (unspec_volatile_element): Add values for
5251 TLB operation and data prefetch.
5252 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5253 functions for TLB operation and data prefetch.
5254 * config/nds32/nds32-intrinsic.md: Likewise.
5255 * config/nds32/nds32_intrinsic.h: Likewise.
5256 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5257 (nds32_print_operand): Likewise.
5258 * config/nds32/nds32.h (nds32_builtins): Likewise.
5260 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5261 Andrew Pinski <pinsika@gcc.gnu.org>
5264 * match.pd: Use constant_boolean_node of correct type instead of
5265 boolean_true_node or boolean_false_node for simplifying
5266 pointer comparisons to zero.
5268 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5270 PR tree-optimization/80021
5271 * tree.c (verify_type_variant): Make error call in verify_variant_match
5272 translatable and remove final full stop.
5274 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5276 * config/nds32/constants.md (unspec_volatile_element): Add
5277 UNSPEC_VOLATILE_EH_RETURN.
5278 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5279 nds32_output_stack_pop): Support dwarf exception handling process.
5280 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5281 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5282 exception handling process.
5283 (nds32_compute_stack_frame): Likewise.
5284 (nds32_return_addr_rtx): Likewise.
5285 (nds32_initial_elimination_offset): Likewise.
5286 (nds32_expand_prologue): Likewise.
5287 (nds32_expand_epilogue): Likewise.
5288 (nds32_dynamic_chain_address): New function.
5289 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5291 (DYNAMIC_CHAIN_ADDRESS): Define.
5292 (EH_RETURN_DATA_REGNO): Define.
5293 (EH_RETURN_STACKADJ_RTX): Define.
5294 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5295 patterns for dwarf exception handling.
5297 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5299 * config/nds32/nds32.h: Clean up obsolete macros.
5301 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5303 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5304 Add enum values for particular instructions.
5305 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5306 particular intrinsic functions.
5307 * config/nds32/nds32-intrinsic.md: Likewise.
5308 * config/nds32/nds32_intrinsic.h: Likewise.
5309 * config/nds32/nds32.h (nds32_builtins): Likewise.
5310 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5311 (btst, ave): New patterns for particular instructions.
5313 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5315 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5316 Add enum values for atomic load/store and memory sync.
5317 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5319 * config/nds32/nds32-intrinsic.md: Likewise.
5320 * config/nds32/nds32_intrinsic.h: Likewise.
5321 * config/nds32/nds32.h (nds32_builtins): Likewise.
5323 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5325 PR tree-optimization/85257
5326 * fold-const.c (native_encode_vector): If not all elts could fit
5327 and off is -1, return 0 rather than offset.
5328 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5329 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5330 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5331 adjust buffer in native_interpret_expr call.
5333 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5335 * config/nds32/constants.md (unspec_volatile_element): Add cache
5336 control enum values.
5337 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5338 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5339 * config/nds32/nds32.c (nds32_cctl_names): New.
5340 (nds32_print_operand): Handle cache control register names.
5341 * config/nds32/nds32.h (nds32_builtins): New enum values.
5342 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5344 * config/nds32/nds32.md (type): Add mmu.
5345 * config/nds32/pipelines.md (simple_insn): Add mmu.
5347 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5349 * config/nds32/nds32.md (type): Remove call.
5350 * config/nds32/pipelines.md (simple_insn): Likewise.
5352 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5354 * config/nds32/constants.md (unspec_volatile_element): Add
5355 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5356 UNSPEC_VOLATILE_FMFCFG.
5357 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5358 description for fmfcfg and fmfcsr.
5359 (bdesc_1arg): Add fmtcsr.
5360 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5361 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5362 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5363 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5364 unspec_fmfcfg): New patterns.
5365 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5366 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5367 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5368 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5369 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5370 __nds32__fmfcfg): Define.
5372 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5374 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5375 intrinsic register names.
5376 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5377 intrinsic register enum values and macros.
5379 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5381 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5382 for load/store addressing form.
5383 (nds32_print_operand_address): Likewise.
5385 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5388 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5389 based on LABEL_REF. Remove useless assertion.
5390 (pic_address_needs_scratch): Fix formatting.
5391 (sparc_legitimize_pic_address): Minor tweaks.
5392 (sparc_delegitimize_address): Adjust assertion accordingly.
5393 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5394 into symbolic_operand.
5395 (movsi_high_pic_label_ref): Likewise.
5396 (movsi_lo_sum_pic_label_ref): Likewise.
5397 (movdi_pic_label_ref): Likewise.
5398 (movdi_high_pic_label_ref): Likewise.
5399 (movdi_lo_sum_pic_label_ref): Likewise.
5401 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5403 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5404 custom LIB_SPEC setup.
5406 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5407 Kito Cheng <kito.cheng@gmail.com>
5409 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5410 * config/riscv/freebsd.h: New.
5412 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5414 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5415 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5418 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5419 Kito Cheng <kito.cheng@gmail.com>
5421 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5422 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5423 * config/nds32/nds32-protos.h (nds32_output_call,
5424 nds32_output_return): Declare.
5425 * config/nds32/nds32.md: Refine all the call and return patterns.
5427 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5430 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5431 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5433 PR rtl-optimization/84872
5434 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5435 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5438 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5440 * expr.c (copy_blkmode_to_reg): Revert 254862.
5441 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5443 2018-04-06 Richard Biener <rguenther@suse.de>
5446 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5447 after seeing a component reference with an adjacent field. Treat
5448 refs to arrays at struct end of external decls similar to
5449 refs to unconstrained commons.
5451 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5454 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5455 look through SAVE_EXPRs with non-side-effects argument. Adjust
5457 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5460 2018-04-06 Richard Biener <rguenther@suse.de>
5463 * alias.c (find_base_term): New wrapper around find_base_term
5464 unwinding CSELIB_VAL_PTR changes.
5465 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5468 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5470 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5472 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5473 constant definitions.
5474 ("nop"): lr 0,0 -> nopr r0
5475 ("nop_lr0", "nop_lr1"): New insn definitions.
5477 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5479 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5480 NDS32_V3PUSH_AVAILABLE_P macro.
5482 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5483 Chung-Ju Wu <jasonwucj@gmail.com>
5485 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5486 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5487 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5488 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5489 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5490 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5491 * config/nds32/constraints.md: New constraints and checking for hard
5492 float configuration.
5493 * config/nds32/iterators.md: New mode iterator and attribute for hard
5494 float configuration.
5495 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5497 * config/nds32/nds32-fpu.md: New file.
5498 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5499 deal with hard float code generation.
5500 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5502 (abi_type, float_reg_number): New enum type.
5503 * config/nds32/nds32-predicates.c: New predicates for hard float.
5504 * config/nds32/nds32-protos.h: Declare functions for hard float.
5505 * config/nds32/nds32.c: Implementation for hard float configuration.
5506 * config/nds32/nds32.h: Definitions for hard float configuration.
5507 * config/nds32/nds32.md: Include hard float machine description and
5508 modify patterns for hard float configuration.
5509 * config/nds32/nds32.opt: New options for hard float configuration.
5510 * config/nds32/predicates.md: New predicates for hard float
5513 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5515 * common/config/nds32/nds32-common.c
5516 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5518 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5521 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5522 CONSTRUCTOR_ELT (ctor, ...)->value.
5524 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5527 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5529 2018-04-05 Tom de Vries <tom@codesourcery.com>
5532 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5535 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5536 Kito Cheng <kito.cheng@gmail.com>
5538 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5539 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5540 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5541 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5543 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5544 Kito Cheng <kito.cheng@gmail.com>
5546 * config/nds32/constraints.md (Ufe): New memory constraint.
5547 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5548 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5549 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5551 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5552 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5554 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5556 * config/nds32/nds32.md: Use optimize_size in the condition for
5557 alu-shift instructions.
5559 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5561 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5563 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5565 * config/nds32/nds32.md (negsi2): Refine pattern.
5567 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5568 Chung-Ju Wu <jasonwucj@gmail.com>
5570 * config/nds32/iterators.md (shift_rotate): New code iterator.
5571 (shift): New code attribute.
5572 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5573 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5574 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5575 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5576 bit-wise operations.
5577 (andsi3, *andsi3): Ditto.
5578 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5579 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5580 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5581 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5582 nds32_ior_operand, nds32_xor_operand): New predicates.
5584 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5586 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5587 (addsi3, subsi3): ... this.
5589 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5591 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5593 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5595 * config/nds32/nds32.md: Adjust indention.
5597 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5599 * config/nds32/nds32.md (feature): New attribute.
5601 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5603 * config/nds32/nds32.md (subtype): New attribute.
5605 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5608 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5609 expansion to perform a bitwise AND of the argument followed by a
5610 boolean negation of the result.
5612 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5614 PR rtl-optimization/84878
5615 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5616 the basic block. Assert the use reference is not artificial and that
5617 it has an associated insn.
5619 2018-04-04 Michael Matz <matz@suse.de>
5621 * builtins.c (compute_objsize): Pass correct operand
5622 to array_at_struct_end_p.
5624 2018-04-04 Richard Biener <rguenther@suse.de>
5627 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5628 from contexts for DINFO_LEVEL_TERSE and below.
5630 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5632 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5633 resiter_operand condition.
5634 * config/nds32/nds32.md (*move<mode>): Ditto.
5636 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5637 Monk Chiang <sh.chiang04@gmail.com>
5639 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5641 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5643 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5645 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5646 Kito Cheng <kito.cheng@gmail.com>
5648 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5649 nds32_cond_code_str, output_cond_branch,
5650 output_cond_branch_compare_zero, nds32_expand_cbranch,
5651 nds32_expand_cstore, nds32_expand_movcc,
5652 nds32_output_cbranchsi4_equality_zero,
5653 nds32_output_cbranchsi4_equality_reg,
5654 nds32_output_cbranchsi4_equality_reg_or_const_int,
5655 nds32_output_cbranchsi4_greater_less_zero: New functions.
5656 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5657 nds32_expand_cstore, nds32_expand_movcc,
5658 nds32_output_cbranchsi4_equality_zero,
5659 nds32_output_cbranchsi4_equality_reg,
5660 nds32_output_cbranchsi4_equality_reg_or_const_int,
5661 nds32_output_cbranchsi4_greater_less_zero): Declare.
5662 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5663 nds32_rimm11s_operand): New predicates.
5664 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5665 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5668 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5670 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5671 * config/nds32/nds32.md: Ditto.
5672 * config/nds32/pipelines.md: Ditto.
5674 2018-04-04 Richard Biener <rguenther@suse.de>
5676 PR tree-optimization/85168
5677 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5678 propagating abnormals.
5680 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5682 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5684 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5685 Kito Cheng <kito.cheng@gmail.com>
5687 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5688 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5689 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5690 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5691 * config/nds32/nds32.md (sibcall_internal): New.
5692 (sibcall_register): Remove.
5693 (sibcall_immediate): Remove.
5694 (sibcall_value_internal): New.
5695 (sibcall_value_register): Remove.
5696 (sibcall_value_immediate): Remove.
5697 * config/nds32/predicates.md (nds32_general_register_operand): New.
5698 (nds32_call_address_operand): New.
5700 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5702 PR rtl-optimization/85167
5703 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5704 bb_defs if *split_p, instead preinitialize it to NULL.
5706 PR tree-optimization/85156
5707 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5708 evaluating the argument multiple times.
5710 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5712 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5714 (_mm_cvtpd_ps): Likewise.
5715 (_mm_cvttpd_epi32): Likewise.
5716 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5717 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5718 vector, pixel, and bool following altivec.h include.
5720 2018-04-03 Martin Sebor <msebor@redhat.com>
5722 * doc/extend.texi (Common Function Attributes): Clarify.
5723 (const attribute): Likewise.
5724 (pure attribute): Likewise.
5726 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5729 * config/i386/i386.c (ix86_expand_vector_set): Use
5730 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5732 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5734 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5735 instructions when changing rounding bits to preserve precision bits
5736 in the x87 control word.
5738 2018-04-03 Martin Liska <mliska@suse.cz>
5740 PR tree-optimization/82491
5741 * rtl.h (strip_offset_and_add): Replace += suboffset with
5742 poly_uint64 () + suboffset.
5744 2018-03-29 Martin Liska <mliska@suse.cz>
5745 Martin Jambor <mjambor@suse.cz>
5748 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5749 param_type is not an integral or pointer type.
5751 2018-04-03 Richard Biener <rguenther@suse.de>
5753 * sese.h (recompute_all_dominators): Remove.
5755 2018-04-02 Martin Sebor <msebor@redhat.com>
5757 * doc/invoke.texi (-Wrestrict): Fix typos.
5759 2018-04-02 Jim Wilson <jimw@sifive.com>
5761 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5762 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5763 (<optab>di3, <optab>si3_extend): Likewise.
5764 (<optab>si3_mask, <optab>si3_mask_1): New.
5765 (<optab>di3_mask, <optab>di3_mask_1): New.
5766 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5767 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5768 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5770 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5772 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5775 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5777 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5778 (nds32_canonicalize_comparison): New function.
5780 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5781 Kito Cheng <kito.cheng@gmail.com>
5782 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5784 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5785 * config/nds32/constants.md (unspec_volatile_element): Add
5786 UNSPEC_VOLATILE_RELAX_GROUP.
5787 * config/nds32/nds32-relax-opt.c: New file.
5788 * config/nds32/nds32-predicates.c
5789 (nds32_symbol_load_store_p): New function.
5790 * config/nds32/nds32-protos.h
5791 (nds32_symbol_load_store_p): Declare function.
5792 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5793 * config/nds32/nds32.c
5794 (nds32_register_pass): New function to register pass.
5795 (nds32_register_passes): New function to register passes.
5796 * config/nds32/nds32.md (relax_group): New pattern.
5797 * config/nds32/nds32.opt (mrelax-hint): New option.
5798 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5800 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5802 * config/nds32/t-nds32: Modify files dependency.
5804 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5806 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5807 (PROFILE_HOOK): Define its implementation.
5809 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5811 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5812 type and 32-bit size.
5814 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5817 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5818 (V_128_256): New mode iterator.
5819 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5820 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5821 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5823 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5824 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5826 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5829 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5830 NaN inputs correctly.
5832 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5835 * config/rs6000/vsx.md (??r): New mode attribute.
5836 (*vsx_mov<mode>_64bit): Use it.
5837 (*vsx_mov<mode>_32bit): Likewise.
5839 2018-03-30 Martin Sebor <msebor@redhat.com>
5841 PR tree-optimization/84818
5842 * builtins.c (check_access): Use warning_n.
5844 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5847 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5849 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5852 2018-03-30 Julia Koval <julia.koval@intel.com>
5855 * x86-tune.def (movx, partial_reg_dependency): Enable for
5858 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5861 * lra-constraints.c (process_alt_operands): Move setting
5862 this_alternative_matches below.
5864 2018-03-29 Martin Liska <mliska@suse.cz>
5867 * doc/invoke.texi: Document how LTO works with debug info.
5868 Describe auto-load support of binutils. Mention 'x86-64'
5869 as valid option value of -march option.
5871 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5873 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5876 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5877 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5878 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5881 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5884 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5885 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5886 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5887 for RS6000_BTM_POWERPC64.
5888 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5889 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5890 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5895 2018-03-28 Carl Love <cel@us.ibm.com>
5898 2017-09-27 Carl Love <cel@us.ibm.com>
5900 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5901 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5902 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5905 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5907 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5908 instead of __vector bool.
5909 (_mm_max_pu8): Likewise.
5910 (_mm_min_pi16): Likewise.
5912 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5915 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5916 (DIVWEUO): Likewise.
5918 (DIVDEUO): Likewise.
5919 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5920 DIVWEUO and DIVDEUO.
5921 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5922 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5923 (div_extend): Likewise.
5924 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5926 (__builtin_divweuo): Likewise.
5927 (__builtin_divdeo): Likewise.
5928 (__builtin_divdeuo): Likewise.
5930 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5933 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5934 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5936 PR tree-optimization/82004
5937 * gimple-match-head.c (optimize_pow_to_exp): New function.
5938 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5939 Don't fold to exp if optimize_pow_to_exp is false.
5941 2018-03-28 Martin Liska <mliska@suse.cz>
5944 * calls.c (initialize_argument_information): Fix trailing space.
5945 * common.opt: Fix typo and provide better explanation for
5946 -fsanitize-coverage option.
5947 * config/i386/i386.opt: Fix typo.
5949 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5950 Martin Liska <mliska@suse.cz>
5953 * gimplify.c (asan_poison_variable): Don't do the check for
5954 gimplify_omp_ctxp here.
5955 (gimplify_decl_expr): Do it here.
5956 (gimplify_target_expr): Likewise.
5958 2018-03-28 Martin Liska <mliska@suse.cz>
5961 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5962 chkp_type_bounds_count if MPX is not enabled.
5964 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5966 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5968 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5971 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5972 function to create the function decl for complex long double
5973 multiply and divide for -mabi=ieeelongdouble.
5974 (init_float128_ieee): Call it.
5976 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5979 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5980 -fcf-protection=branch -mibt.
5981 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5983 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5986 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5987 arm_word_relocations.
5989 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5992 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5993 extern array declarations.
5995 2018-03-27 Richard Biener <rguenther@suse.de>
5998 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5999 explicit single_use checks.
6001 2018-03-27 Richard Biener <rguenther@suse.de>
6003 PR tree-optimization/85082
6004 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
6007 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6009 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
6010 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
6011 Turn on fasynchronous-unwind-tables and funwind-tables.
6013 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
6016 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
6017 (*bmi_blsr_<mode>_ccz): Ditto.
6019 2018-03-26 Tom de Vries <tom@codesourcery.com>
6021 PR tree-optimization/85063
6022 * omp-general.c (offloading_function_p): New function. Factor out
6024 * omp-offload.c (pass_omp_target_link::gate): ... here.
6025 * omp-general.h (offloading_function_p): Declare.
6026 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
6027 with attribute omp declare target for offloading functions.
6029 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
6031 PR tree-optimization/84005
6032 * tree-data-ref.h (get_base_for_alignment): Declare.
6033 * tree-data-ref.c (get_base_for_alignment_1): New function.
6034 (get_base_for_alignment): Likewise.
6035 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
6036 get_base_for_alignment to find a suitable base object, instead
6037 of always using drb->base_address.
6039 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6042 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
6043 known size by default.
6045 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
6048 * lra-constraints.c (process_alt_operands): Don't match BLKmode
6049 and non BLKmode operands.
6051 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6054 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
6055 Clean up attributes.
6057 2018-03-23 Richard Biener <rguenther@suse.de>
6060 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
6061 we are going to emit early debug for LTO.
6063 2018-03-23 Jakub Jelinek <jakub@redhat.com>
6066 * function.c (match_asm_constraints_1): Don't optimize if input
6067 doesn't satisfy general_operand predicate for output's mode.
6070 * alias.c (write_dependence_p): Don't require for x_canonicalized
6071 non-VOIDmode if x has VOIDmode.
6074 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
6075 just don't try to optimize it rather than assert it never happens.
6077 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6079 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
6080 macro expansions for definition of ST_INTERNAL_<mode> and
6081 LD_INTERNAL_<mode> builtins.
6082 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
6084 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
6086 (altivec_expand_st_builtin): Likewise.
6087 (altivec_expand_builtin): Remove calls to deleted functions.
6088 (rs6000_address_for_altivec): Delete this function.
6089 * config/rs6000/vector.md: Remove expands for
6090 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6092 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6095 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6096 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6097 re-computing once computed.
6098 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6099 (arm_init_machine_status): Initialize
6100 machine->static_chain_stack_bytes.
6102 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6105 * doc/extend.texi: Add four new prototypes for vec_ld.
6106 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6107 definitions for more logical presentation.
6108 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6109 entries for V1TI variants of __builtin_altivec_ld builtin.
6110 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6111 handling of V1TI variant of LVX icode pattern.
6112 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6113 (rs6000_gimple_fold_builtin): Likewise.
6114 (altivec_init_builtins): Add code to define
6115 __builtin_altivec_lvx_v1ti function.
6117 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6120 * function.c (match_asm_constraints_1): Don't do the optimization
6121 if input isn't a REG, SUBREG, MEM or constant.
6123 2018-03-22 Tom de Vries <tom@codesourcery.com>
6125 PR tree-optimization/84956
6126 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6127 bb_has_abnormal_pred.
6129 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6132 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6133 DECL_INITIAL (decl) to decl at the end.
6134 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6137 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6139 * doc/extend.texi (__builtin_tgmath): Document when complex
6140 integer types are treated as _Complex _Float64.
6142 2018-03-21 Tom de Vries <tom@codesourcery.com>
6144 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6146 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6148 PR tree-optimization/84960
6149 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6150 if it is ENTRY block, move them into single succ of ENTRY in that case.
6152 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6154 PR tree-optimization/84811
6155 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6156 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6157 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6158 (known_subrange_p): Update accordingly. Cast each value involved
6159 in the size comparison, rather than casting the result of the
6162 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6164 PR tree-optimization/84982
6165 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6166 by flipping the least significant bit rather than all bits from
6167 bitpos to bitpos + bitsize - 1.
6169 2018-03-21 Nathan Sidwell <nathan@acm.org>
6171 * doc/extend.texi (Deprecated Features): Remove mention of
6172 long-deleted deprecations.
6174 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6177 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6178 * configure: Regenerate.
6180 2018-03-21 Tom de Vries <tom@codesourcery.com>
6182 PR tree-optimization/83126
6183 * tree-parloops.c (num_phis): New function.
6184 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6186 2018-03-21 Nathan Sidwell <nathan@acm.org>
6188 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6189 mention anon-struct/union members and trailing attributes.
6191 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6193 PR tree-optimization/84969
6194 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6195 builtin memset partitions if they set different rhs values.
6197 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6199 PR rtl-optimization/84989
6200 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6201 VEC_DUPLICATE with scalar result mode.
6203 2018-03-21 Martin Liska <mliska@suse.cz>
6206 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6207 not intended return statement.
6209 2018-03-21 Martin Liska <mliska@suse.cz>
6212 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6213 (chkp_find_bound_slots_1): Limit number of iterations.
6215 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6218 * Minor grammar fixes for x86 options.
6220 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6223 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6224 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6226 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6229 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6230 (altivec_lvx_<mode>_1op): Likewise.
6231 (altivec_stvx_<mode>_2op): Likewise.
6232 (altivec_stvx_<mode>_1op): Likewise.
6233 (altivec_lvx_<VM2:mode>): New define_expand.
6234 (altivec_stvx_<VM2:mode>): Likewise.
6235 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6236 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6237 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6238 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6239 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6240 (rs6000_gen_lvx): Likewise.
6241 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6242 (altivec_expand_stv_builtin): Likewise.
6243 (altivec_expand_builtin): Likewise.
6244 * config/rs6000/vector.md: Likewise.
6246 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6249 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6252 2018-03-20 Richard Biener <rguenther@suse.de>
6255 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6256 sign-conversions as zero, fall back to standard scalar_stmt
6259 2018-03-20 Martin Liska <mliska@suse.cz>
6262 * predict.c (rebuild_frequencies): Handle case when we have
6263 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6265 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6268 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6269 flag_section_anchors.
6270 * varasm.c (use_blocks_for_decl_p): Remove hack for
6271 dw2_force_const_mem.
6274 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6276 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6277 be created, use lowpart_subreg of operands[0] rather than operands[0]
6279 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6280 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6281 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6282 and n constraint instead of aarch64_shift_imm_di and Usd.
6283 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6284 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6286 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6289 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6290 to favor GPR over NEON registers.
6291 (<shift>di3_neon): Likewise.
6293 2018-03-20 Tom de Vries <tom@codesourcery.com>
6296 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6297 (nvptx_process_pars): Emit bar.sync asap and alap.
6299 2018-03-20 Tom de Vries <tom@codesourcery.com>
6302 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6303 seen_label if seen_label is already set.
6305 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6308 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6309 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6310 Use 1U instead of 1. Formatting fixes.
6313 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6314 instead of TREE_TYPE (s1) for the return value.
6316 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6318 PR tree-optimization/84946
6319 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6320 bitsize + bitsize in poly_uint64 rather than poly_int64.
6323 * dwarf2asm.c: Include fold-const.c.
6324 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6325 of decl rather than decl itself.
6327 PR rtl-optimization/84643
6328 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6330 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6333 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6334 calling assemble_variable.
6336 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6339 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6340 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6342 2018-03-19 Jim Wilson <jimw@sifive.com>
6345 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6346 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6347 (riscv_first_stack_step): Likewise.
6348 (riscv_option_override): Use STACK_BOUNDARY instead of
6350 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6352 (BIGGEST_ALIGNMENT): Set to 128.
6353 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6354 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6357 2018-03-19 Richard Biener <rguenther@suse.de>
6359 PR tree-optimization/84933
6360 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6361 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6363 2018-03-19 Richard Biener <rguenther@suse.de>
6365 PR tree-optimization/84859
6366 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6367 (cond_if_else_store_replacement): Perform sinking operation on
6368 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6369 Generalize what a BB with a single eligible store is.
6371 2018-03-19 Richard Biener <rguenther@suse.de>
6373 PR tree-optimization/84929
6374 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6375 chrec_is_positive against non-chrec arg.
6377 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6380 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6382 2018-03-18 Martin Liska <mliska@suse.cz>
6384 PR rtl-optimization/84635
6385 * regrename.c (build_def_use): Use matches_mode only when
6388 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6390 PR tree-optimization/84913
6391 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6392 vectorize chains of COND_EXPRs.
6394 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6396 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6398 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6400 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6402 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6404 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6406 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6407 Kito Cheng <kito.cheng@gmail.com>
6409 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6410 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6411 (nds32_adjust_reg_alloc_order): New function.
6412 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6414 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6416 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6417 nds32_print_operand, nds32_print_operand_address): Use
6418 HOST_WIDE_INT_PRINT_DEC instead.
6420 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6422 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6424 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6427 * config/i386/i386.c (initial_ix86_tune_features,
6428 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6430 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6431 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6432 rather than 1u << ix86_tune. Formatting fix.
6433 (ix86_option_override_internal): Change ix86_arch_mask from
6434 unsigned int to unsigned HOST_WIDE_INT, initialize to
6435 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6436 (ix86_function_specific_restore): Likewise.
6438 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6441 * postreload.c (reload_combine_recognize_pattern): Perform
6442 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6443 truncate_int_for_mode the result for the destination's mode.
6446 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6447 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6449 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6452 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6455 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6456 regno_allocno_class_array and sorted_pseudos.
6457 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6458 insns where regno is used.
6460 2018-03-16 Martin Liska <mliska@suse.cz>
6463 * multiple_target.c (create_dispatcher_calls): Redirect
6464 reference in the symbol table.
6466 2018-03-16 Martin Liska <mliska@suse.cz>
6469 * multiple_target.c (create_dispatcher_calls): Redirect also
6472 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6476 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6477 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6478 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6480 2018-03-16 Julia Koval <julia.koval@intel.com>
6482 * doc/invoke.texi (Skylake Server): Add CLWB.
6483 Cannonlake): Remove CLWB.
6485 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6487 PR tree-optimization/84841
6488 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6490 (FLOAT_ONE_CONST_TYPE): Define.
6491 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6492 (sort_by_operand_rank): Put entries with higher constant_type last
6493 rather than first to match comments.
6495 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6497 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6500 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6503 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6504 check and use address of target always.
6506 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6509 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6510 (indirect_thunk_bnd_needed): Likewise.
6511 (indirect_thunks_used): Likewise.
6512 (indirect_thunks_bnd_used): Likewise.
6513 (indirect_return_needed): New.
6514 (indirect_return_bnd_needed): Likewise.
6515 (output_indirect_thunk_function): Add a bool argument for
6517 (output_indirect_thunk_function): Don't generate alias for
6518 function return thunk.
6519 (ix86_code_end): Call output_indirect_thunk_function to generate
6520 function return thunks.
6521 (ix86_output_function_return): Set indirect_return_bnd_needed
6522 and indirect_return_needed instead of indirect_thunk_bnd_needed
6523 and indirect_thunk_needed.
6525 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6527 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6528 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6529 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6531 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6532 Paul Hua <paul.hua.gm@gmail.com>
6535 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6537 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6539 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6540 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6543 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6546 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6547 instead of GET_MODE_SIZE when comparing Units.
6549 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6552 * varasm.c (hash_section): Return an unchangeble hash value
6553 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6554 Return !aarch64_can_use_per_function_literal_pools_p ().
6556 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6559 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6560 rather than address of cmode as last argument to prepare_cmp_insn.
6562 2018-03-15 Julia Koval <julia.koval@intel.com>
6564 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6565 F_AVX512VNNI, F_AVX512BITALG): New.
6567 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6570 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6571 insn for floating-point loads and stores.
6573 2018-03-14 Carl Love <cel@us.ibm.com>
6575 * config/rs6000/rs6000-c.c: Add macro definitions for
6576 ALTIVEC_BUILTIN_VEC_PERMXOR.
6577 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6578 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6579 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6580 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6582 * config/doc/extend.texi: Add prototypes for vec_permxor.
6584 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6587 * diagnostic-show-locus.c (class layout_point): Convert m_line
6588 from int to linenum_type.
6589 (line_span::comparator): Use linenum "compare" function when
6590 comparing line numbers.
6591 (test_line_span): New function.
6592 (layout_range::contains_point): Convert param "row" from int to
6594 (layout_range::intersects_line_p): Likewise.
6595 (layout::will_show_line_p): Likewise.
6596 (layout::print_source_line): Likewise.
6597 (layout::should_print_annotation_line_p): Likewise.
6598 (layout::print_annotation_line): Likewise.
6599 (layout::print_leading_fixits): Likewise.
6600 (layout::annotation_line_showed_range_p): Likewise.
6601 (struct line_corrections): Likewise for field m_row.
6602 (line_corrections::line_corrections): Likewise for param "row".
6603 (layout::print_trailing_fixits): Likewise.
6604 (layout::get_state_at_point): Likewise.
6605 (layout::get_x_bound_for_row): Likewise.
6606 (layout::print_line): Likewise.
6607 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6608 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6609 * input.c (selftest::test_linenum_comparisons): New function.
6610 (selftest::input_c_tests): Call it.
6611 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6612 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6613 * selftest.h (ASSERT_GT): New macro.
6614 (ASSERT_GT_AT): New macro.
6615 (ASSERT_LT): New macro.
6616 (ASSERT_LT_AT): New macro.
6618 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6620 PR rtl-optimization/84780
6621 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6623 2018-03-14 Martin Liska <mliska@suse.cz>
6625 * tree.c (record_node_allocation_statistics): Use
6626 get_stats_node_kind.
6627 (get_stats_node_kind): New function extracted from
6628 record_node_allocation_statistics.
6629 (free_node): Use get_stats_node_kind.
6631 2018-03-14 Richard Biener <rguenther@suse.de>
6633 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6634 that the value-set of ANTIC_IN doesn't grow.
6637 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6639 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6640 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6642 2018-03-14 Julia Koval <julia.koval@intel.com>
6644 * config.gcc (icelake-client, icelake-server): New.
6646 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6647 (initial_ix86_arch_features): Ditto.
6648 (PTA_SKYLAKE): Add SGX.
6649 (PTA_ICELAKE): Remove.
6650 (PTA_ICELAKE_CLIENT): New.
6651 (PTA_ICELAKE_SERVER): New.
6652 (ix86_option_override_internal): Split up icelake on icelake client and
6654 (get_builtin_code_for_version): Ditto.
6655 (fold_builtin_cpu): Ditto.
6656 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6657 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6658 * config/i386/i386.h (processor_type): Ditto.
6659 * doc/invoke.texi: Ditto.
6661 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6664 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6665 INTEGER_CST offset, add it together with bitpos / 8 and
6666 sign extend based on POINTER_SIZE.
6670 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6673 * config/i386/constraints.md (Yc): New register constraint.
6674 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6675 Use Yc constraint for alternative 2 of operand 0. Remove
6676 preferred_for_speed attribute.
6678 2018-03-14 Richard Biener <rguenther@suse.de>
6680 PR tree-optimization/84830
6681 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6682 with the old one to avoid oscillations.
6684 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6687 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6689 (assign_by_spills): Return a flag of reload assignment failure.
6690 Do not process the reload assignment failures. Do not spill other
6691 reload pseudos if they has the same reg class. Update n if
6693 (lra_assign): Add a return arg. Set up from the result of
6694 assign_by_spills call.
6695 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6696 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6697 usage_insns if it is not NULL.
6698 (spill_hard_reg_in_range): New function.
6699 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6700 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6701 function prototypes.
6702 (lra_assign): Change prototype.
6703 * lra.c (lra): Add code to deal with fails by splitting hard reg
6706 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6708 * config/riscv/riscv.opt (mrelax): New option.
6709 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6710 "norelax" when riscv_mrelax is disabled.
6711 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6713 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6716 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6717 reassociation for int modes.
6719 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6721 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6722 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6724 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6725 * config/aarch64/aarch64-sve.md
6726 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6727 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6728 (*extend<mode><Vwide>2): Rename to...
6729 (aarch64_sve_extend<mode><Vwide>2): ...this.
6730 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6731 renaming the old pattern to...
6732 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6734 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6735 define_expand, renaming the old pattern to...
6736 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6737 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6738 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6739 account when deciding which SVE instruction the optab should use.
6740 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6742 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6744 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6745 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6746 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6747 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6748 (tlsdesc_small_<mode>): Turn a define_expand and use
6749 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6750 (tlsdesc_small_advsimd_<mode>): ...this.
6751 (tlsdesc_small_sve_<mode>): New pattern.
6753 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6755 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6756 (UNSPEC_UMUL_HIGHPART): New constants.
6757 (MUL_HIGHPART): New int iteraor.
6758 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6759 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6761 (*<su>mul<mode>3_highpart): New define_insn.
6763 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6766 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6769 2018-03-13 Martin Liska <mliska@suse.cz>
6772 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6774 (sem_item_optimizer::~sem_item_optimizer): Release it.
6775 (sem_item_optimizer::merge_classes): Register variable aliases.
6776 (sem_item_optimizer::fixup_pt_set): New function.
6777 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6778 * ipa-icf.h: Declare new variables and functions.
6780 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6783 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6784 integer_pow2p@2 and test integer_pow2p in condition.
6785 (A < 0 ? C : 0): Similarly for @1.
6788 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6789 characters starting at p contain '\0' character, don't look beyond
6793 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6794 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6797 * reg-stack.c (change_stack): Change update_end var from int to
6798 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6799 also call set_block_for_insn on the newly added insns and rescan.
6802 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6803 on the last operand.
6806 * tree.c (stabilize_reference_1): Return save_expr (e) for
6807 STATEMENT_LIST even if it doesn't have side-effects.
6809 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6811 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6813 2018-03-12 Renlin Li <renlin.li@arm.com>
6815 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6816 aarch64_output_scalar_simd_mov_immediate.
6818 2018-03-12 Martin Sebor <msebor@redhat.com>
6820 PR tree-optimization/83456
6821 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6822 for perfectly overlapping calls to memcpy.
6823 (gimple_fold_builtin_memory_chk): Same.
6824 (gimple_fold_builtin_strcpy): Handle no-warning.
6825 (gimple_fold_builtin_stxcpy_chk): Same.
6826 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6828 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6830 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6831 parameter. Use it for SFmode.
6832 (rs6000_function_arg_advance_1): Adjust.
6833 (rs6000_function_arg): Adjust.
6834 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6836 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6838 PR rtl-optimization/84169
6839 PR rtl-optimization/84780
6840 * combine.c (can_combine_p): Check for a 2-insn combination whether
6841 the destination register is used between the two insns, too.
6843 2018-03-12 Richard Biener <rguenther@suse.de>
6845 PR tree-optimization/84803
6846 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6847 for refs DR analysis didn't process.
6849 2018-03-12 Richard Biener <rguenther@suse.de>
6851 PR tree-optimization/84777
6852 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6853 force-vectorize loops ignore whether we are optimizing for size.
6855 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6857 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6858 (TARGET_MD_ASM_ADJUST): Define.
6860 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6861 Kito Cheng <kito.cheng@gmail.com>
6862 Chung-Ju Wu <jasonwucj@gmail.com>
6864 * config/nds32/nds32.c (nds32_compute_stack_frame,
6865 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6866 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6867 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6868 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6869 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6870 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6871 * config/nds32/nds32.md (prologue, epilogue): Use macro
6872 NDS32_V3PUSH_AVAILABLE_P to do checking.
6874 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6877 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6878 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6879 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6880 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6881 addition of most attributes on !orig_type_die or the attribute not
6882 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6884 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6885 Chung-Ju Wu <jasonwucj@gmail.com>
6887 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6889 * config/nds32/nds32.opt (mvh): New option.
6891 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6892 Chung-Ju Wu <jasonwucj@gmail.com>
6894 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6896 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6897 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6900 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6901 Chung-Ju Wu <jasonwucj@gmail.com>
6903 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6905 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6906 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6908 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6909 Kito Cheng <kito.cheng@gmail.com>
6910 Chung-Ju Wu <jasonwucj@gmail.com>
6912 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6913 UNSPEC_FFMISM and UNSPEC_FLMISM.
6914 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6915 for ffb, ffmism and flmism.
6916 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6917 (unspec_ffmism): Ditto.
6918 (unspec_flmism): Ditto.
6919 (nds32_expand_builtin_impl): Check if string extension is available.
6920 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6921 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6923 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6926 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6929 * lra-assigns.c (assign_by_spills): Return a flag of reload
6930 assignment failure. Do not process the reload assignment
6931 failures. Do not spill other reload pseudos if they has the same
6933 (lra_assign): Add a return arg. Set up from the result of
6934 assign_by_spills call.
6935 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6936 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6937 usage_insns if it is not NULL.
6938 (spill_hard_reg_in_range): New function.
6939 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6940 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6941 function prototypes.
6942 (lra_assign): Change prototype.
6943 * lra.c (lra): Add code to deal with fails by splitting hard reg
6946 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6949 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6951 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6954 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6955 (dw_val_node): Add val_symbolic_view.
6956 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6957 (symview_upper_bound): New.
6958 (new_line_info_table): Initialize symviews_since_reset.
6959 (dwarf2out_source_line): Count symviews_since_reset and set
6960 symview_upper_bound.
6961 (dw_val_equal_p): Handle symview.
6962 (add_AT_symview): New.
6963 (print_dw_val): Handle symview.
6964 (attr_checksum, attr_checksum_ordered): Likewise.
6965 (same_dw_val_p, size_of_die): Likewise.
6966 (value_format, output_die): Likewise.
6967 (add_high_low_attributes): Use add_AT_symview for entry_view.
6968 (dwarf2out_finish): Reset symview_upper_bound, clear
6971 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6974 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6975 Add strict argument and use it.
6976 (rs6000_split_multireg_move): Update for new strict argument.
6977 (mem_operand_gpr): Disallow all non-offsettable addresses.
6978 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6980 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6983 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6984 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6985 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6988 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6989 decl, use remap_type if we want to use the type.
6991 2018-03-09 Martin Sebor <msebor@redhat.com>
6993 PR tree-optimization/84526
6994 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6996 (builtin_access::generic_overlap): Be prepared to handle non-array
6999 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
7001 PR rtl-optimization/84682
7002 * lra-constraints.c (process_address_1): Check is_address flag
7003 for address constraints.
7004 (process_alt_operands): Likewise.
7005 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
7006 preprocess_constraints.
7007 * recog.h (preprocess_constraints): Add oploc parameter.
7009 * recog.c (preprocess_constraints): Test address_operand for
7010 CT_ADDRESS constraints.
7012 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
7015 * lra-assigns.c (assign_by_spills): Return a flag of reload
7016 assignment failure. Do not process the reload assignment
7017 failures. Do not spill other reload pseudos if they has the same
7019 (lra_assign): Add a return arg. Set up from the result of
7020 assign_by_spills call.
7021 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7022 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7023 usage_insns if it is not NULL.
7024 (spill_hard_reg_in_range): New function.
7025 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7026 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7027 function prototypes.
7028 (lra_assign): Change prototype.
7029 * lra.c (lra): Add code to deal with fails by splitting hard reg
7032 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7035 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
7036 Accept complain bool parameter. Only emit errors if it is true.
7037 (arm_parse_cpu_option_name): Likewise.
7038 (arm_target_thumb_only): Adjust callers of the above.
7039 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
7040 prototype to take a default true bool parameter.
7041 (arm_parse_arch_option_name): Likewise.
7043 2018-03-09 David Malcolm <dmalcolm@redhat.com>
7044 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
7048 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
7049 * configure: Regenerate.
7050 * configure.ac ("linker --version-script option"): New.
7051 ("linker soname option"): New.
7053 2018-03-09 Richard Biener <rguenther@suse.de>
7055 PR tree-optimization/84775
7056 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
7057 immediate uses of predicate stmts and mark them modified.
7060 PR tree-optimization/84178
7061 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7063 (version_loop_for_if_conversion): Delay update_ssa call.
7064 (tree_if_conversion): Delay update_ssa until after predicate
7067 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
7070 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
7071 when the function accesses prior frames.
7073 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7076 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
7077 gen_llsym, otherwise call maybe_gen_llsym.
7080 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
7081 has ',' character inside of it.
7083 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7086 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7087 as clobbering CC_REGNUM.
7089 2018-03-08 Richard Biener <rguenther@suse.de>
7092 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7093 (follow_copies_to_constant): Do not follow SSA names registered
7096 2018-03-08 Richard Biener <rguenther@suse.de>
7098 PR tree-optimization/84178
7099 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7101 (version_loop_for_if_conversion): Delay update_ssa call.
7102 (tree_if_conversion): Delay update_ssa until after predicate
7105 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7107 PR tree-optimization/84178
7108 * tree-if-conv.c (release_bb_predicate): Remove the
7109 the assertion that the stmts have NULL use_ops.
7110 Discard the statements, asserting that they haven't
7111 yet been added to a BB.
7113 2018-03-08 Richard Biener <rguenther@suse.de>
7115 PR tree-optimization/84746
7116 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7117 (phi_translate): Pass in destination ANTIC_OUT set.
7118 (phi_translate_1): Likewise. For a simplified result lookup
7119 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7120 (phi_translate_set): Adjust.
7121 (do_pre_regular_insertion): Likewise.
7122 (do_pre_partial_partial_insertion): Likewise.
7124 2018-03-08 Martin Liska <mliska@suse.cz>
7126 PR gcov-profile/84735
7127 * doc/gcov.texi: Document usage of profile files.
7128 * gcov-io.h: Document changes in the format.
7130 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7134 * dwarf2out.c (struct dw_line_info_table): Update comments for
7136 (FORCE_RESET_NEXT_VIEW): New.
7137 (FORCE_RESETTING_VIEW_P): New.
7138 (RESETTING_VIEW_P): Check for -1 too.
7139 (ZERO_VIEW_P): Likewise.
7140 (new_line_info_table): Force-reset next view.
7141 (dwarf2out_begin_function): Likewise.
7142 (dwarf2out_source_line): Simplify zero_view_p initialization.
7143 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7144 view directly. Omit view when omitting .loc at line 0.
7146 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7148 PR tree-optimization/84740
7149 * tree-switch-conversion.c (process_switch): Call build_constructors
7150 only if info.phi_count is non-zero.
7152 PR tree-optimization/84739
7153 * tree-tailcall.c (find_tail_calls): Check call arguments against
7154 DECL_ARGUMENTS (current_function_decl) rather than
7155 DECL_ARGUMENTS (func) when checking for tail recursion.
7157 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7159 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7160 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7161 Volker Reichelt's entry and add entries for people that perform
7162 GCC fuzzy testing and report numerous bugs.
7164 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7167 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7168 readonly data in sdata, if that is disabled.
7169 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7170 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7171 -mreadonly-in-sdata option.
7173 2018-03-07 Martin Sebor <msebor@redhat.com>
7175 PR tree-optimization/84468
7176 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7177 basic block when looking for nul assignment.
7179 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7182 * except.h (output_function_exception_table): Adjust prototype.
7183 * except.c (output_function_exception_table): Remove FNNAME parameter
7184 and add SECTION parameter. Ouput one part of the table at a time.
7185 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7186 the first part of the exception table and emit unwind directives.
7187 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7188 (i386_pe_seh_cold_init): Likewise.
7189 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7190 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7191 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7192 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7193 * config/i386/winnt.c: Include except.h.
7194 (struct seh_frame_state): Add reg_offset, after_prologue and
7195 in_cold_section fields.
7196 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7197 (i386_pe_seh_cold_init): New function.
7198 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7199 to seh->in_cold_section.
7200 (seh_emit_push): Record the offset of the push.
7201 (seh_emit_save): Record the offet of the save.
7202 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7203 Test seh->after_prologue to disregard the epilogue.
7204 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7205 (i386_pe_end_cold_function): New function.
7207 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7210 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7211 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7214 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7216 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7217 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7220 * multiple_target.c: Include tree-inline.h and intl.h.
7221 (expand_target_clones): Diagnose and fail if node->definition and
7222 !tree_versionable_function_p (node->decl).
7224 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7226 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7228 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7229 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7230 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7232 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7235 * combine.c (try_combine): Use reg_or_subregno instead of handling
7236 just paradoxical SUBREGs and REGs.
7238 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7240 * config/arc/arc.c (arc_finalize_pic): Remove function.
7241 (arc_must_save_register): We use single base PIC register, remove
7242 checks to save/restore the PIC register.
7243 (arc_expand_prologue): Likewise.
7244 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7246 (arc_verify_short): Likewise.
7247 (arc_attr_type): Likewise.
7248 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7249 (walk_stores): Likewise.
7250 (arc_address_cost): Make it static.
7251 (arc_verify_short): Likewise.
7252 (branch_dest): Likewise.
7253 (arc_attr_type): Likewise.
7254 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7255 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7256 (arc_final_prescan_insn): Remove inserting the nops due to
7257 hardware hazards. It is done in reorg step.
7258 (insn_length_variant_t): Remove.
7259 (insn_length_parameters_t): Likewise.
7260 (arc_insn_length_parameters): Likewise.
7261 (arc_get_insn_variants): Likewise.
7262 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7264 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7267 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7270 PR tree-optimization/84687
7271 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7273 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7275 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7277 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7278 Rename to ppc_speculation_barrier.
7279 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7280 __builtin_ppc_speculation_barrier.
7282 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7285 * combine.c (combine_simplify_rtx): Don't try to simplify if
7286 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7289 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7291 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7292 to 32 bytes when compiling for POWER9.
7294 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7297 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7298 regparm >= 3 with no arg reg available also for calls with
7299 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7302 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7304 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7306 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7309 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7311 2018-03-05 Richard Biener <rguenther@suse.de>
7313 PR tree-optimization/84486
7314 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7315 When inserting a __builtin_assume_aligned call set the LHS
7316 SSA name alignment info accordingly.
7318 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7320 PR tree-optimization/84114
7321 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7322 Avoid reassociation of FLOAT_MODE addition.
7324 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7326 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7327 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7328 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7329 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7330 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7331 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7332 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7334 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7335 __builtin_ia32_wbinvd): New builtins.
7336 (SPECIAL_ARGS2): New.
7337 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7338 (SPECIAL_ARGS2): New.
7339 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7340 (ix86_valid_target_attribute_inner_p): Ditto.
7341 (ix86_init_mmx_sse_builtins): Add special_args2.
7342 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7343 TARGET_WBNOINVD_P): New.
7344 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7345 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7346 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7347 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7348 * config/i386/pconfigintrin.h: New file.
7349 * config/i386/wbnoinvdintrin.h: Ditto.
7350 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7352 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7354 2018-03-05 Richard Biener <rguenther@suse.de>
7356 PR tree-optimization/84670
7357 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7359 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7360 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7361 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7362 doesn't grow if all successors (recursively) were visited at least
7365 2018-03-05 Richard Biener <rguenther@suse.de>
7367 PR tree-optimization/84650
7368 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7369 if executed in the loop pipeline.
7371 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7373 * doc/configfiles.texi (Configuration Files): Move info about
7374 conditionalizing $target-protos.h to...
7375 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7376 differs from $target-protos.h.
7378 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7379 Chung-Ju Wu <jasonwucj@gmail.com>
7381 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7382 * config/nds32/nds32-multiple.md (setmemsi): Define.
7383 * config/nds32/nds32-memory-manipulation.c
7384 (nds32_gen_dup_4_byte_to_word_value): New.
7385 (emit_setmem_word_loop): New.
7386 (emit_setmem_byte_loop): New.
7387 (nds32_expand_setmem_loop): New.
7388 (nds32_expand_setmem_loop_v3m): New.
7389 (nds32_expand_setmem_unroll): New.
7390 (nds32_expand_setmem): New.
7392 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7393 Chung-Ju Wu <jasonwucj@gmail.com>
7395 * config/nds32/nds32-memory-manipulation.c
7396 (nds32_emit_load_store): New.
7397 (nds32_emit_post_inc_load_store): New.
7398 (nds32_emit_mem_move): New.
7399 (nds32_emit_mem_move_block): New.
7400 (nds32_expand_movmemsi_loop_unknown_size): New.
7401 (nds32_expand_movmemsi_loop_known_size): New.
7402 (nds32_expand_movmemsi_loop): New.
7403 (nds32_expand_movmemsi_unroll): New.
7404 (nds32_expand_movmemqi): Rename ...
7405 (nds32_expand_movmemsi): ... to this.
7406 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7407 (movmemsi): ... to this.
7408 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7409 (nds32_expand_movmemsi): ... to this.
7411 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7412 Monk Chiang <sh.chiang04@gmail.com>
7413 Chung-Ju Wu <jasonwucj@gmail.com>
7415 * config/nds32/nds32-protos.h
7416 (nds32_expand_load_multiple): New arguments.
7417 (nds32_expand_store_multiple): Ditto.
7418 (nds32_valid_multiple_load_store): Rename ...
7419 (nds32_valid_multiple_load_store_p): ... to this.
7420 * config/nds32/nds32-memory-manipulation.c
7421 (nds32_expand_load_multiple): Refine implementation.
7422 (nds32_expand_store_multiple): Ditto.
7423 * config/nds32/nds32-multiple.md
7424 (load_multiple): Update nds32_expand_load_multiple interface.
7425 (store_multiple): Update nds32_expand_store_multiple interface.
7426 * config/nds32/nds32-predicates.c
7427 (nds32_valid_multiple_load_store): Rename ...
7428 (nds32_valid_multiple_load_store_p): ... to this and refine
7430 * config/nds32/predicates.md
7431 (nds32_load_multiple_and_update_address_operation): New predicate.
7432 (nds32_store_multiple_and_update_address_operation): New predicate.
7434 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7435 Chung-Ju Wu <jasonwucj@gmail.com>
7437 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7438 (combo): New attribute.
7439 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7441 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7443 * config/nds32/nds32.opt: Change -mcmodel= default value.
7445 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7446 Monk Chiang <sh.chiang04@gmail.com>
7447 Chung-Ju Wu <jasonwucj@gmail.com>
7449 * config/nds32/constants.md (unspec_element): New enum.
7450 * config/nds32/constraints.md (Umw): New constraint.
7451 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7452 * config/nds32/nds32-intrinsic.md: Likewise.
7453 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7454 (nds32_valid_smw_lwm_base_p): New.
7455 (nds32_output_smw_single_word): New.
7456 (nds32_output_lmw_single_word): New.
7457 (nds32_expand_unaligned_load): New.
7458 (nds32_expand_unaligned_store): New.
7459 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7460 (nds32_output_smw_single_word): Declare.
7461 (nds32_output_lmw_single_word): Declare.
7462 (nds32_expand_unaligned_load): Declare.
7463 (nds32_expand_unaligned_store): Declare.
7464 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7465 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7466 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7467 NDS32_BUILTIN_UASTORE_DW.
7468 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7471 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7472 Kito Cheng <kito.cheng@gmail.com>
7473 Chung-Ju Wu <jasonwucj@gmail.com>
7475 * config/nds32/nds32-intrinsic.c
7476 (nds32_expand_builtin_null_ftype_reg): Delete.
7477 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7478 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7479 (nds32_read_argument): New.
7480 (nds32_legitimize_target): Ditto.
7481 (nds32_legitimize_argument): Ditto.
7482 (nds32_check_constant_argument): Ditto.
7483 (nds32_expand_unop_builtin): Ditto.
7484 (nds32_expand_unopimm_builtin): Ditto.
7485 (nds32_expand_binop_builtin): Ditto.
7486 (nds32_builtin_decl_impl): Ditto.
7487 (builtin_description): Ditto.
7488 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7489 (nds32_init_builtins_impl): Ditto.
7490 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7491 (nds32_builtin_decl): New.
7492 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7493 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7495 2018-03-02 Jeff Law <law@redhat.com>
7497 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7498 (redundant_insn, fill_simple_delay_slots): Likewise.
7499 (fill_slots_from_thread): Likewise.
7500 * resource.c (mark_referenced_resources): Likewise.
7501 (mark_set_resources, find_dead_or_set_registers): Likewise.
7503 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7505 * substring-locations.h (format_warning_va): Formatting fix for
7507 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7509 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7510 * substring-locations.c: Include intl.h.
7511 (format_warning_va): Turned into small wrapper around
7512 format_warning_n_va, renamed to ...
7513 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7514 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7516 (format_warning_at_substring_n): New function.
7517 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7518 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7519 format_warning_at_substring with just a shorter name instead of
7520 const function pointer.
7521 (fmtwarn_n): New function.
7522 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7523 appropriate, get rid of all the fmtstr temporaries, move conditionals
7524 with G_() wrapped string literals directly into fmtwarn arguments,
7525 cast dir.len to (int), formatting fixes.
7527 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7529 * doc/invoke.texi: Remove "Cilk Plus" references.
7531 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7532 Richard Biener <rguenther@suse.de>
7535 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7536 for error or warning attributes if CALL_FROM_THUNK_P is set.
7539 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7542 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7543 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7546 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7547 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7549 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7550 instead of -1U in last predictors element's probability member.
7552 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7555 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7556 arguments if they are comparable.
7558 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7560 PR tree-optimization/84634
7561 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7562 masks and masked_loop_p with a single loop_masks, making sure it's
7563 null for bb vectorization.
7565 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7567 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7568 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7569 loop->force_vectorize to check whether there is no alias.
7571 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7574 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7576 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7578 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7579 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7580 instead of a loop around prev_real_insn.
7581 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7585 * config/i386/i386.c (ix86_print_operand): Use conditional
7586 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7589 2018-03-02 Richard Biener <rguenther@suse.de>
7591 PR tree-optimization/84427
7592 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7593 (bitmap_set_subtract_values): Rewrite to handle multiple
7596 (prune_clobbered_mems): Likewise.
7597 (phi_translate): Take edge instead of pred/phiblock.
7598 (phi_translate_1): Likewise.
7599 (phi_translate_set): Likewise. Insert all translated
7600 exprs for a value into the set, keeping possibly multiple
7601 expressions per value.
7602 (compute_antic_aux): Adjust for phi_translate changes.
7603 When intersecting union the expressions and prune those
7604 not in the final value set, keeping possibly multiple
7605 expressions per value. Do not use value-insertion
7606 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7607 all expressions. Add verification that the value-sets
7608 only shrink during iteration.
7609 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7610 (do_pre_regular_insertion): Likewise.
7611 (do_pre_partial_partial_insertion): Likewise.
7613 2018-03-02 Richard Biener <rguenther@suse.de>
7616 * config/darwin.c (saved_debug_info_level): New static global.
7617 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7618 (darwin_asm_lto_end): Restore debug info generation settings.
7620 2018-03-01 Martin Liska <mliska@suse.cz>
7623 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7626 2018-03-01 Richard Biener <rguenther@suse.de>
7629 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7630 type annotation with LTO when debug was disabled at compile-time.
7632 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7634 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7636 (mips_final_postscan_insn): Likewise.
7638 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7640 PR rtl-optimization/84528
7641 * alias.c (init_alias_target): Add commentary.
7642 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7643 a unique base value if the frame pointer is not eliminated
7644 to the stack pointer.
7646 2018-03-01 Tom de Vries <tom@codesourcery.com>
7648 PR rtl-optimization/83327
7649 * lra-int.h (hard_regs_spilled_into): Declare.
7650 * lra.c (hard_regs_spilled_into): Define.
7651 (init_reg_info): Init hard_regs_spilled_into.
7652 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7653 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7654 (process_bb_lives): Handle hard_regs_spilled_into.
7655 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7656 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7658 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7660 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7661 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7662 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7663 * config/rs6000/aix72.h: New file.
7665 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7667 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7668 instead of warning_at with conditional singular and plural messages
7672 * stor-layout.c (update_alignment_for_field): For
7673 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7674 && !DECL_PACKED (field), do the alignment update, just use
7675 only desired_align instead of MAX (type_align, desired_align)
7677 (place_field): Don't do known_align < desired_align handling
7678 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7679 is non-NULL, instead do it after rli->prev_field handling and
7680 only if not within a bitfield word. For DECL_PACKED (field)
7681 use type_align of BITS_PER_UNIT.
7683 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7685 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7686 superfluous parentheses and trailing spaces.
7688 2018-02-28 Richard Biener <rguenther@suse.de>
7690 PR tree-optimization/84584
7691 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7692 SCoPs with fake exit edge.
7694 2018-02-28 Martin Liska <mliska@suse.cz>
7697 * timevar.c (timer::print): Fix format to properly print 100%
7700 2018-02-28 Richard Biener <rguenther@suse.de>
7703 * genmatch.c (capture_info::walk_match): Do not mark
7704 captured expressions without operands as expr_p given
7705 they act more like predicates and should be subject to
7706 "lost tail" side-effect preserving.
7708 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7710 PR rtl-optimization/81611
7711 * auto-inc-dec.c (attempt_change): Move dead note from
7712 mem_insn if it's the next use of regno
7713 (find_address): Take address use of reg holding
7714 non-incremented value. Add parm to limit search to the named
7716 (merge_in_block): Attempt to use a mem insn that is the next
7717 use of the original regno.
7719 2018-02-27 Martin Sebor <msebor@redhat.com>
7722 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7723 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7725 2018-02-27 Martin Sebor <msebor@redhat.com>
7727 PR translation/84207
7728 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7729 n argument to unsigned HOST_WIDE_INT.
7730 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7731 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7732 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7733 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7735 2018-02-27 Richard Biener <rguenther@suse.de>
7737 PR tree-optimization/84512
7738 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7739 Do not use the estimate returned from record_stmt_cost for
7740 the scalar iteration cost but sum properly using add_stmt_cost.
7742 2018-02-27 Richard Biener <rguenther@suse.de>
7744 PR tree-optimization/84466
7745 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7746 Adjust last change to less strictly validate use operands.
7748 2018-02-27 Martin Liska <mliska@suse.cz>
7750 PR gcov-profile/84548
7751 * gcov.c (process_file): Allow partial overlap and consider it
7752 also as group functions.
7753 (output_lines): Properly calculate range of lines for a group.
7755 2018-02-27 Martin Liska <mliska@suse.cz>
7757 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7758 'ggc' suffixes. Change first column width.
7759 (timer::print): Fix formatting of the column.
7761 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7763 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7764 preserve inline entry blocks for the sake of debug inline
7765 entry point markers alone.
7766 (remove_unused_locals): Suggest in comments a better place to
7767 force the preservation of inline entry blocks that are
7768 otherwise unused, but do not preserve them.
7770 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7772 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7774 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7777 * config/i386/constraints.md (Bs): Replace
7778 ix86_indirect_branch_register with
7779 TARGET_INDIRECT_BRANCH_REGISTER.
7781 * config/i386/i386.md (indirect_jump): Likewise.
7782 (tablejump): Likewise.
7783 (*sibcall_memory): Likewise.
7784 (*sibcall_value_memory): Likewise.
7785 Peepholes of indirect call and jump via memory: Likewise.
7786 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7787 (*sibcall_value_GOT_32): Likewise.
7788 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7789 (GOT_memory_operand): Likewise.
7790 (call_insn_operand): Likewise.
7791 (sibcall_insn_operand): Likewise.
7792 (GOT32_symbol_operand): Likewise.
7793 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7795 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7797 PR rtl-optimization/83496
7798 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7799 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7800 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7801 redundant insn, if any.
7802 (relax_delay_slots): Likewise.
7803 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7805 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7807 PR tree-optimization/83965
7808 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7809 that grouped statements are part of a reduction chain. Return
7810 true if the statement is not marked as a reduction itself but
7812 (vect_recog_dot_prod_pattern): Don't check whether the statement
7813 is part of a group here.
7814 (vect_recog_sad_pattern): Likewise.
7815 (vect_recog_widen_sum_pattern): Likewise.
7817 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7820 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7822 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7825 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7827 (ix86_output_indirect_function_return): New prototype.
7828 (ix86_split_simple_return_pop_internal): Likewise.
7829 * config/i386/i386.c (indirect_return_via_cx): New.
7830 (indirect_return_via_cx_bnd): Likewise.
7831 (indirect_thunk_name): Handle return va CX_REG.
7832 (output_indirect_thunk_function): Create alias for
7833 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7834 (ix86_output_indirect_jmp): Remove the bool argument.
7835 (ix86_output_indirect_function_return): New function.
7836 (ix86_split_simple_return_pop_internal): Likewise.
7837 * config/i386/i386.md (*indirect_jump): Don't pass false
7838 to ix86_output_indirect_jmp.
7839 (*tablejump_1): Likewise.
7840 (simple_return_pop_internal): Change it to define_insn_and_split.
7841 Call ix86_split_simple_return_pop_internal to split it for
7843 (simple_return_indirect_internal): Call
7844 ix86_output_indirect_function_return instead of
7845 ix86_output_indirect_jmp.
7847 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7850 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7851 memset and value initialization afterwards.
7853 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7855 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7857 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7860 * common/config/aarch64/aarch64-common.c
7861 (aarch_option_optimization_table[]): Switch
7862 off fomit-frame-pointer
7864 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7865 Chung-Ju Wu <jasonwucj@gmail.com>
7867 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7869 (store_multiple): Ditto.
7871 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7873 * config.gcc: Add --with-cpu support for nds32 target.
7874 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7875 * config/nds32/nds32.opt: Add -mcpu= option.
7877 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7879 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7880 isel=yes): Warn for these deprecated options.
7882 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7884 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7885 ISA_2_5_MASKS_EMBEDDED.
7887 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7889 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7890 p->max as pointers rather than using iterative_hash_expr.
7892 2018-02-23 Carl Love <cel@us.ibm.com>
7894 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7895 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7897 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7898 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7899 P8V_BUILTIN_VEC_VUNSIGNED2.
7901 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7904 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7905 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7907 * lra-constraints.c (curr_insn_transform): Set up
7908 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7910 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7911 * lra-eliminations.c (spill_pseudos): Ditto.
7912 (process_insn_for_elimination): Ditto.
7913 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7914 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7915 LRA_NON_CLOBBERED_ALT.
7917 2018-02-22 Martin Sebor <msebor@redhat.com>
7919 PR tree-optimization/84480
7920 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7921 to maybe_diag_stxncpy_trunc. Call it.
7922 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7923 from gimple_fold_builtin_strcpy. Print inlining stack.
7924 (handle_builtin_stxncpy): Print inlining stack.
7925 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7927 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7930 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7931 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7932 and -fcheck-pointer-bounds are used together.
7933 (indirect_thunk_prefix): New enum.
7934 (indirect_thunk_need_prefix): New function.
7935 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7936 "_nt" instead of "_bnd" for NOTRACK prefix.
7937 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7938 (output_indirect_thunk_function): Likewise.
7940 (ix86_code_end): Update output_indirect_thunk_function calls.
7941 (ix86_output_indirect_branch_via_reg): Replace
7942 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7943 (ix86_output_indirect_branch_via_push): Likewise.
7944 (ix86_output_function_return): Likewise.
7945 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7946 incompatible with -fcf-protection=branch and
7947 -fcheck-pointer-bounds.
7949 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7952 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7953 Change gcc_assert call to output_operand_lossage.
7955 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7957 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7959 2018-02-22 DJ Delorie <dj@redhat.com>
7960 Sebastian Perta <sebastian.perta@renesas.com>
7961 Oleg Endo <olegendo@gcc.gnu.org>
7963 * config/rx/rx.c (rx_rtx_costs): New function.
7964 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7966 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7968 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7970 2018-02-22 Martin Liska <mliska@suse.cz>
7973 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7974 Add "native" as a possible value.
7976 2018-02-22 Martin Liska <mliska@suse.cz>
7979 * config/i386/i386.c (ix86_option_override_internal):
7980 Add "native" as a possible value for -march and -mtune.
7982 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7985 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7986 to all type variants.
7988 PR tree-optimization/84503
7989 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7990 width as info->bitpos + info->bitsize - start.
7991 (merged_store_group::merge_overlapping): Simplify width computation.
7992 (check_no_overlap): New function.
7993 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7994 start + width and last_order of the group, fail if check_no_overlap
7996 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7997 to group if check_no_overlap fails.
7999 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8001 * config/rs6000/altivec.md: Delete contraint arguments to
8002 define_expand, define_split, and define_peephole2, and in
8003 define_insn_and_split if always unused.
8004 * config/rs6000/darwin.md: Ditto.
8005 * config/rs6000/dfp.md: Ditto.
8006 * config/rs6000/rs6000.md: Ditto.
8007 * config/rs6000/sync.md: Ditto.
8008 * config/rs6000/vector.md: Ditto.
8009 * config/rs6000/vsx.md: Ditto.
8011 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
8013 * config/rs6000/altivec.md: Write output control strings as braced
8014 blocks instead of double-quoted strings.
8015 * config/rs6000/darwin.md: Ditto.
8016 * config/rs6000/rs6000.md: Ditto.
8017 * config/rs6000/vector.md: Ditto.
8018 * config/rs6000/vsx.md: Ditto.
8020 2018-02-21 Jason Merrill <jason@redhat.com>
8022 PR c++/84314 - ICE with templates and fastcall attribute.
8023 * attribs.c (build_type_attribute_qual_variant): Remove assert.
8025 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8027 * ipa-cp.c (determine_versionability): Fix comment typos.
8029 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
8032 * ipa-cp.c (determine_versionability): Do not version functions caling
8035 2018-02-21 Martin Liska <mliska@suse.cz>
8038 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
8039 Add "native" as a possible value.
8040 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
8041 the macro when native cpu detection is available.
8043 2018-02-21 Martin Liska <mliska@suse.cz>
8046 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
8047 Add "native" as a possible value.
8048 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
8049 when native cpu detection is available.
8051 2018-02-21 Jakub Jelinek <jakub@redhat.com>
8052 Martin Sebor <msebor@redhat.com>
8054 PR tree-optimization/84478
8055 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
8057 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
8058 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
8059 support which is conservatively correct, for 2 only stay conservative
8060 for maxlen. Formatting and comment capitalization fixes. Add STRICT
8061 argument to the 2 argument get_range_strlen, adjust 6 arg
8062 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
8064 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
8065 (gimple_fold_builtin_strlen): Pass true as last argument to
8068 2018-02-20 Martin Sebor <msebor@redhat.com>
8071 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
8072 (builtin_memref::set_base_and_offset): Same. Handle inner references.
8073 (builtin_memref::builtin_memref): Factor out parts into
8074 set_base_and_offset and call it.
8076 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8079 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
8080 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
8081 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
8082 search at the associated MODE_INT.
8084 2018-02-20 Jeff Law <law@redhat.com>
8087 PR tree-optimization/81592
8089 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8090 for range data rather than using global data.
8091 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8092 range data rather than using global data.
8093 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8094 pass it to children as needed.
8095 (struct directive::fmtresult): Similarly.
8096 (struct directive::set_width): Similarly.
8097 (struct directive::set_precision): Similarly.
8098 (format_integer, format_directive, parse_directive): Similarly.
8099 (format_none): Accept unnamed vr_values parameter.
8100 (format_percent, format_floating, format_character): Similarly.
8101 (format_string, format_plain): Similarly.
8102 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8103 the EVRP range analyzer for range data rather than using global data.
8104 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8105 gimple-ssa-evrp-analyze.h
8106 (class sprintf_dom_walker): Add after_dom_children member function.
8107 Add evrp_range_analyzer member.
8108 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8109 range analyzer as needed.
8110 (sprintf_dom_walker::after_dom_children): New member function.
8111 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8113 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8114 (evrp_range_analyzer::pop_to_marker): Likewise.
8116 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8118 PR tree-optimization/84419
8119 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8120 with the required type if its current type is compatible but
8123 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8126 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8127 after vectorization.
8129 2018-02-20 Martin Liska <mliska@suse.cz>
8132 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8133 possible values if we don't have a hint.
8135 2018-02-20 Martin Liska <mliska@suse.cz>
8139 * final.c (shorten_branches): Build align_tab array with one
8141 * opts.c (finish_options): Add alignment option limit check.
8142 (MAX_CODE_ALIGN): Likewise.
8143 (MAX_CODE_ALIGN_VALUE): Likewise.
8144 * doc/invoke.texi: Document maximum allowed option value for
8145 all -falign-* options.
8147 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8150 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8151 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8152 * var-tracking.c (emit_note_insn_var_location): Remove all references
8153 to NOTE_INSN_CALL_ARG_LOCATION.
8154 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8155 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8156 Use copy_rtx_if_shared.
8157 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8158 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8159 (dwarf2out_var_location): Remove handling of
8160 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8162 * final.c (final_scan_insn): Remove all references to
8163 NOTE_INSN_CALL_ARG_LOCATION.
8164 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8165 before dumping final insns.
8166 * except.c (emit_note_eh_region_end): Remove all references to
8167 NOTE_INSN_CALL_ARG_LOCATION.
8168 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8169 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8170 * config/arc/arc.c (hwloop_optimize): Likewise.
8171 * config/arm/arm.c (create_fix_barrier): Likewise.
8172 * config/s390/s390.c (s390_chunkify_start): Likewise.
8173 * config/sh/sh.c (find_barrier): Likewise.
8174 * config/i386/i386.c (rest_of_insert_endbranch,
8175 ix86_seh_fixup_eh_fallthru): Likewise.
8176 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8177 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8178 * config/frv/frv.c (frv_function_prologue): Likewise.
8179 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8181 (note_outside_basic_block_p): Remove all references to
8182 NOTE_INSN_CALL_ARG_LOCATION.
8183 * gengtype.c (adjust_field_rtx_def): Likewise.
8184 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8186 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8187 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8190 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8193 PR tree-optimization/84452
8194 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8195 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8198 2018-02-19 Martin Liska <mliska@suse.cz>
8201 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8203 2018-02-19 Martin Liska <mliska@suse.cz>
8204 Richard Sandiford <richard.sandiford@linaro.org>
8206 PR tree-optimization/82491
8207 * gimple-fold.c (get_base_constructor): Make earlier bail out
8210 2018-02-19 Carl Love <cel@us.ibm.com>
8212 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8213 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8215 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8216 P8V_BUILTIN_VEC_NEG.
8218 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8220 * config/rl78/rl78.md (movdf): New define expand.
8222 2018-02-19 Martin Liska <mliska@suse.cz>
8225 * doc/invoke.texi: Fix typo.
8226 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8228 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8230 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8231 handle rs6000_single_float and rs6000_double_float specially for
8234 2018-02-16 Jeff Law <law@redhat.com>
8236 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8239 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8242 * ipa-inline.c (inline_small_functions): Fix a typo.
8244 2018-02-16 Nathan Sidwell <nathan@acm.org>
8246 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8248 2018-02-16 Carl Love <cel@us.ibm.com>
8250 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8251 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8252 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8253 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8254 expansion to P8V_BUILTIN_VEC_FLOAT2.
8256 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8258 PR rtl-optimization/70023
8259 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8260 src_regno into account.
8262 2018-02-16 Carl Love <cel@us.ibm.com>
8264 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8265 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8266 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8267 * config/rs6000/rs6000.c: Remove case statements for
8268 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8269 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8270 and P9V_BUILTIN_VEC_VINSERT4B.
8271 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8272 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8273 * config/rs6000/vsx.md:
8274 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8277 2018-02-16 Carl Love <cel@us.ibm.com>
8279 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8281 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8283 * config/rs6000/rs6000-c.c: Add the definitions for
8284 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8285 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8286 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8287 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8288 definition for insert4b and define insn *insert3b_internal.
8289 * doc/extend.texi: Add documentation for vec_extract4b.
8291 2018-02-16 Nathan Sidwell <nathan@acm.org>
8293 * doc/extend.texi (Backwards Compatibility): Mention friend
8294 injection. Note for-scope is deprecated.
8295 * doc/invoke.texi (-ffriend-injection): Deprecate.
8297 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8299 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8300 that moved to I2, also allow destinations that are a paradoxical
8301 subreg (instead of a normal reg).
8303 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8306 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8309 2018-02-16 Richard Biener <rguenther@suse.de>
8311 PR tree-optimization/84037
8312 PR tree-optimization/84016
8314 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8315 Adjust vec_construct for the fact we need additional higher latency
8316 128bit inserts for AVX256 and AVX512 vector builds.
8317 (ix86_add_stmt_cost): Scale vector construction cost for
8320 2018-02-16 Richard Biener <rguenther@suse.de>
8322 PR tree-optimization/84417
8323 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8324 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8325 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8327 2018-02-16 Martin Liska <mliska@suse.cz>
8330 * internal-fn.def (ASAN_CHECK): Set proper flags.
8331 (ASAN_MARK): Likewise.
8333 2018-02-16 Julia Koval <julia.koval@intel.com>
8335 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8336 from PTA_CANNONLAKE.
8338 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8341 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8342 Use ++iter rather than iter++ for std::list iterators.
8343 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8344 defer deleting them until all nodes in the forest are processed. Do
8345 free even leaf nodes. Change to_process into auto_vec.
8348 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8349 * vec.h (vec_default_construct): Use memset instead of placement new
8350 if BROKEN_VALUE_INITIALIZATION is defined.
8351 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8352 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8355 PR rtl-optimization/83723
8356 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8357 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8358 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8360 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8362 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8364 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8366 PR rtl-optimization/81443
8367 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8368 from inner REGs to paradoxical SUBREGs.
8370 2018-02-16 Richard Biener <rguenther@suse.de>
8372 PR tree-optimization/84399
8373 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8374 For operands we can analyze at their definition make sure we can
8375 analyze them at each use as well.
8377 2018-02-16 Richard Biener <rguenther@suse.de>
8379 PR tree-optimization/84190
8380 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8381 volatile accesses if the decl isn't volatile.
8383 2018-02-15 Jason Merrill <jason@redhat.com>
8385 PR c++/84314 - ICE with templates and fastcall attribute.
8386 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8387 TYPE_CANONICAL on an existing type.
8389 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8391 PR tree-optimization/84383
8392 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8393 dstoff nor call operand_equal_p if dstbase is NULL.
8395 PR tree-optimization/84334
8396 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8397 also a CONSTANT_CLASS_P, punt.
8399 2018-02-14 Jim Wilson <jimw@sifive.com>
8401 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8402 first SMALL_OPERAND check. New local min_second_step. Move assert
8403 to where locals are set. Add TARGET_RVC support.
8404 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8406 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8408 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8410 2018-02-14 Martin Sebor <msebor@redhat.com>
8412 PR tree-optimization/83698
8413 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8414 arrays constrain the offset range to their bounds.
8415 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8416 (builtin_access::overlap): Avoid setting the size of overlap if it's
8418 (maybe_diag_overlap): Also consider arrays when deciding what values
8419 of offsets to include in diagnostics.
8421 2018-02-14 Martin Sebor <msebor@redhat.com>
8424 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8425 that correspond to the kind of a declaration.
8427 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8430 * config/pa/pa.md: Load address of PIC label using the linkage table
8431 if the label is nonlocal.
8433 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8435 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8436 warning message if user requests -maltivec=be.
8437 * doc/invoke.texi: Document deprecation of -maltivec=be.
8439 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8442 * config/rs6000/rs6000-c.c: Update definitions for
8443 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8444 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8446 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8449 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8450 add _get_ssp intrinsics. Remove argument from
8451 __builtin_ia32_rdssp[d|q].
8452 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8453 * config/i386/i386-builtin.def: Remove argument from
8454 __builtin_ia32_rdssp[d|q].
8455 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8456 ix86_expand_special_args_builtin for _rdssp[d|q].
8457 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8458 Clear register before usage.
8459 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8460 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8462 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8464 PR tree-optimization/84357
8465 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8466 operand 1 of an ARRAY_REF too.
8468 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8471 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8472 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8474 (set_of_reg): New struct.
8475 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8476 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8477 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8479 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8480 Split into bitclr, bitset, bitinvert patterns if appropriate.
8481 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8482 use rx_fuse_in_memory_bitop.
8483 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8484 to named insn, correct maximum insn length.
8486 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8489 * machmode.def: Define a complex mode for PARTIAL_INT.
8490 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8492 * doc/rtl.texi: Document CSPImode.
8493 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8495 (msp430_hard_regno_nregs_with_padding): Likewise.
8497 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8500 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8502 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8504 PR rtl-optimization/84169
8505 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8506 we generated a parallel as new i3 and we split that to new i2 and i3
8507 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8508 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8509 those to i2, not i1. Partially rewrite this scan code.
8511 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8514 * stor-layout.c (place_field): For variable length fields, adjust
8515 offset_align afterwards not just based on the field's alignment,
8516 but also on the size.
8519 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8520 of exps and logs in the use_exp2 case.
8522 2018-02-13 Jeff Law <law@redhat.com>
8524 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8527 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8530 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8534 * common.opt (gas-loc-support, gas-locview-support): New.
8535 (ginline-points, ginternal-reset-location-views): New.
8536 * doc/invoke.texi: Document them. Use @itemx where intended.
8537 (gvariable-location-views): Adjust.
8538 * target.def (reset_location_view): New.
8539 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8540 (TARGET_RESET_LOCATION_VIEW): New.
8541 * doc/tm.texi: Rebuilt.
8542 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8543 (dwarf2out_default_as_locview_support): New.
8544 (output_asm_line_debug_info): Use option variables.
8545 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8546 (output_loc_list): Likewise.
8547 (add_high_low_attributes): Check option variables.
8548 Don't output entry view attribute in strict mode.
8549 (gen_inlined_subroutine_die): Check option variables.
8550 (dwarf2out_inline_entry): Likewise.
8551 (init_sections_and_labels): Likewise.
8552 (dwarf2out_early_finish): Likewise.
8553 (maybe_reset_location_view): New, from...
8554 (dwarf2out_var_location): ... here. Call it.
8555 * debug.h (dwarf2out_default_as_loc_support): Declare.
8556 (dwarf2out_default_as_locview_support): Declare.
8557 * hooks.c (hook_int_rtx_insn_0): New.
8558 * hooks.h (hook_int_rtx_insn_0): Declare.
8559 * toplev.c (process_options): Take -gas-loc-support and
8560 -gas-locview-support from dwarf2out. Enable
8561 -gvariable-location-views by default only with locview
8562 assembler support. Enable -ginternal-reset-location-views by
8563 default only if the target defines the corresponding hook.
8564 Enable -ginline-points by default if location views are
8565 enabled; force it disabled if statement frontiers are
8567 * tree-inline.c (expand_call_inline): Check option variables.
8568 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8570 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8572 PR tree-optimization/84321
8573 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8574 handling. Also check whether the anti-range contains any values
8575 that satisfy the mask; switch to a VR_RANGE if not.
8577 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8580 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8582 2018-02-13 Martin Jambor <mjambor@suse.cz>
8585 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8586 of call statements, also set location of a load to a temporary.
8588 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8590 * config/rl78/rl78.c (add_vector_labels): New function.
8591 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8592 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8593 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8594 which checks that no arguments are passed.
8595 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8596 * doc/extend.texi: Documentation for the new attribute.
8598 2018-02-13 Andreas Schwab <schwab@suse.de>
8600 * config/riscv/linux.h (CPP_SPEC): Define.
8602 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8605 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8606 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8607 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8608 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8609 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8610 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8611 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8612 temporarily for AES and PCLMUL builtins.
8614 PR tree-optimization/84339
8615 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8616 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8620 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8621 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8622 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8624 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8626 * omp-simd-clone.h: New file.
8627 * omp-simd-clone.c: Include omp-simd-clone.h.
8628 (expand_simd_clones): No longer static.
8629 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8630 cgraph.h and omp-simd-clone.h.
8631 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8632 (vect_recog_widen_shift_pattern): Formatting fix.
8633 (vect_pattern_recog_1): Don't check optab for calls.
8636 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8637 operands[2] into a REG before using gen_lowpart on it.
8639 2018-02-12 Jeff Law <law@redhat.com>
8642 * config/sh/sh.c (find_barrier): Consider a sibling call
8645 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8646 successfully back substituting a reg.
8648 2018-02-12 Richard Biener <rguenther@suse.de>
8650 PR tree-optimization/84037
8651 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8652 parameter, move visited init to caller.
8653 (vect_slp_analyze_operations): Separate cost from validity
8654 check, initialize visited once for all instances.
8655 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8657 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8659 (vect_model_promotion_demotion_cost): Likewise.
8660 (vectorizable_bswap): Guard cost modeling with !slp_node
8661 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8663 (vectorizable_call): Likewise.
8664 (vectorizable_conversion): Likewise.
8665 (vectorizable_assignment): Likewise.
8666 (vectorizable_shift): Likewise.
8667 (vectorizable_operation): Likewise.
8668 (vectorizable_store): Likewise.
8669 (vectorizable_load): Likewise.
8670 (vectorizable_condition): Likewise.
8671 (vectorizable_comparison): Likewise.
8673 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8676 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8677 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8678 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8680 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8683 * params.def (inline-min-speedup): Increase from 8 to 15.
8684 (max-inline-insns-auto): Decrease from 40 to 30.
8685 * ipa-split.c (consider_split): Add some buffer for function to
8686 be considered inlining candidate.
8687 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8690 2018-02-12 Richard Biener <rguenther@suse.de>
8692 PR tree-optimization/84037
8693 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8694 matched stmts if we cannot swap the non-matched ones.
8696 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8698 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8699 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8700 _mm_maskz_scalef_round_ss): New intrinsics.
8701 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8702 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8703 __builtin_ia32_scalefss_round): Remove.
8704 (__builtin_ia32_scalefsd_mask_round,
8705 __builtin_ia32_scalefss_mask_round): New intrinsics.
8706 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8707 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8708 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8709 "<round_constraint>")): Changed to ...
8710 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8711 "<round_scalar_constraint>")): ... this.
8712 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8713 %0, %1, %2<round_op3>}"): Changed to ...
8714 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8715 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8716 %2<round_scalar_mask_op3>}"): ... this.
8717 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8719 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8721 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8722 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8723 (_mm_maskz_sqrt_round_ss): New intrinsics.
8724 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8725 (__builtin_ia32_sqrtsd_mask_round)
8726 (__builtin_ia32_sqrtss_mask_round): New builtins.
8727 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8728 (__builtin_ia32_sqrtss_round): Remove.
8729 (__builtin_ia32_sqrtsd_mask_round)
8730 (__builtin_ia32_sqrtss_mask_round): New builtins.
8731 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8732 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8733 ((match_operand:VF_128 1 "vector_operand"
8734 "xBm,<round_constraint>")): Changed to ...
8735 ((match_operand:VF_128 1 "vector_operand"
8736 "xBm,<round_scalar_constraint>")): ... this.
8737 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8738 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8739 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8740 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8741 %<iptr>1<round_scalar_mask_op3>}): ... this.
8742 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8743 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8745 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8748 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8749 Cast vec_cmpeq result to correct type.
8750 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8751 Cast vec_cmpgt result to correct type.
8753 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8755 * final.c (final_scan_insn_1): Renamed from...
8756 (final_scan_insn): ... this. New wrapper, to recover
8757 seen from the outermost call in recursive ones.
8758 * config/sparc/sparc.c (output_return): Drop seen from call.
8759 (output_sibcall): Likewise.
8760 * config/visium/visium.c (output_branch): Likewise.
8762 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8764 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8767 2018-02-10 Alan Modra <amodra@gmail.com>
8770 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8771 Specify LR as an input.
8773 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8776 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8777 remove_member_access_dummy_vars): New functions.
8778 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8779 lower_omp_1, execute_lower_omp): Use them.
8781 PR rtl-optimization/84308
8782 * shrink-wrap.c (spread_components): Release todo vector.
8784 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8786 PR rtl-optimization/57193
8787 * ira-color.c (struct allocno_color_data): Add member
8788 conflict_allocno_hard_prefs.
8789 (update_conflict_allocno_hard_prefs): New.
8790 (bucket_allocno_compare_func): Add a preference based on
8791 conflict_allocno_hard_prefs.
8792 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8793 (color_allocnos): Remove a dead code. Initiate
8794 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8796 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8799 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8800 constraint from =wa to wa. Avoid a subreg on the output operand,
8801 instead use a pseudo and subreg it in a move.
8802 (p9_xxbrd_<mode>): Changed to ...
8803 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8804 (p9_xxbrd_v2df): New expander.
8805 (p9_xxbrw_<mode>): Changed to ...
8806 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8807 (p9_xxbrw_v4sf): New expander.
8809 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8811 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8813 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8816 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8817 multiply in 32-bit mode.
8818 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8819 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8822 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8824 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8825 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8826 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8827 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8829 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8832 * dwarf2out.c (is_trivial_indirect_ref): New function.
8833 (dwarf2out_late_global_decl): Do not generate a location
8834 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8835 and that are not defined in the current unit.
8837 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8839 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8840 instead of a libcall for UNORDERED.
8842 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8845 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8846 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8848 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8851 * config/s390/s390.c (s390_set_current_function): Invoke
8852 s390_indirect_branch_settings also if fndecl didn't change.
8854 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8856 * config/rs6000/rs6000.md (blockage): Set length to zero.
8858 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8860 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8862 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8865 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8866 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8870 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8871 PARALLEL incoming that failed vt_get_decl_and_offset check.
8874 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8875 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8877 (get_variable_section): For decls in named .bss* sections pass true as
8878 second argument to bss_initializer_p.
8880 2018-02-09 Marek Polacek <polacek@redhat.com>
8881 Jakub Jelinek <jakub@redhat.com>
8884 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8885 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8886 Sync some changes from cxx_fold_indirect_ref.
8888 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8890 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8892 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8893 (BLOCK_INLINE_ENTRY_LABEL): New.
8894 (dwarf2out_var_location): Disregard inline entry markers.
8895 (inline_entry_data): New struct.
8896 (inline_entry_data_hasher): New hashtable type.
8897 (inline_entry_data_hasher::hash): New.
8898 (inline_entry_data_hasher::equal): New.
8899 (inline_entry_data_table): New variable.
8900 (add_high_low_attributes): Add DW_AT_entry_pc and
8901 DW_AT_GNU_entry_view attributes if a pending entry is found
8902 in inline_entry_data_table. Add old entry_pc attribute only
8903 if debug nonbinding markers are disabled.
8904 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8905 markers are enabled.
8906 (block_within_block_p, dwarf2out_inline_entry): New.
8907 (dwarf2out_finish): Check that no entries remained in
8908 inline_entry_data_table.
8909 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8910 (final_scan_insn, notice_source_line): Likewise.
8911 (rest_of_clean_state): Skip inline entry markers.
8912 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8914 * gimple.c (gimple_build_debug_inline_entry): New.
8915 * gimple.h (enum gimple_debug_subcode): Add
8916 GIMPLE_DEBUG_INLINE_ENTRY.
8917 (gimple_build_debug_inline_entry): Declare.
8918 (gimple_debug_inline_entry_p): New.
8919 (gimple_debug_nonbind_marker_p): Adjust.
8920 * insn-notes.def (INLINE_ENTRY): New.
8921 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8922 inline entry marker notes.
8923 (print_insn): Likewise.
8924 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8925 (INSN_DEBUG_MARKER_KIND): Likewise.
8926 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8927 * tree-inline.c (expand_call_inline): Build and insert
8928 debug_inline_entry stmt.
8929 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8930 inline entry blocks early, if nonbind markers are enabled.
8931 (dump_scope_block): Dump fragment info.
8932 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8933 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8934 (gimple_build_debug_inline_entry): New.
8935 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8936 Enable/disable inline entry points too.
8937 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8938 (DEBUG_INSN): Describe inline entry markers.
8940 * common.opt (gvariable-location-views): New.
8941 (gvariable-location-views=incompat5): New.
8942 * config.in: Rebuilt.
8943 * configure: Rebuilt.
8944 * configure.ac: Test assembler for view support.
8945 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8946 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8947 * dwarf2out.c (var_loc_view): New typedef.
8948 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8949 (dwarf2out_locviews_in_attribute): New.
8950 (dwarf2out_locviews_in_loclist): New.
8951 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8952 (enum dw_line_info_opcode): Add LI_adv_address.
8953 (struct dw_line_info_table): Add view.
8954 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8955 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8956 (zero_view_p): New variable.
8957 (ZERO_VIEW_P): New macro.
8958 (output_asm_line_debug_info): New.
8959 (struct var_loc_node): Add view.
8960 (add_AT_view_list, AT_loc_list): New.
8961 (add_var_loc_to_decl): Add view param. Test it against last.
8962 (new_loc_list): Add view params. Record them.
8963 (AT_loc_list_ptr): Handle loc and view lists.
8964 (view_list_to_loc_list_val_node): New.
8965 (print_dw_val): Handle dw_val_class_view_list.
8966 (size_of_die): Likewise.
8967 (value_format): Likewise.
8968 (loc_list_has_views): New.
8969 (gen_llsym): Set vl_symbol too.
8970 (maybe_gen_llsym, skip_loc_list_entry): New.
8971 (dwarf2out_maybe_output_loclist_view_pair): New.
8972 (output_loc_list): Output view list or entries too.
8973 (output_view_list_offset): New.
8974 (output_die): Handle dw_val_class_view_list.
8975 (output_dwarf_version): New.
8976 (output_compilation_unit_header): Use it.
8977 (output_skeleton_debug_sections): Likewise.
8978 (output_rnglists, output_line_info): Likewise.
8979 (output_pubnames, output_aranges): Update version comments.
8980 (output_one_line_info_table): Output view numbers in asm comments.
8981 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8982 Call maybe_gen_llsym.
8983 (loc_list_from_tree_1): Adjust.
8984 (add_AT_location_description): Create view list attribute if
8985 needed, check it's absent otherwise.
8986 (convert_cfa_to_fb_loc_list): Adjust.
8987 (maybe_emit_file): Call output_asm_line_debug_info for test.
8988 (dwarf2out_var_location): Reset views as needed. Precompute
8989 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8990 attribute. Set view.
8991 (new_line_info_table): Reset next view.
8992 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8993 (dwarf2out_source_line): Likewise. Output view resets and labels to
8994 the assembler, or select appropriate line info opcodes.
8995 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8996 (optimize_string_length): Catch it. Adjust.
8997 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8998 dw_val_class_view_list, and remove it if no longer needed.
8999 (hash_loc_list): Hash view numbers.
9000 (loc_list_hasher::equal): Compare them.
9001 (optimize_location_lists): Check whether a view list symbol is
9002 needed, and whether the locview attribute is present, and
9003 whether they match. Remove the locview attribute if no longer
9005 (index_location_lists): Call skip_loc_list_entry for test.
9006 (dwarf2out_finish): Call output_asm_line_debug_info for test.
9007 Use output_dwarf_version.
9008 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
9009 (struct dw_val_node): Add val_view_list.
9010 * final.c (SEEN_NEXT_VIEW): New.
9011 (set_next_view_needed): New.
9012 (clear_next_view_needed): New.
9013 (maybe_output_next_view): New.
9014 (final_start_function): Rename to...
9015 (final_start_function_1): ... this. Take pointer to FIRST,
9016 add SEEN parameter. Emit param bindings in the initial view.
9017 (final_start_function): Reintroduce SEEN-less interface.
9018 (final): Rename to...
9019 (final_1): ... this. Take SEEN parameter. Output final pending
9020 next view at the end.
9021 (final): Reintroduce seen-less interface.
9022 (final_scan_insn): Output pending next view before switching
9023 sections or ending a block. Mark the next view as needed when
9024 outputting variable locations. Notify debug backend of section
9025 changes, and of location view changes.
9026 (rest_of_handle_final): Adjust.
9027 * toplev.c (process_options): Autodetect value for debug variable
9028 location views option. Warn on incompat5 without -gdwarf-5.
9029 * doc/invoke.texi (gvariable-location-views): New.
9030 (gvariable-location-views=incompat5): New.
9031 (gno-variable-location-views): New.
9033 2018-02-08 David Malcolm <dmalcolm@redhat.com>
9035 PR tree-optimization/84136
9036 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
9037 that the result of find_edge is non-NULL.
9039 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9042 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
9043 storing integer register in SImode. Fix cost of 256 and 512
9044 byte aligned SSE register store.
9046 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
9048 * config/i386/i386.c (ix86_multiplication_cost): Fix
9049 multiplication cost for TARGET_AVX512DQ.
9051 2018-02-08 Marek Polacek <polacek@redhat.com>
9053 PR tree-optimization/84238
9054 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
9057 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9059 PR tree-optimization/84265
9060 * tree-vect-stmts.c (vectorizable_store): Don't treat
9061 VMAT_CONTIGUOUS accesses as grouped.
9062 (vectorizable_load): Likewise.
9064 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9066 PR tree-optimization/81635
9067 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
9068 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
9069 (test_round_for_mask): New functions.
9070 (wide_int_cc_tests): Call test_round_for_mask.
9071 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
9072 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
9073 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
9074 range returned by get_range_info.
9076 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9079 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
9080 * symtab.c: Include builtins.h
9081 (symtab_node::output_to_lto_symbol_table_p): Move here
9082 from lto-streamer-out.c:output_symbol_p.
9083 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
9084 (output_symbol_p): Move all logic to symtab.c
9085 (produce_symtab): Update.
9087 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9089 * config/s390/s390-opts.h (enum indirect_branch): Define.
9090 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9091 (s390_indirect_branch_via_thunk)
9092 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9093 (enum s390_indirect_branch_type): Define.
9094 * config/s390/s390.c (struct s390_frame_layout, struct
9095 machine_function): Remove.
9096 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9097 (indirect_branch_table_label_no, indirect_branch_table_name):
9099 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9100 (enum s390_indirect_branch_option): Define.
9101 (s390_return_addr_from_memory): New function.
9102 (s390_handle_string_attribute): New function.
9103 (s390_attribute_table): Add new attribute handler.
9104 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9105 (s390_indirect_branch_via_thunk): New function.
9106 (s390_indirect_branch_via_inline_thunk): New function.
9107 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9108 sibling call optimization for non z10 compiles.
9109 (s390_emit_call): Force indirect branch target to be a single
9110 register. Add r1 clobber for non-z10 compiles.
9111 (s390_emit_epilogue): Emit return jump via return_use expander.
9112 (s390_reorg): Handle JUMP_INSNs as execute targets.
9113 (s390_option_override_internal): Perform validity checks for the
9114 new command line options.
9115 (s390_indirect_branch_attrvalue): New function.
9116 (s390_indirect_branch_settings): New function.
9117 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9118 (s390_output_indirect_thunk_function): New function.
9119 (s390_code_end): Implement target hook.
9120 (s390_case_values_threshold): Implement target hook.
9121 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9123 * config/s390/s390.h (struct s390_frame_layout)
9124 (struct machine_function): Move here from s390.c.
9125 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9126 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9127 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9128 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9129 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9130 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9131 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9132 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9133 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9134 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9135 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9136 (mnemonic attribute): Add values which aren't recognized
9138 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9139 pattern for branch conversion. Fix mnemonic attribute.
9140 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9141 indirect branch via thunk if requested.
9142 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9143 ("*indirect_jump"): Disable for branch conversion using out of
9145 ("indirect_jump_via_thunk<mode>_z10")
9146 ("indirect_jump_via_thunk<mode>")
9147 ("indirect_jump_via_inlinethunk<mode>_z10")
9148 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9149 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9150 ("casesi_jump_via_inlinethunk<mode>_z10")
9151 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9152 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9153 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9154 ("*indirect2_jump"): Disable for branch conversion.
9155 ("casesi_jump"): Turn into expander and expand patterns for branch
9157 ("return_use"): New expander.
9158 ("*return"): Emit return via thunk and rename it to ...
9159 ("*return<mode>"): ... this one.
9160 * config/s390/s390.opt: Add new options and and enum for the
9163 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9165 * lra-constraints.c (match_reload): Unconditionally use
9166 gen_lowpart_SUBREG, rather than selecting between that
9167 and equivalent gen_rtx_SUBREG code.
9169 2018-02-08 Richard Biener <rguenther@suse.de>
9171 PR tree-optimization/84233
9172 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9173 changed flag instead of boguously re-using phi_inserted.
9175 2018-02-08 Martin Jambor <mjambor@suse.cz>
9177 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9178 static local variables.
9180 2018-02-08 Richard Biener <rguenther@suse.de>
9182 PR tree-optimization/84278
9183 * tree-vect-stmts.c (vectorizable_store): When looking for
9184 smaller vector types to perform grouped strided loads/stores
9185 make sure the mode is supported by the target.
9186 (vectorizable_load): Likewise.
9188 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9190 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9191 Increase LDP/STP opportunities by adding adjacent callee-saves.
9193 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9195 PR rtl-optimization/84068
9196 PR rtl-optimization/83459
9197 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9199 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9201 PR tree-optimization/84224
9202 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9203 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9206 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9209 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9210 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9211 count, remove one USE.
9213 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9215 * doc/install.texi (Configuration): Document the
9216 --with-long-double-format={ibm,ieee} PowerPC configuration
9220 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9221 Convert from define_expand to be define_insn_and_split. Rework
9222 float/double/_Float128 conversions to QI/HI/SImode to work with
9223 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9224 conversions to QI/HImode types did a store and then a load to
9225 truncate the value. For conversions to VSX registers, don't split
9226 the insn, instead emit the code directly. Use the code iterator
9227 any_fix to combine signed and unsigned conversions.
9228 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9229 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9230 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9231 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9232 (fix_<mode>di2_hw): Likewise.
9233 (fixuns_<mode>di2_hw): Likewise.
9234 (fix_<mode>si2_hw): Likewise.
9235 (fixuns_<mode>si2_hw): Likewise.
9236 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9237 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9238 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9239 fix<uns>_trunc<SFDF:mode>si2_p8.
9240 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9241 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9242 (fix<uns>_<mode>_mem): Likewise.
9243 (fctiw<u>z_<mode>_mem): Likewise.
9244 (fix<uns>_<mode>_mem): Likewise.
9245 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9246 the register allocator from doing a direct move to the GPRs to do
9247 a store, and instead use the ISA 3.0 store byte/half-word from
9248 vector register instruction. For IEEE 128-bit floating point,
9249 also optimize stores of 32-bit ints.
9250 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9252 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9254 * genextract.c (push_pathstr_operand): New function to support
9256 (walk_rtx): Call push_pathstr_operand.
9257 (print_path): Support [a-zA-Z].
9259 2018-02-07 Richard Biener <rguenther@suse.de>
9261 PR tree-optimization/84037
9262 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9263 (cse_and_gimplify_to_preheader): Declare.
9264 (vect_get_place_in_interleaving_chain): Likewise.
9265 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9267 (_loop_vec_info::~_loop_vec_info): Delete it.
9268 (cse_and_gimplify_to_preheader): New function.
9269 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9270 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9271 (vectorizable_load): Likewise. For grouped stores always base
9272 the IV on the first element.
9273 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9274 condition before gimplifying.
9276 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9278 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9279 *DIV_EXPR and *MOD_EXPR.
9281 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9284 * config/i386/i386.c (ix86_option_override_internal): Mask out
9285 the CF_SET bit when checking -fcf-protection.
9287 2018-02-07 Tom de Vries <tom@codesourcery.com>
9290 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9293 2018-02-07 Richard Biener <rguenther@suse.de>
9295 PR tree-optimization/84204
9296 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9299 PR tree-optimization/84205
9300 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9301 special-case isl_ast_op_zdiv_r.
9303 PR tree-optimization/84223
9304 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9305 Only add conditions from within the region.
9306 (gather_bbs::after_dom_children): Adjust.
9308 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9311 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9312 * config/avr/avr.md: Only post-reload split REG-REG moves if
9313 either register is GENERAL_REG_P.
9315 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9317 PR tree-optimization/84235
9318 * tree-ssa-scopedtables.c
9319 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9320 if the subtraction is performed in floating point type where NaNs are
9321 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9322 build 1. Formatting fix.
9324 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9327 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9328 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9329 and skip it regardless of bb boundaries. Use CALL_P macro,
9330 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9331 unnecessarily, formatting fix.
9333 2018-02-06 Michael Collison <michael.collison@arm.com>
9335 * config/arm/thumb2.md:
9336 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9337 (*thumb_mov_notscc): Ditto.
9339 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9342 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9343 unsigned_fix, not "s".
9345 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9347 * configure.ac (gcc_fn_eh_frame_ro): New function.
9348 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9349 correct .eh_frame permissions.
9350 * configure: Regenerate.
9352 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9354 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9357 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9359 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9360 Display warning message for -mno-speculate-indirect-jumps.
9362 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9364 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9366 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9368 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9370 PR tree-optimization/84225
9371 * tree-eh.c (find_trapping_overflow): Only call
9372 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9374 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9377 * config/i386/i386.c: Reimplement the check of possible options
9378 -mibt/-mshstk conbination. Change error messages.
9379 * doc/invoke.texi: Fix a typo: remove extra '='.
9381 2018-02-06 Marek Polacek <polacek@redhat.com>
9383 PR tree-optimization/84228
9384 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9386 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9389 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9390 emitted arch directives.
9391 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9392 __ARM_FEATURE_COPROC before changing architectures.
9394 2018-02-06 Richard Biener <rguenther@suse.de>
9396 * config/i386/i386.c (print_reg): Fix typo.
9397 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9399 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9401 * configure: Regenerate.
9403 2018-02-05 Martin Sebor <msebor@redhat.com>
9405 PR tree-optimization/83369
9406 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9409 2018-02-05 Martin Liska <mliska@suse.cz>
9411 * doc/invoke.texi: Cherry-pick upstream r323995.
9413 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9415 * ira.c (ira_init_register_move_cost): Adjust comment.
9417 2018-02-05 Martin Liska <mliska@suse.cz>
9419 PR gcov-profile/84137
9420 * doc/gcov.texi: Fix typo in documentation.
9422 2018-02-05 Martin Liska <mliska@suse.cz>
9424 PR gcov-profile/83879
9425 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9426 using dlopen functionality.
9428 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9430 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9431 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9432 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9433 _mm_maskz_range_round_ss): New intrinsics.
9434 (__builtin_ia32_rangesd128_round)
9435 (__builtin_ia32_rangess128_round): Remove.
9436 (__builtin_ia32_rangesd128_mask_round,
9437 __builtin_ia32_rangess128_mask_round): New builtins.
9438 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9439 __builtin_ia32_rangess128_round): Remove.
9440 (__builtin_ia32_rangesd128_mask_round,
9441 __builtin_ia32_rangess128_mask_round): New builtins.
9442 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9443 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9444 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9445 "<round_saeonly_constraint>")): Changed to ...
9446 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9447 "<round_saeonly_scalar_constraint>")): ... this.
9448 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9449 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9450 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9451 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9452 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9454 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9456 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9458 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9459 Remove all values except native, 8540 and 8548.
9461 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9463 * config/i386/i386.c (ix86_output_function_return): Pass
9464 INVALID_REGNUM, instead of -1, as invalid register number to
9465 indirect_thunk_name and output_indirect_thunk.
9467 2018-02-02 Julia Koval <julia.koval@intel.com>
9469 * config.gcc: Add -march=icelake.
9470 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9471 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9472 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9473 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9474 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9475 (processor_target_table): Add icelake.
9476 (ix86_option_override_internal): Handle new PTAs.
9477 (get_builtin_code_for_version): Handle icelake.
9478 (M_INTEL_COREI7_ICELAKE): New.
9479 (fold_builtin_cpu): Handle icelake.
9480 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9481 * doc/invoke.texi: Add -march=icelake.
9483 2018-02-02 Julia Koval <julia.koval@intel.com>
9485 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9486 to wide_int_bitmask.
9487 * wide-int-bitmask.h: New.
9489 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9492 * config/i386/i386.md: Replace Pmode with word_mode in
9493 builtin_setjmp_setup and builtin_longjmp to support x32.
9495 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9499 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9501 (rs6000_supported_cpu_names): New static variable.
9502 (linux_cpu_translation_table): Likewise.
9503 (elf_platform) <cpu>: Define new static variable and use it.
9504 Translate kernel AT_PLATFORM name to canonical name if needed.
9505 Error if platform name is unknown.
9507 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9510 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9512 2018-02-01 Jeff Law <law@redhat.com>
9515 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9516 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9517 the scratch if RELEASE_VIA_POP is false.
9518 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9519 If we have to save a temporary register, decrement SIZE appropriately.
9520 Pass new arguments to release_scratch_register_on_entry.
9521 (ix86_adjust_stack_and_probe): Likewise.
9522 (ix86_emit_probe_stack_range): Pass new arguments to
9523 release_scratch_register_on_entry.
9525 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9527 PR rtl-optimization/84157
9528 * combine.c (change_zero_ext): Use REG_P predicate in
9529 front of HARD_REGISTER_P predicate.
9531 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9533 * config/avr/avr.c (avr_option_override): Move disabling of
9534 -fdelete-null-pointer-checks to...
9535 * common/config/avr/avr-common.c (avr_option_optimization_table):
9538 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9540 PR tree-optimization/81635
9541 * tree-data-ref.c (split_constant_offset_1): For types that
9542 wrap on overflow, try to use range info to prove that wrapping
9545 2018-02-01 Renlin Li <renlin.li@arm.com>
9548 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9550 (aarch64_register_move_cost): Likewise.
9551 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9553 (REG_CLASS_NAMES): Likewise.
9554 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9555 TAILCALL_ADDR_REGS. Remove IP registers.
9556 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9558 2018-02-01 Richard Biener <rguenther@suse.de>
9560 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9561 for specifying RPO order and allow NULL for that.
9562 * domwalk.c (dom_walker::dom_walker): Likewise.
9563 (dom_walker::walk): Handle NULL RPO order.
9564 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9566 (rewrite_update_dom_walker): Likewise.
9567 (mark_def_dom_walker): Likewise.
9569 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9571 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9572 (aarch64_maybe_expand_sve_subreg_move): Declare.
9573 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9574 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9576 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9577 that are semantically a reverse operation.
9578 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9579 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9580 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9582 (aarch64_can_change_mode_class): For big-endian, forbid changes
9583 between two SVE modes if they have different element sizes.
9585 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9587 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9588 the TImode handling for big-endian targets.
9590 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9592 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9593 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9595 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9596 Remove BSWAP handing for big-endian targets and use the form of
9597 LD1RQ appropariate for the mode.
9599 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9601 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9602 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9605 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9608 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9609 check for operands that need to go through aarch64_sve_reload_be.
9611 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9613 PR tree-optimization/81661
9614 PR tree-optimization/84117
9615 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9616 * tree-eh.c: Include gimplify.h.
9617 (find_trapping_overflow, replace_trapping_overflow,
9618 rewrite_to_non_trapping_overflow): New functions.
9619 * tree-vect-loop.c: Include tree-eh.h.
9620 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9621 * tree-data-ref.c: Include tree-eh.h.
9622 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9624 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9626 PR rtl-optimization/84123
9627 * combine.c (change_zero_ext): Check if hard register satisfies
9628 can_change_dest_mode before calling gen_lowpart_SUBREG.
9630 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9633 * ira.c (ira_init_register_move_cost): Remove assert.
9635 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9637 PR rtl-optimization/84071
9638 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9639 * doc/tm.texi: Regenerate.
9641 2018-01-31 Richard Biener <rguenther@suse.de>
9643 PR tree-optimization/84132
9644 * tree-data-ref.c (analyze_miv_subscript): Properly
9645 check whether evolution_function_is_affine_multivariate_p
9646 before calling gcd_of_steps_may_divide_p.
9648 2018-01-31 Julia Koval <julia.koval@intel.com>
9651 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9652 * config/i386/i386.md (rdpid_rex64) New.
9653 (rdpid): Make 32bit only.
9655 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9658 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9659 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9661 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9664 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9666 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9668 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9670 PR rtl-optimization/84071
9671 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9672 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9674 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9676 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9677 (arc_attribute_table): Add 'aux' attribute.
9678 (arc_in_small_data_p): Consider aux like variables.
9679 (arc_is_aux_reg_p): New function.
9680 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9681 (arc_get_aux_arg): New function.
9682 (prepare_move_operands): Handle aux-register access.
9683 (arc_handle_aux_attribute): New function.
9684 * doc/extend.texi (ARC Variable attributes): Add subsection.
9686 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9688 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9689 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9690 (arc_attribute_table): Add 'uncached' attribute.
9691 (arc_print_operand): Print '.di' flag for uncached memory
9693 (arc_in_small_data_p): Do not consider for small data the uncached
9695 (arc_is_uncached_mem_p): New function.
9696 * config/arc/predicates.md (compact_store_memory_operand): Check
9697 for uncached memory accesses.
9698 (nonvol_nonimm_operand): Likewise.
9699 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9701 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9704 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9705 falign-loops=): Add Optimization flag.
9707 2018-01-30 Jeff Law <law@redhat.com>
9710 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9711 INT_REGISTERS_SAVED. Check it prior to calling
9712 get_scratch_register_on_entry.
9713 (ix86_adjust_stack_and_probe): Similarly.
9714 (ix86_emit_probe_stack_range): Similarly.
9715 (ix86_expand_prologue): Corresponding changes.
9717 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9720 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9721 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9723 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9726 * lra-constraints.c (curr_insn_transform): Process AND in the
9729 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9731 PR rtl-optimization/83986
9732 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9733 dependence against last_pending_memory_flush in addition to
9736 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9738 PR tree-optimization/81611
9739 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9742 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9745 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9748 2018-01-30 Richard Biener <rguenther@suse.de>
9749 Jakub Jelinek <jakub@redhat.com>
9751 PR tree-optimization/84111
9752 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9753 inner loops added during recursion, as they don't have up-to-date
9756 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9759 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9760 (can_inline_edge_by_limits_p): ... here.
9761 (can_early_inline_edge_p, check_callers,
9762 update_caller_keys, update_callee_keys, recursive_inlining,
9763 add_new_edges_to_heap, speculation_useful_p,
9764 inline_small_functions,
9765 inline_small_functions, flatten_function,
9766 inline_to_all_callers_1): Update.
9768 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9770 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9773 2018-01-30 Richard Biener <rguenther@suse.de>
9775 PR tree-optimization/83008
9776 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9777 invariant and constant vector uses in stmts when they need
9780 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9783 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9784 * configure: Regenerate.
9786 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9788 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9790 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9791 Use gen_rtx_REG rather than gen_lowpart.
9793 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9795 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9796 rather than 0 when creating partial subregs.
9798 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9800 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9803 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9806 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9807 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9808 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9809 flags. This restores the settings used before the 2017-07-24.
9810 Turning off pre increment/decrement/modify allows IVOPTS to
9811 optimize DF/SF loops where the index is an int.
9813 2018-01-29 Richard Biener <rguenther@suse.de>
9814 Kelvin Nilsen <kelvin@gcc.gnu.org>
9817 * tree-vect-stmts.c (vectorizable_call): Don't call
9818 targetm.vectorize_builtin_md_vectorized_function if callee is
9821 2018-01-22 Carl Love <cel@us.ibm.com>
9823 * doc/extend.tex: Fix typo in second arg in
9824 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9826 2018-01-29 Richard Biener <rguenther@suse.de>
9828 PR tree-optimization/84086
9829 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9830 (flush_ssaname_freelist): When SSA names were released reset
9831 the SCEV hash table.
9833 2018-01-29 Richard Biener <rguenther@suse.de>
9835 PR tree-optimization/84057
9836 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9837 removed paths when removing edges.
9839 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9841 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9842 -mfunction-return=@var{choice}.
9844 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9847 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9849 (layout::print_source_line): Likewise.
9850 (test_get_line_width_without_trailing_whitespace): Add test cases.
9852 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9855 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9858 2018-01-26 Jim Wilson <jimw@sifive.com>
9860 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9862 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9865 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9867 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9868 and CMP + SUB-immediate -> SUBS.
9870 2018-01-26 Martin Sebor <msebor@redhat.com>
9872 PR tree-optimization/83896
9873 * tree-ssa-strlen.c (get_string_len): Rename...
9874 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9875 Avoid assuming length is constant.
9876 (handle_char_store): Use HOST_WIDE_INT for string length.
9878 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9881 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9882 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9884 2018-01-26 Richard Biener <rguenther@suse.de>
9886 PR rtl-optimization/84003
9887 * dse.c (record_store): Only record redundant stores when
9888 the earlier store aliases at least all accesses the later one does.
9890 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9892 PR rtl-optimization/83985
9893 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9894 REG_CFA_RESTORE insns.
9895 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9896 REG_CFA_RESTORE insns here.
9899 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9900 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9902 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9904 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9905 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9906 (arc_init): Likewise.
9907 (arc_override_options): Likewise.
9908 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9910 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9912 * config/arc/arc.h (TARGET_DBNZ): Define.
9913 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9914 properly set the tune attribute.
9915 (dbnz): Use TARGET_DBNZ guard.
9916 * config/arc/arc.opt (mtune): Add core3 option.
9918 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9920 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9921 recognize new pic like addresses.
9922 (arc_delegitimize_address): Clean up.
9924 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9926 * config/arc/arc-arches.def: Option mrf16 valid for all
9928 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9929 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9930 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9931 * config/arc/arc-tables.opt: Regenerate.
9932 * config/arc/arc.c (arc_conditional_register_usage): Handle
9933 reduced register file case.
9934 (arc_file_start): Set must have build attributes.
9935 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9937 * config/arc/arc.opt (mrf16): Add new option.
9938 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9939 * config/arc/genmultilib.awk: Handle new mrf16 option.
9940 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9941 * config/arc/t-multilib: Regenerate.
9942 * doc/invoke.texi (ARC Options): Document mrf16 option.
9944 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9946 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9947 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9948 (arc_attribute_table): Add 'secure_call' attribute.
9949 (arc_print_operand): Print secure call operand.
9950 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9952 (arc_is_secure_call_p): New function. * config/arc/arc.md
9953 (call_i): Add support for sjli instruction.
9954 (call_value_i): Likewise.
9955 * config/arc/constraints.md (Csc): New constraint.
9957 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9958 John Eric Martin <John.Martin@emmicro-us.com>
9960 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9961 * config/arc/arc.c (_arc_jli_section): New struct.
9962 (arc_jli_section): New type.
9963 (rc_jli_sections): New static variable.
9964 (arc_handle_jli_attribute): New function.
9965 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9966 (arc_file_end): New function.
9967 (TARGET_ASM_FILE_END): Define.
9968 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9969 (arc_add_jli_section): New function.
9970 (jli_call_scan): Likewise.
9971 (arc_reorg): Call jli_call_scan.
9972 (arc_output_addsi): Remove 'S' from printing asm operand.
9973 (arc_is_jli_call_p): New function.
9974 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9976 (movhi_insn): Likewise.
9977 (movsi_insn): Likewise.
9978 (movsi_set_cc_insn): Likewise.
9979 (loadqi_update): Likewise.
9980 (load_zeroextendqisi_update): Likewise.
9981 (load_signextendqisi_update): Likewise.
9982 (loadhi_update): Likewise.
9983 (load_zeroextendhisi_update): Likewise.
9984 (load_signextendhisi_update): Likewise.
9985 (loadsi_update): Likewise.
9986 (loadsf_update): Likewise.
9987 (movsicc_insn): Likewise.
9988 (bset_insn): Likewise.
9989 (bxor_insn): Likewise.
9990 (bclr_insn): Likewise.
9991 (bmsk_insn): Likewise.
9992 (bicsi3_insn): Likewise.
9993 (cmpsi_cc_c_insn): Likewise.
9994 (movsi_ne): Likewise.
9995 (movsi_cond_exec): Likewise.
9996 (clrsbsi2): Likewise.
10007 (call_i): Remove 'S' asm letter, add jli instruction.
10008 (call_value_i): Likewise.
10009 * config/arc/arc.op (mjli-always): New option.
10010 * config/arc/constraints.md (Cji): New constraint.
10011 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
10013 (subsf3_fpx): Likewise.
10014 (mulsf3_fpx): Likewise.
10015 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
10017 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
10018 function attrbutes.
10019 * doc/invoke.texi (ARC): Document mjli-always option.
10021 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
10023 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
10024 avoid addition with 0 and use incw and decw where possible.
10026 2018-01-26 Richard Biener <rguenther@suse.de>
10028 PR tree-optimization/81082
10029 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
10030 association if it requires casting to unsigned.
10031 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
10032 from fold_plusminus_mult_expr to catch important cases late when
10033 range info is available.
10035 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10037 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
10038 * configure.ac (hidden_linkonce): New test.
10039 * configure: Regenerate.
10040 * config.in: Regenerate.
10042 2018-01-26 Julia Koval <julia.koval@intel.com>
10044 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
10045 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
10046 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
10047 _mm_mask_bitshuffle_epi64_mask): Fix type.
10048 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
10049 USI_FTYPE_V4DI_V4DI_USI): Remove.
10050 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
10051 __builtin_ia32_vpshufbitqmb256_mask,
10052 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
10053 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
10054 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
10056 2018-01-26 Alan Modra <amodra@gmail.com>
10059 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
10060 UNSPEC_VBPERMQ. Sort other unspecs.
10062 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
10064 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
10066 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
10068 PR middle-end/83055
10069 * predict.c (drop_profile): Do not push/pop cfun; update also
10071 (handle_missing_profiles): Fix logic looking for zero profiles.
10073 2018-01-25 Jakub Jelinek <jakub@redhat.com>
10075 PR middle-end/83977
10076 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
10077 on functions with #pragma omp declare simd or functions with simd
10079 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
10080 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
10081 Remove trailing \n from warning_at calls.
10083 2018-01-25 Tom de Vries <tom@codesourcery.com>
10086 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10087 for neutered workers.
10089 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10092 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10093 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10095 2018-01-24 Jeff Law <law@redhat.com>
10098 * i386.c (get_probe_interval): Move to earlier point.
10099 (ix86_compute_frame_layout): If -fstack-clash-protection and
10100 the frame is larger than the probe interval, then use pushes
10101 to save registers rather than reg->mem moves.
10102 (ix86_expand_prologue): Remove conditional for int_registers_saved
10105 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10108 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10109 min/max for never referenced object.
10111 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10113 PR middle-end/83977
10114 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10116 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10117 attributes from DECL_ATTRIBUTES (decl) without affecting
10118 DECL_ATTRIBUTES (current_function_decl).
10119 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10120 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10122 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10124 PR tree-optimization/83979
10125 * fold-const.c (fold_comparison): Use constant_boolean_node
10126 instead of boolean_{true,false}_node.
10128 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10130 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10133 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10135 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10136 Simplify the clause that sets the length attribute.
10137 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10138 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10139 clause that sets the length attribute.
10140 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10142 2018-01-24 Tom de Vries <tom@codesourcery.com>
10145 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10146 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10147 Add strict parameter.
10148 (prevent_branch_around_nothing): Insert dummy insn between branch to
10149 label and label with no ptx insn inbetween.
10150 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10152 2018-01-24 Tom de Vries <tom@codesourcery.com>
10155 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10156 for neutered threads in warp.
10157 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10159 2018-01-24 Richard Biener <rguenther@suse.de>
10161 PR tree-optimization/83176
10162 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10165 2018-01-24 Richard Biener <rguenther@suse.de>
10167 PR tree-optimization/82819
10168 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10169 code generating pluses that are no-ops in the target precision.
10171 2018-01-24 Richard Biener <rguenther@suse.de>
10173 PR middle-end/84000
10174 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10176 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10178 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10179 to merge probabilities.
10180 * predict.c (probably_never_executed): Also mark as cold functions
10181 with global 0 profile and guessed local profile.
10182 * profile-count.c (profile_probability::combine_with_count): New
10184 * profile-count.h (profile_probability::operator*,
10185 profile_probability::operator*=, profile_probability::operator/,
10186 profile_probability::operator/=): Reduce precision to adjusted
10187 and set value to guessed on contradictory divisions.
10188 (profile_probability::combine_with_freq): Remove.
10189 (profile_probability::combine_wiht_count): Declare.
10190 (profile_count::force_nonzero):: Set to adjusted.
10191 (profile_count::probability_in):: Set quality to adjusted.
10192 * tree-ssa-tail-merge.c (replace_block_by): Use
10193 combine_with_count.
10195 2018-01-23 Andrew Waterman <andrew@sifive.com>
10196 Jim Wilson <jimw@sifive.com>
10198 * config/riscv/riscv.c (riscv_stack_boundary): New.
10199 (riscv_option_override): Set riscv_stack_boundary. Handle
10200 riscv_preferred_stack_boundary_arg.
10201 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10202 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10203 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10204 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10205 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10206 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10208 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10211 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10212 of struct ix86_frame.
10213 (ix86_expand_epilogue): Likewise. Add a local variable for
10214 the reg_save_offset field in struct ix86_frame.
10216 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10218 PR tree-optimization/82604
10219 * tree-loop-distribution.c (enum partition_kind): New enum item
10220 PKIND_PARTIAL_MEMSET.
10221 (partition_builtin_p): Support above new enum item.
10222 (generate_code_for_partition): Ditto.
10223 (compute_access_range): Differentiate cases that equality can be
10224 proven at all loops, the innermost loops or no loops.
10225 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10226 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10227 (finalize_partitions, distribute_loop): Don't fuse partition of
10228 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10229 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10230 parloop is enabled.
10232 2018-01-23 Martin Liska <mliska@suse.cz>
10234 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10235 order to ignore the predictor.
10236 (PRED_POLYMORPHIC_CALL): Likewise.
10237 (PRED_RECURSIVE_CALL): Likewise.
10239 2018-01-23 Martin Liska <mliska@suse.cz>
10241 * tree-profile.c (tree_profiling): Print function header to
10242 aware reader which function we are working on.
10243 * value-prof.c (gimple_find_values_to_profile): Do not print
10244 not interesting value histograms.
10246 2018-01-23 Martin Liska <mliska@suse.cz>
10248 * profile-count.h (enum profile_quality): Add
10249 profile_uninitialized as the first value. Do not number values
10250 as they are zero based.
10251 (profile_count::verify): Update sanity check.
10252 (profile_probability::verify): Likewise.
10254 2018-01-23 Nathan Sidwell <nathan@acm.org>
10256 * doc/invoke.texi (ffor-scope): Deprecate.
10258 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10260 PR tree-optimization/83510
10261 * domwalk.c (set_all_edges_as_executable): New function.
10262 (dom_walker::dom_walker): Convert bool param
10263 "skip_unreachable_blocks" to enum reachability. Move setup of
10264 edge flags to set_all_edges_as_executable and only do it when
10265 reachability is REACHABLE_BLOCKS.
10266 * domwalk.h (enum dom_walker::reachability): New enum.
10267 (dom_walker::dom_walker): Convert bool param
10268 "skip_unreachable_blocks" to enum reachability.
10269 (set_all_edges_as_executable): New decl.
10270 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10271 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10273 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10274 but converting true to REACHABLE_BLOCKS.
10275 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10277 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10278 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10279 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10281 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10282 if check_all_array_refs will be called.
10284 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10286 * tree.c (selftest::test_location_wrappers): Add more test
10289 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10291 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10292 (selftest::test_bit_in_range): Likewise.
10294 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10297 * doc/sourcebuild.texi (vect_float): Say that the selector
10298 only describes the situation when -funsafe-math-optimizations is on.
10299 (vect_float_strict): Document.
10301 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10303 PR tree-optimization/83965
10304 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10305 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10306 instead of checking only for a reduction.
10307 (vect_recog_widen_sum_pattern): Likewise.
10309 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10311 * predict.c (probably_never_executed): Only use precise profile info.
10312 (compute_function_frequency): Skip after inlining hack since we now
10313 have quality checking.
10315 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10317 * profile-count.h (profile_probability::very_unlikely,
10318 profile_probability::unlikely, profile_probability::even): Set
10319 precision to guessed.
10321 2018-01-23 Richard Biener <rguenther@suse.de>
10323 PR tree-optimization/83963
10324 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10325 Properly terminate dominator walk when crossing the exit edge not
10326 when visiting its source block.
10328 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10331 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10332 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10334 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10336 PR tree-optimization/83957
10337 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10338 semicolon after for body surrounded by braces.
10340 PR tree-optimization/83081
10341 * profile-count.h (profile_probability::split): New method.
10342 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10343 Use profile_probability::split.
10344 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10345 when splitting a single conditional jump into 2.
10347 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10349 PR tree-optimization/69452
10350 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10353 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10355 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10356 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10357 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10359 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10361 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10363 * config/rl78/rl78.md (movdi): New define_expand.
10364 * config/rl78/rl78.c (rl78_split_movdi): New function.
10366 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10369 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10371 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10372 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10373 128-bit to produce an UNSPEC move to get the double word with the
10374 signbit and then a shift directly to do signbit.
10375 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10376 implementation with a new version that just does either a direct
10377 move or a regular move. Move memory interface to separate insns.
10378 Move insns so they are next to the expander.
10379 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10380 with signbit move. Split big and little endian case.
10381 (signbit<mode>2_dm_mem_le): Likewise.
10382 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10383 (signbit<mode>2_dm2): Likewise.
10385 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10387 * config/rl78/rl78.md (anddi3): New define_expand.
10389 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10391 * config/rl78/rl78.md (umindi3): New define_expand.
10393 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10395 * config/rl78/rl78.md (smindi3): New define_expand.
10397 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10399 * config/rl78/rl78.md (smaxdi3): New define_expand.
10401 2018-01-22 Carl Love <cel@us.ibm.com>
10403 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10404 LVX_V1TI): Add macro expansion.
10405 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10406 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10407 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10408 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10409 Change check to determine if the instruction is a byte reversing
10410 entry. Fix typo in comment.
10411 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10412 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10413 Add def_builtin calls for new builtins.
10414 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10415 Add define_insn expansion.
10417 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10419 * config/rl78/rl78.md (umaxdi3): New define_expand.
10421 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10423 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10424 for non-QImode registers.
10426 2018-01-22 Richard Biener <rguenther@suse.de>
10428 PR tree-optimization/83963
10429 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10430 including the loop exit block.
10431 (scop_detection::merge_sese): Likewise.
10432 (scop_detection::add_scop): Do it here instead.
10434 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10436 * doc/sourcebuild.texi (arm_softfloat): Document.
10438 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10441 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10442 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10443 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10445 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10446 David Edelsohn <dje.gcc@gmail.com>
10449 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10450 Change "crset eq" to "crset 2".
10451 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10452 (*call_indirect_aix<mode>_nospec): Likewise.
10453 (*call_value_indirect_aix<mode>_nospec): Likewise.
10454 (*call_indirect_elfv2<mode>_nospec): Likewise.
10455 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10456 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10457 change assembly output from . to $.
10458 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10459 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10460 (*tablejump<mode>_internal1_nospec): Likewise.
10462 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10465 * config/sh/sh_optimize_sett_clrt.cc:
10466 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10468 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10470 PR tree-optimization/83940
10471 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10472 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10473 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10474 use it to pass back the definition type.
10475 (vect_check_store_rhs): Likewise rhs_dt_out.
10476 (vect_build_gather_load_calls): Add a mask_dt argument and use
10477 it instead of a call to vect_is_simple_use.
10478 (vectorizable_store): Update calls to vect_check_load_store_mask
10479 and vect_check_store_rhs. Use the dt returned by the latter instead
10480 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10481 instead of calls to vect_is_simple_use. Pass the scalar rather
10482 than the vector operand to vect_is_simple_use when handling
10483 second and subsequent copies of an rhs value.
10484 (vectorizable_load): Update calls to vect_check_load_store_mask
10485 and vect_build_gather_load_calls. Use the cached mask_dt and
10486 gs_info.offset_dt instead of calls to vect_is_simple_use.
10488 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10490 PR middle-end/83945
10491 * tree-emutls.c: Include gimplify.h.
10492 (lower_emutls_2): New function.
10493 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10494 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10495 it before further processing.
10498 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10499 UINTVAL (trueop1) instead of INTVAL (op1).
10501 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10505 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10506 INCOMING_FRAME_SP_OFFSET if not defined.
10507 (scan_trace): Add ENTRY argument. If true and
10508 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10509 emit a note to adjust the CFA offset.
10510 (create_cfi_notes): Adjust scan_trace callers.
10511 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10512 INCOMING_FRAME_SP_OFFSET in the CIE.
10513 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10514 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10516 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10517 * doc/tm.texi: Regenerated.
10519 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10521 PR rtl-optimization/83147
10522 * lra-constraints.c (remove_inheritance_pseudos): Use
10523 lra_substitute_pseudo_within_insn.
10525 2018-01-19 Tom de Vries <tom@codesourcery.com>
10526 Cesar Philippidis <cesar@codesourcery.com>
10529 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10531 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10534 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10535 spaces for function labels.
10537 2018-01-19 Martin Liska <mliska@suse.cz>
10539 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10540 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10541 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10542 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10543 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10544 (PRED_CONST_RETURN): Change from 69 to 65.
10545 (PRED_NULL_RETURN): Change from 91 to 71.
10546 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10547 (PRED_LOOP_GUARD): Change from 66 to 73.
10549 2018-01-19 Martin Liska <mliska@suse.cz>
10551 * predict.c (predict_insn_def): Add new assert.
10552 (struct branch_predictor): Change type to signed integer.
10553 (test_prediction_value_range): Amend test to cover
10554 PROB_UNINITIALIZED.
10555 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10556 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10557 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10558 (PRED_LOOP_IV_COMPARE): Likewise.
10559 * predict.h (PROB_UNINITIALIZED): Define new constant.
10561 2018-01-19 Martin Liska <mliska@suse.cz>
10563 * predict.c (dump_prediction): Add new format for
10564 analyze_brprob.py script which is enabled with -details
10566 * profile-count.h (precise_p): New function.
10568 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10570 PR tree-optimization/83922
10571 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10572 there are no statements that need masking.
10573 (vect_active_double_reduction_p): New function.
10574 (vect_analyze_loop_operations): Use it when handling phis that
10575 are not in the loop header.
10577 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10579 PR tree-optimization/83914
10580 * tree-vect-loop.c (vectorizable_induction): Don't convert
10581 init_expr or apply the peeling adjustment for inductions
10582 that are nested within the vectorized loop.
10584 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10586 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10589 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10593 * function.h (gimplify_parameters): Add gimple_seq * argument.
10594 * function.c: Include gimple.h and options.h.
10595 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10596 for the added local temporaries if needed.
10597 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10598 if there are any parameter cleanups, wrap whole body into a
10599 try/finally with the cleanups.
10601 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10604 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10605 Use GET_MODE_CLASS for scalar floating point.
10607 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10611 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10612 Fix call of call_cgraph_insertion_hooks.
10614 2018-01-18 Martin Sebor <msebor@redhat.com>
10616 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10618 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10621 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10624 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10627 * common.opt: (-ffile-prefix-map): New option.
10628 * opts.c (common_handle_option): Defer it.
10629 * opts-global.c (handle_common_deferred_options): Handle it.
10630 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10631 * file-prefix-map.h: New file.
10632 (remap_debug_filename, add_debug_prefix_map): ...here.
10633 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10634 * final.c (debug_prefix_map, add_debug_prefix_map
10635 remap_debug_filename): Move to...
10636 * file-prefix-map.c: New file.
10637 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10638 generalize, get rid of alloca(), use strrchr() instead of strchr().
10639 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10640 Implement in terms of add_prefix_map().
10641 (remap_macro_filename, remap_debug_filename): Implement in term of
10643 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10644 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10645 * dbxout.c: Include file-prefix-map.h.
10646 * varasm.c: Likewise.
10647 * vmsdbgout.c: Likewise.
10648 * xcoffout.c: Likewise.
10649 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10650 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10651 * doc/invoke.texi (-ffile-prefix-map): Document.
10652 (-fdebug-prefix-map): Update description.
10654 2018-01-18 Martin Liska <mliska@suse.cz>
10656 * config/i386/i386.c (indirect_thunk_name): Document that also
10658 (output_indirect_thunk): Document why both instructions
10659 (pause and lfence) are generated.
10661 2018-01-18 Richard Biener <rguenther@suse.de>
10663 PR tree-optimization/83887
10664 * graphite-scop-detection.c
10665 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10666 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10667 (scop_detection::merge_sese): Re-implement with a flood-fill
10668 algorithm that properly finds a SESE region if it exists.
10670 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10673 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10674 pointer_diff optimizations use view_convert instead of convert.
10676 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10678 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10679 Generate different code for -mno-speculate-indirect-jumps.
10680 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10681 (*call_indirect_aix<mode>): Disable for
10682 -mno-speculate-indirect-jumps.
10683 (*call_indirect_aix<mode>_nospec): New define_insn.
10684 (*call_value_indirect_aix<mode>): Disable for
10685 -mno-speculate-indirect-jumps.
10686 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10687 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10688 -mno-speculate-indirect-jumps.
10689 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10691 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10693 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10694 long double type, set the flags for noting the default long double
10695 type, even if we don't pass or return a long double type.
10697 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10700 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10703 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10705 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10706 support for merge[hl].
10707 (fold_mergehl_helper): New helper function.
10708 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10709 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10710 (altivec_vmrglw_direct): Add xxmrglw insn.
10712 2018-01-17 Andrew Waterman <andrew@sifive.com>
10714 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10715 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10717 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10720 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10721 call the lto_location_cache before reading the
10722 DECL_SOURCE_LOCATION of the types.
10724 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10725 Richard Sandiford <richard.sandiford@linaro.org>
10727 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10728 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10729 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10730 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10731 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10733 * config/aarch64/constraints.md (aarch64_movti_operand):
10735 * config/aarch64/predicates.md (Uti): Add new constraint.
10737 2018-01-17 Carl Love <cel@us.ibm.com>
10739 * config/rs6000/vsx.md (define_expand xl_len_r,
10740 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10741 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10743 (define_expand, define_insn): Move the shift left from the
10744 define_insn to the define_expand for lxvl and stxvl instructions.
10745 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10746 and XL_LEN_R definitions to PURE.
10748 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10750 * config/i386/i386.c (indirect_thunk_name): Declare regno
10751 as unsigned int. Compare regno with INVALID_REGNUM.
10752 (output_indirect_thunk): Ditto.
10753 (output_indirect_thunk_function): Ditto.
10754 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10755 in the call to output_indirect_thunk_function.
10757 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10759 PR middle-end/83884
10760 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10761 rather than the size of inner_type to determine the stack slot size
10762 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10764 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10767 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10770 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10772 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10773 endian Linux systems to optionally enable multilibs for selecting
10774 the long double type if the user configured an explicit type.
10775 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10776 have no long double multilibs if not defined.
10777 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10778 warn if the user used -mabi={ieee,ibm}longdouble and we built
10779 multilibs for long double.
10780 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10781 appropriate multilib option.
10782 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10784 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10785 for building long double multilibs.
10786 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10788 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10790 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10793 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10795 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10798 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10801 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10804 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10806 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10807 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10809 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10811 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10812 different rtl trees depending on TARGET_64BIT.
10813 (rs6000_gen_lvx): Likewise.
10815 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10817 * config/visium/visium.md (nop): Tweak comment.
10818 (hazard_nop): Likewise.
10820 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10822 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10823 -mspeculate-indirect-jumps.
10824 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10825 for -mno-speculate-indirect-jumps.
10826 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10827 (*call_value_indirect_elfv2<mode>): Disable for
10828 -mno-speculate-indirect-jumps.
10829 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10830 (indirect_jump): Emit different RTL for
10831 -mno-speculate-indirect-jumps.
10832 (*indirect_jump<mode>): Disable for
10833 -mno-speculate-indirect-jumps.
10834 (*indirect_jump<mode>_nospec): New define_insn.
10835 (tablejump): Emit different RTL for
10836 -mno-speculate-indirect-jumps.
10837 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10838 (tablejumpsi_nospec): New define_expand.
10839 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10840 (tablejumpdi_nospec): New define_expand.
10841 (*tablejump<mode>_internal1): Disable for
10842 -mno-speculate-indirect-jumps.
10843 (*tablejump<mode>_internal1_nospec): New define_insn.
10844 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10847 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10849 * caller-save.c (insert_save): Drop unnecessary parameter. All
10852 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10853 Richard Biener <rguenth@suse.de>
10856 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10857 return early, inline manually is_gimple_sizepos. Make sure if we
10858 call gimplify_expr we don't end up with a gimple constant.
10859 * tree.c (variably_modified_type_p): Don't return true for
10860 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10861 * gimplify.h (is_gimple_sizepos): Remove.
10863 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10865 PR tree-optimization/83857
10866 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10867 vectorizable_live_operation for pure SLP statements.
10868 (vectorizable_live_operation): Handle PHIs.
10870 2018-01-16 Richard Biener <rguenther@suse.de>
10872 PR tree-optimization/83867
10873 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10874 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10876 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10879 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10880 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10881 If off is not INTEGER_CST, issue a may not be aligned warning
10882 rather than isn't aligned. Use isn%'t rather than isn't.
10883 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10885 <case MULT_EXPR>: Improve the case when bottom and one of the
10886 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10887 operand, in that case check if the other operand is multiple of
10888 bottom divided by the INTEGER_CST operand.
10890 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10893 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10894 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10895 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10896 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10897 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10898 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10899 (pa_function_arg_size): New function.
10901 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10903 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10904 in a separate statement.
10906 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10908 PR tree-optimization/83847
10909 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10910 group gathers and scatters.
10912 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10914 PR rtl-optimization/86620
10915 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10917 PR rtl-optimization/83213
10918 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10919 to last if both are JUMP_INSNs.
10921 PR tree-optimization/83843
10922 * gimple-ssa-store-merging.c
10923 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10924 store_immediate_info for bswap/nop orig_stores.
10926 2018-01-15 Andrew Waterman <andrew@sifive.com>
10928 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10930 <UDIV>: Increase cost if !TARGET_DIV.
10932 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10934 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10935 (define_attr "cr_logical_3op"): New.
10936 (cceq_ior_compare): Adjust.
10937 (cceq_ior_compare_complement): Adjust.
10938 (*cceq_rev_compare): Adjust.
10939 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10940 (is_cracked_insn): Adjust.
10941 (insn_must_be_first_in_group): Adjust.
10942 * config/rs6000/40x.md: Adjust.
10943 * config/rs6000/440.md: Adjust.
10944 * config/rs6000/476.md: Adjust.
10945 * config/rs6000/601.md: Adjust.
10946 * config/rs6000/603.md: Adjust.
10947 * config/rs6000/6xx.md: Adjust.
10948 * config/rs6000/7450.md: Adjust.
10949 * config/rs6000/7xx.md: Adjust.
10950 * config/rs6000/8540.md: Adjust.
10951 * config/rs6000/cell.md: Adjust.
10952 * config/rs6000/e300c2c3.md: Adjust.
10953 * config/rs6000/e500mc.md: Adjust.
10954 * config/rs6000/e500mc64.md: Adjust.
10955 * config/rs6000/e5500.md: Adjust.
10956 * config/rs6000/e6500.md: Adjust.
10957 * config/rs6000/mpc.md: Adjust.
10958 * config/rs6000/power4.md: Adjust.
10959 * config/rs6000/power5.md: Adjust.
10960 * config/rs6000/power6.md: Adjust.
10961 * config/rs6000/power7.md: Adjust.
10962 * config/rs6000/power8.md: Adjust.
10963 * config/rs6000/power9.md: Adjust.
10964 * config/rs6000/rs64.md: Adjust.
10965 * config/rs6000/titan.md: Adjust.
10967 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10969 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10970 ix86_indirect_branch_register logic.
10972 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10974 * config/i386/constraints.md (Bs): Update
10975 ix86_indirect_branch_register check. Don't check
10976 ix86_indirect_branch_register with GOT_memory_operand.
10978 * config/i386/predicates.md (GOT_memory_operand): Don't check
10979 ix86_indirect_branch_register here.
10980 (GOT32_symbol_operand): Likewise.
10982 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10984 * config/i386/predicates.md (constant_call_address_operand):
10985 Rewrite ix86_indirect_branch_register logic.
10986 (sibcall_insn_operand): Likewise.
10988 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10990 * config/i386/constraints.md (Bs): Replace
10991 ix86_indirect_branch_thunk_register with
10992 ix86_indirect_branch_register.
10994 * config/i386/i386.md (indirect_jump): Likewise.
10995 (tablejump): Likewise.
10996 (*sibcall_memory): Likewise.
10997 (*sibcall_value_memory): Likewise.
10998 Peepholes of indirect call and jump via memory: Likewise.
10999 * config/i386/i386.opt: Likewise.
11000 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11001 (GOT_memory_operand): Likewise.
11002 (call_insn_operand): Likewise.
11003 (sibcall_insn_operand): Likewise.
11004 (GOT32_symbol_operand): Likewise.
11006 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11008 PR middle-end/83837
11009 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
11010 type rather than type addr's type points to.
11011 (expand_omp_atomic_mutex): Likewise.
11012 (expand_omp_atomic): Likewise.
11014 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
11017 * config/i386/i386.c (output_indirect_thunk_function): Use
11018 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
11019 for __x86_return_thunk.
11021 2018-01-15 Richard Biener <rguenther@suse.de>
11023 PR middle-end/83850
11024 * expmed.c (extract_bit_field_1): Fix typo.
11026 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11029 * config/arm/iterators.md (VF): New mode iterator.
11030 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
11031 Remove integer-related logic from pattern.
11032 (neon_vabd<mode>_3): Likewise.
11034 2018-01-15 Jakub Jelinek <jakub@redhat.com>
11036 PR middle-end/82694
11037 * common.opt (fstrict-overflow): No longer an alias.
11038 (fwrapv-pointer): New option.
11039 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
11040 also for pointer types based on flag_wrapv_pointer.
11041 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
11042 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
11043 opts->x_flag_wrapv got set.
11044 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
11045 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
11046 POINTER_TYPE_OVERFLOW_UNDEFINED.
11047 * match.pd: Likewise in address comparison pattern.
11048 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
11050 2018-01-15 Richard Biener <rguenther@suse.de>
11053 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
11054 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
11055 Reset type names to their identifier if their TYPE_DECL doesn't
11056 have linkage (and thus is used for ODR and devirt).
11057 (save_debug_info_for_decl): Remove.
11058 (save_debug_info_for_type): Likewise.
11059 (add_tree_to_fld_list): Adjust.
11060 * tree-pretty-print.c (dump_generic_node): Make dumping of
11061 type names more robust.
11063 2018-01-15 Richard Biener <rguenther@suse.de>
11065 * BASE-VER: Bump to 8.0.1.
11067 2018-01-14 Martin Sebor <msebor@redhat.com>
11070 * builtins.c (check_access): Avoid warning when the no-warning bit
11073 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
11075 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
11076 * ira-color (allocno_hard_regs_compare): Likewise.
11078 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
11081 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
11082 Use .pushsection/.popsection.
11084 2018-01-14 Martin Sebor <msebor@redhat.com>
11087 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11089 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11091 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11092 entry from extra_headers.
11093 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11094 extra_headers, make the list bitwise identical to the i?86-*-* one.
11096 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11098 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11099 -mcmodel=large with -mindirect-branch=thunk,
11100 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11101 -mfunction-return=thunk-extern.
11102 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11103 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11104 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11106 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11108 * config/i386/i386.c (print_reg): Print the name of the full
11109 integer register without '%'.
11110 (ix86_print_operand): Handle 'V'.
11111 * doc/extend.texi: Document 'V' modifier.
11113 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11115 * config/i386/constraints.md (Bs): Disallow memory operand for
11116 -mindirect-branch-register.
11118 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11119 (GOT_memory_operand): Likewise.
11120 (call_insn_operand): Likewise.
11121 (sibcall_insn_operand): Likewise.
11122 (GOT32_symbol_operand): Likewise.
11123 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11124 for -mindirect-branch-register.
11125 (tablejump): Likewise.
11126 (*sibcall_memory): Likewise.
11127 (*sibcall_value_memory): Likewise.
11128 Disallow peepholes of indirect call and jump via memory for
11129 -mindirect-branch-register.
11130 (*call_pop): Replace m with Bw.
11131 (*call_value_pop): Likewise.
11132 (*sibcall_pop_memory): Replace m with Bs.
11133 * config/i386/i386.opt (mindirect-branch-register): New option.
11134 * doc/invoke.texi: Document -mindirect-branch-register option.
11136 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11138 * config/i386/i386-protos.h (ix86_output_function_return): New.
11139 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11140 set function_return_type.
11141 (indirect_thunk_name): Add ret_p to indicate thunk for function
11143 (output_indirect_thunk_function): Pass false to
11144 indirect_thunk_name.
11145 (ix86_output_indirect_branch_via_reg): Likewise.
11146 (ix86_output_indirect_branch_via_push): Likewise.
11147 (output_indirect_thunk_function): Create alias for function
11148 return thunk if regno < 0.
11149 (ix86_output_function_return): New function.
11150 (ix86_handle_fndecl_attribute): Handle function_return.
11151 (ix86_attribute_table): Add function_return.
11152 * config/i386/i386.h (machine_function): Add
11153 function_return_type.
11154 * config/i386/i386.md (simple_return_internal): Use
11155 ix86_output_function_return.
11156 (simple_return_internal_long): Likewise.
11157 * config/i386/i386.opt (mfunction-return=): New option.
11158 (indirect_branch): Mention -mfunction-return=.
11159 * doc/extend.texi: Document function_return function attribute.
11160 * doc/invoke.texi: Document -mfunction-return= option.
11162 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11164 * config/i386/i386-opts.h (indirect_branch): New.
11165 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11166 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11167 with local indirect jump when converting indirect call and jump.
11168 (ix86_set_indirect_branch_type): New.
11169 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11170 (indirectlabelno): New.
11171 (indirect_thunk_needed): Likewise.
11172 (indirect_thunk_bnd_needed): Likewise.
11173 (indirect_thunks_used): Likewise.
11174 (indirect_thunks_bnd_used): Likewise.
11175 (INDIRECT_LABEL): Likewise.
11176 (indirect_thunk_name): Likewise.
11177 (output_indirect_thunk): Likewise.
11178 (output_indirect_thunk_function): Likewise.
11179 (ix86_output_indirect_branch_via_reg): Likewise.
11180 (ix86_output_indirect_branch_via_push): Likewise.
11181 (ix86_output_indirect_branch): Likewise.
11182 (ix86_output_indirect_jmp): Likewise.
11183 (ix86_code_end): Call output_indirect_thunk_function if needed.
11184 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11186 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11187 (ix86_attribute_table): Add indirect_branch.
11188 * config/i386/i386.h (machine_function): Add indirect_branch_type
11189 and has_local_indirect_jump.
11190 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11192 (tablejump): Likewise.
11193 (*indirect_jump): Use ix86_output_indirect_jmp.
11194 (*tablejump_1): Likewise.
11195 (simple_return_indirect_internal): Likewise.
11196 * config/i386/i386.opt (mindirect-branch=): New option.
11197 (indirect_branch): New.
11200 (thunk-inline): Likewise.
11201 (thunk-extern): Likewise.
11202 * doc/extend.texi: Document indirect_branch function attribute.
11203 * doc/invoke.texi: Document -mindirect-branch= option.
11205 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11208 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11210 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11212 * ipa-inline.c (want_inline_small_function_p): Return false if
11213 inlining has already failed with CIF_FINAL_ERROR.
11214 (update_caller_keys): Call want_inline_small_function_p before
11216 (update_callee_keys): Likewise.
11218 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11220 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11222 (rs6000_quadword_masked_address_p): Likewise.
11223 (quad_aligned_load_p): Likewise.
11224 (quad_aligned_store_p): Likewise.
11225 (const_load_sequence_p): Add comment to describe the outer-most loop.
11226 (mimic_memory_attributes_and_flags): New function.
11227 (rs6000_gen_stvx): Likewise.
11228 (replace_swapped_aligned_store): Likewise.
11229 (rs6000_gen_lvx): Likewise.
11230 (replace_swapped_aligned_load): Likewise.
11231 (replace_swapped_load_constant): Capitalize argument name in
11232 comment describing this function.
11233 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11234 and stores that access quad-word aligned addresses and replace
11235 with stvx or lvx instructions when appropriate.
11236 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11237 New function prototype.
11238 (rs6000_quadword_masked_address_p): Likewise.
11239 (rs6000_gen_lvx): Likewise.
11240 (rs6000_gen_stvx): Likewise.
11241 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11242 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11243 when memory address is aligned.
11244 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11245 this split to select lvx instruction when memory address is aligned.
11246 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11247 instruction when memory address is aligned.
11248 (*vsx_le_perm_load_v16qi): Likewise.
11249 (four unnamed splitters): Modify to select the stvx instruction
11250 when memory is aligned.
11252 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11254 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11255 which appears in the queue multiple times.
11257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11258 Alan Hayward <alan.hayward@arm.com>
11259 David Sherwood <david.sherwood@arm.com>
11261 * tree-vectorizer.h (vec_lower_bound): New structure.
11262 (_loop_vec_info): Add check_nonzero and lower_bounds.
11263 (LOOP_VINFO_CHECK_NONZERO): New macro.
11264 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11265 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11266 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11267 fields. Make seg_len the distance travelled, not including the
11269 (dr_direction_indicator): Declare.
11270 (dr_zero_step_indicator): Likewise.
11271 (dr_known_forward_stride_p): Likewise.
11272 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11274 (runtime_alias_check_p): Allow runtime alias checks with
11276 (operator ==): Compare access_size and align.
11277 (prune_runtime_alias_test_list): Rework for new distinction between
11278 the access_size and seg_len.
11279 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11281 (get_segment_min_max): New function.
11282 (create_intersect_range_checks): Use it.
11283 (dr_step_indicator): New function.
11284 (dr_direction_indicator): Likewise.
11285 (dr_zero_step_indicator): Likewise.
11286 (dr_known_forward_stride_p): Likewise.
11287 * tree-loop-distribution.c (data_ref_segment_size): Return
11288 DR_STEP * (niters - 1).
11289 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11291 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11292 (vect_preserves_scalar_order_p): New function, split out from...
11293 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11294 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11295 (vect_vfa_access_size): New function.
11296 (vect_vfa_align): Likewise.
11297 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11298 (dump_lower_bound): New function.
11299 (vect_check_lower_bound): Likewise.
11300 (vect_small_gap_p): Likewise.
11301 (vectorizable_with_step_bound_p): Likewise.
11302 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11303 depencies if the vectorization factor is 1. Convert the checks
11304 for nonzero steps into checks on the bounds of DR_STEP. Try using
11305 a bunds check for variable steps if the minimum required step is
11306 relatively small. Update calls to the dr_with_seg_len
11307 constructor and to vect_compile_time_alias.
11308 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11310 (vect_loop_versioning): Call it.
11311 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11313 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11316 Alan Hayward <alan.hayward@arm.com>
11317 David Sherwood <david.sherwood@arm.com>
11319 * doc/sourcebuild.texi (vect_scatter_store): Document.
11320 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11322 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11324 * genopinit.c (main): Add supports_vec_scatter_store and
11325 supports_vec_scatter_store_cached to target_optabs.
11326 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11327 IFN_MASK_SCATTER_STORE.
11328 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11330 * internal-fn.h (internal_store_fn_p): Declare.
11331 (internal_fn_stored_value_index): Likewise.
11332 * internal-fn.c (scatter_store_direct): New macro.
11333 (expand_scatter_store_optab_fn): New function.
11334 (direct_scatter_store_optab_supported_p): New macro.
11335 (internal_store_fn_p): New function.
11336 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11337 IFN_MASK_SCATTER_STORE.
11338 (internal_fn_mask_index): Likewise.
11339 (internal_fn_stored_value_index): New function.
11340 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11341 for scatter stores.
11342 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11343 * optabs-query.c (supports_vec_scatter_store_p): New function.
11344 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11345 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11346 true for scatter stores.
11347 (vect_gather_scatter_fn_p): Handle scatter stores too.
11348 (vect_check_gather_scatter): Consider using scatter stores if
11349 supports_vec_scatter_store_p.
11350 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11351 scatter stores too.
11352 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11353 internal_fn_stored_value_index.
11354 (check_load_store_masking): Handle scatter stores too.
11355 (vect_get_store_rhs): Make public.
11356 (vectorizable_call): Use internal_store_fn_p.
11357 (vectorizable_store): Handle scatter store internal functions.
11358 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11359 when deciding whether the end of the group has been reached.
11360 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11361 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11362 (mask_scatter_store<mode>): New insns.
11364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11365 Alan Hayward <alan.hayward@arm.com>
11366 David Sherwood <david.sherwood@arm.com>
11368 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11369 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11370 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11372 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11373 Use vect_truncate_gather_scatter_offset if we can't treat the
11374 operation as a normal gather load or scatter store.
11375 (get_group_load_store_type): Take the gather_scatter_info
11376 as argument. Try using a gather load or scatter store for
11377 single-element groups.
11378 (get_load_store_type): Update calls to get_group_load_store_type
11379 and vect_use_strided_gather_scatters_p.
11381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11382 Alan Hayward <alan.hayward@arm.com>
11383 David Sherwood <david.sherwood@arm.com>
11385 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11386 optional tree argument.
11387 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11389 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11390 but continue to use the current value as a fallback.
11391 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11392 to compare the updates.
11393 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11394 (get_load_store_type): Use it when handling a strided access.
11395 (vect_get_strided_load_store_ops): New function.
11396 (vect_get_data_ptr_increment): Likewise.
11397 (vectorizable_load): Handle strided gather loads. Always pass
11398 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11401 Alan Hayward <alan.hayward@arm.com>
11402 David Sherwood <david.sherwood@arm.com>
11404 * doc/md.texi (gather_load@var{m}): Document.
11405 (mask_gather_load@var{m}): Likewise.
11406 * genopinit.c (main): Add supports_vec_gather_load and
11407 supports_vec_gather_load_cached to target_optabs.
11408 * optabs-tree.c (init_tree_optimization_optabs): Use
11409 ggc_cleared_alloc to allocate target_optabs.
11410 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11411 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11413 * internal-fn.h (internal_load_fn_p): Declare.
11414 (internal_gather_scatter_fn_p): Likewise.
11415 (internal_fn_mask_index): Likewise.
11416 (internal_gather_scatter_fn_supported_p): Likewise.
11417 * internal-fn.c (gather_load_direct): New macro.
11418 (expand_gather_load_optab_fn): New function.
11419 (direct_gather_load_optab_supported_p): New macro.
11420 (direct_internal_fn_optab): New function.
11421 (internal_load_fn_p): Likewise.
11422 (internal_gather_scatter_fn_p): Likewise.
11423 (internal_fn_mask_index): Likewise.
11424 (internal_gather_scatter_fn_supported_p): Likewise.
11425 * optabs-query.c (supports_at_least_one_mode_p): New function.
11426 (supports_vec_gather_load_p): Likewise.
11427 * optabs-query.h (supports_vec_gather_load_p): Declare.
11428 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11429 and memory_type field.
11430 (NUM_PATTERNS): Bump to 15.
11431 * tree-vect-data-refs.c: Include internal-fn.h.
11432 (vect_gather_scatter_fn_p): New function.
11433 (vect_describe_gather_scatter_call): Likewise.
11434 (vect_check_gather_scatter): Try using internal functions for
11435 gather loads. Recognize existing calls to a gather load function.
11436 (vect_analyze_data_refs): Consider using gather loads if
11437 supports_vec_gather_load_p.
11438 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11439 (vect_get_gather_scatter_offset_type): Likewise.
11440 (vect_convert_mask_for_vectype): Likewise.
11441 (vect_add_conversion_to_patterm): Likewise.
11442 (vect_try_gather_scatter_pattern): Likewise.
11443 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11444 (vect_vect_recog_func_ptrs): Add it.
11445 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11446 internal_fn_mask_index and internal_gather_scatter_fn_p.
11447 (check_load_store_masking): Take the gather_scatter_info as an
11448 argument and handle gather loads.
11449 (vect_get_gather_scatter_ops): New function.
11450 (vectorizable_call): Check internal_load_fn_p.
11451 (vectorizable_load): Likewise. Handle gather load internal
11453 (vectorizable_store): Update call to check_load_store_masking.
11454 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11455 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11456 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11457 (aarch64_gather_scale_operand_d): New predicates.
11458 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11459 (mask_gather_load<mode>): New insns.
11461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11462 Alan Hayward <alan.hayward@arm.com>
11463 David Sherwood <david.sherwood@arm.com>
11465 * optabs.def (fold_left_plus_optab): New optab.
11466 * doc/md.texi (fold_left_plus_@var{m}): Document.
11467 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11468 * internal-fn.c (fold_left_direct): Define.
11469 (expand_fold_left_optab_fn): Likewise.
11470 (direct_fold_left_optab_supported_p): Likewise.
11471 * fold-const-call.c (fold_const_fold_left): New function.
11472 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11473 * tree-parloops.c (valid_reduction_p): New function.
11474 (gather_scalar_reductions): Use it.
11475 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11476 (vect_finish_replace_stmt): Declare.
11477 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11478 (needs_fold_left_reduction_p): New function, split out from...
11479 (vect_is_simple_reduction): ...here. Accept reductions that
11480 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11481 (vect_force_simple_reduction): Also store the reduction type in
11482 the assignment's STMT_VINFO_REDUC_TYPE.
11483 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11484 (merge_with_identity): New function.
11485 (vect_expand_fold_left): Likewise.
11486 (vectorize_fold_left_reduction): Likewise.
11487 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11488 scalar phi in place for it. Check for target support and reject
11489 cases that would reassociate the operation. Defer the transform
11490 phase to vectorize_fold_left_reduction.
11491 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11492 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11493 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11497 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11498 call to ifc_temp_var.
11500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11501 Alan Hayward <alan.hayward@arm.com>
11502 David Sherwood <david.sherwood@arm.com>
11504 * target.def (legitimize_address_displacement): Take the original
11505 offset as a poly_int.
11506 * targhooks.h (default_legitimize_address_displacement): Update
11508 * targhooks.c (default_legitimize_address_displacement): Likewise.
11509 * doc/tm.texi: Regenerate.
11510 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11511 as an argument, moving assert of ad->disp == ad->disp_term to...
11512 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11513 Try calling targetm.legitimize_address_displacement before expanding
11514 the address rather than afterwards, and adjust for the new interface.
11515 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11516 Match the new hook interface. Handle SVE addresses.
11517 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11518 new hook interface.
11520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11522 * Makefile.in (OBJS): Add early-remat.o.
11523 * target.def (select_early_remat_modes): New hook.
11524 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11525 * doc/tm.texi: Regenerate.
11526 * targhooks.h (default_select_early_remat_modes): Declare.
11527 * targhooks.c (default_select_early_remat_modes): New function.
11528 * timevar.def (TV_EARLY_REMAT): New timevar.
11529 * passes.def (pass_early_remat): New pass.
11530 * tree-pass.h (make_pass_early_remat): Declare.
11531 * early-remat.c: New file.
11532 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11534 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11536 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11537 Alan Hayward <alan.hayward@arm.com>
11538 David Sherwood <david.sherwood@arm.com>
11540 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11541 vfm1 with a bound_epilog parameter.
11542 (vect_do_peeling): Update calls accordingly, and move the prologue
11543 call earlier in the function. Treat the base bound_epilog as 0 for
11544 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11545 this base when peeling for gaps.
11546 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11547 with fully-masked loops.
11548 (vect_estimate_min_profitable_iters): Handle the single peeled
11549 iteration in that case.
11551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11552 Alan Hayward <alan.hayward@arm.com>
11553 David Sherwood <david.sherwood@arm.com>
11555 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11556 single-element interleaving even if the size is not a power of 2.
11557 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11558 accesses for single-element interleaving if the group size is
11561 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11562 Alan Hayward <alan.hayward@arm.com>
11563 David Sherwood <david.sherwood@arm.com>
11565 * doc/md.texi (fold_extract_last_@var{m}): Document.
11566 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11567 * optabs.def (fold_extract_last_optab): New optab.
11568 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11569 * internal-fn.c (fold_extract_direct): New macro.
11570 (expand_fold_extract_optab_fn): Likewise.
11571 (direct_fold_extract_optab_supported_p): Likewise.
11572 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11573 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11574 EXTRACT_LAST_REDUCTION.
11575 (get_initial_def_for_reduction): Do not create an initial vector
11576 for EXTRACT_LAST_REDUCTION reductions.
11577 (vectorizable_reduction): Leave the scalar phi in place for
11578 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11579 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11580 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11581 transform phase to vectorizable_condition.
11582 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11584 (vect_finish_stmt_generation): ...here.
11585 (vect_finish_replace_stmt): New function.
11586 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11587 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11589 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11591 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11592 Alan Hayward <alan.hayward@arm.com>
11593 David Sherwood <david.sherwood@arm.com>
11595 * doc/md.texi (extract_last_@var{m}): Document.
11596 * optabs.def (extract_last_optab): New optab.
11597 * internal-fn.def (EXTRACT_LAST): New internal function.
11598 * internal-fn.c (cond_unary_direct): New macro.
11599 (expand_cond_unary_optab_fn): Likewise.
11600 (direct_cond_unary_optab_supported_p): Likewise.
11601 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11602 loops using EXTRACT_LAST.
11603 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11604 (extract_last_<mode>): ...this optab.
11605 (vec_extract<mode><Vel>): Update accordingly.
11607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11608 Alan Hayward <alan.hayward@arm.com>
11609 David Sherwood <david.sherwood@arm.com>
11611 * target.def (empty_mask_is_expensive): New hook.
11612 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11613 * doc/tm.texi: Regenerate.
11614 * targhooks.h (default_empty_mask_is_expensive): Declare.
11615 * targhooks.c (default_empty_mask_is_expensive): New function.
11616 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11617 if the target says that empty masks are expensive.
11618 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11620 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11623 Alan Hayward <alan.hayward@arm.com>
11624 David Sherwood <david.sherwood@arm.com>
11626 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11627 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11628 (vect_use_loop_mask_for_alignment_p): New function.
11629 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11630 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11631 niters_skip argument. Make sure that the first niters_skip elements
11632 of the first iteration are inactive.
11633 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11634 Update call to vect_set_loop_masks_directly.
11635 (get_misalign_in_elems): New function, split out from...
11636 (vect_gen_prolog_loop_niters): ...here.
11637 (vect_update_init_of_dr): Take a code argument that specifies whether
11638 the adjustment should be added or subtracted.
11639 (vect_update_init_of_drs): Likewise.
11640 (vect_prepare_for_masked_peels): New function.
11641 (vect_do_peeling): Skip prologue peeling if we're using a mask
11642 instead. Update call to vect_update_inits_of_drs.
11643 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11645 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11646 alignment. Do not include the number of peeled iterations in
11647 the minimum threshold in that case.
11648 (vectorizable_induction): Adjust the start value down by
11649 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11650 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11651 Take the number of skipped iterations into account when calculating
11653 * tree-vect-stmts.c (vect_gen_while_not): New function.
11655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11656 Alan Hayward <alan.hayward@arm.com>
11657 David Sherwood <david.sherwood@arm.com>
11659 * doc/sourcebuild.texi (vect_fully_masked): Document.
11660 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11661 default value to 0.
11662 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11664 (vect_analyze_loop_2): ...here. Don't check the vectorization
11665 factor against the number of loop iterations if the loop is
11668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11669 Alan Hayward <alan.hayward@arm.com>
11670 David Sherwood <david.sherwood@arm.com>
11672 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11673 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11674 (dump_groups): Update accordingly.
11675 (iv_use::mem_type): New member variable.
11676 (address_p): New function.
11677 (record_use): Add a mem_type argument and initialize the new
11679 (record_group_use): Add a mem_type argument. Use address_p.
11680 Remove obsolete null checks of base_object. Update call to record_use.
11681 (find_interesting_uses_op): Update call to record_group_use.
11682 (find_interesting_uses_cond): Likewise.
11683 (find_interesting_uses_address): Likewise.
11684 (get_mem_type_for_internal_fn): New function.
11685 (find_address_like_use): Likewise.
11686 (find_interesting_uses_stmt): Try find_address_like_use before
11687 calling find_interesting_uses_op.
11688 (addr_offset_valid_p): Use the iv mem_type field as the type
11689 of the addressed memory.
11690 (add_autoinc_candidates): Likewise.
11691 (get_address_cost): Likewise.
11692 (split_small_address_groups_p): Use address_p.
11693 (split_address_groups): Likewise.
11694 (add_iv_candidate_for_use): Likewise.
11695 (autoinc_possible_for_pair): Likewise.
11696 (rewrite_groups): Likewise.
11697 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11698 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11699 (get_alias_ptr_type_for_ptr_address): New function.
11700 (rewrite_use_address): Rewrite address uses in calls that were
11701 identified by find_address_like_use.
11703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11704 Alan Hayward <alan.hayward@arm.com>
11705 David Sherwood <david.sherwood@arm.com>
11707 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11709 * gimple-expr.h (is_gimple_addressable: Likewise.
11710 * gimple-expr.c (is_gimple_address): Likewise.
11711 * internal-fn.c (expand_call_mem_ref): New function.
11712 (expand_mask_load_optab_fn): Use it.
11713 (expand_mask_store_optab_fn): Likewise.
11715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11716 Alan Hayward <alan.hayward@arm.com>
11717 David Sherwood <david.sherwood@arm.com>
11719 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11720 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11721 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11722 (cond_umax@var{mode}): Document.
11723 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11724 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11725 (cond_umin_optab, cond_umax_optab): New optabs.
11726 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11727 (COND_IOR, COND_XOR): New internal functions.
11728 * internal-fn.h (get_conditional_internal_fn): Declare.
11729 * internal-fn.c (cond_binary_direct): New macro.
11730 (expand_cond_binary_optab_fn): Likewise.
11731 (direct_cond_binary_optab_supported_p): Likewise.
11732 (get_conditional_internal_fn): New function.
11733 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11734 Cope with reduction statements that are vectorized as calls rather
11736 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11737 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11738 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11739 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11740 (UNSPEC_COND_EOR): New unspecs.
11741 (optab): Add mappings for them.
11742 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11743 (sve_int_op, sve_fp_op): New int attributes.
11745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11746 Alan Hayward <alan.hayward@arm.com>
11747 David Sherwood <david.sherwood@arm.com>
11749 * optabs.def (while_ult_optab): New optab.
11750 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11751 * internal-fn.def (WHILE_ULT): New internal function.
11752 * internal-fn.h (direct_internal_fn_supported_p): New override
11753 that takes two types as argument.
11754 * internal-fn.c (while_direct): New macro.
11755 (expand_while_optab_fn): New function.
11756 (convert_optab_supported_p): Likewise.
11757 (direct_while_optab_supported_p): New macro.
11758 * wide-int.h (wi::udiv_ceil): New function.
11759 * tree-vectorizer.h (rgroup_masks): New structure.
11760 (vec_loop_masks): New typedef.
11761 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11762 and fully_masked_p.
11763 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11764 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11765 (vect_max_vf): New function.
11766 (slpeel_make_loop_iterate_ntimes): Delete.
11767 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11768 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11769 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11770 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11771 internal-fn.h, stor-layout.h and optabs-query.h.
11772 (vect_set_loop_mask): New function.
11773 (add_preheader_seq): Likewise.
11774 (add_header_seq): Likewise.
11775 (interleave_supported_p): Likewise.
11776 (vect_maybe_permute_loop_masks): Likewise.
11777 (vect_set_loop_masks_directly): Likewise.
11778 (vect_set_loop_condition_masked): Likewise.
11779 (vect_set_loop_condition_unmasked): New function, split out from
11780 slpeel_make_loop_iterate_ntimes.
11781 (slpeel_make_loop_iterate_ntimes): Rename to..
11782 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11783 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11784 (vect_do_peeling): Update call accordingly.
11785 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11787 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11788 mask_compare_type, can_fully_mask_p and fully_masked_p.
11789 (release_vec_loop_masks): New function.
11790 (_loop_vec_info): Use it to free the loop masks.
11791 (can_produce_all_loop_masks_p): New function.
11792 (vect_get_max_nscalars_per_iter): Likewise.
11793 (vect_verify_full_masking): Likewise.
11794 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11795 retries, and free the mask rgroups before retrying. Check loop-wide
11796 reasons for disallowing fully-masked loops. Make the final decision
11797 about whether use a fully-masked loop or not.
11798 (vect_estimate_min_profitable_iters): Do not assume that peeling
11799 for the number of iterations will be needed for fully-masked loops.
11800 (vectorizable_reduction): Disable fully-masked loops.
11801 (vectorizable_live_operation): Likewise.
11802 (vect_halve_mask_nunits): New function.
11803 (vect_double_mask_nunits): Likewise.
11804 (vect_record_loop_mask): Likewise.
11805 (vect_get_loop_mask): Likewise.
11806 (vect_transform_loop): Handle the case in which the final loop
11807 iteration might handle a partial vector. Call vect_set_loop_condition
11808 instead of slpeel_make_loop_iterate_ntimes.
11809 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11810 (check_load_store_masking): New function.
11811 (prepare_load_store_mask): Likewise.
11812 (vectorizable_store): Handle fully-masked loops.
11813 (vectorizable_load): Likewise.
11814 (supportable_widening_operation): Use vect_halve_mask_nunits for
11816 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11817 (vect_gen_while): New function.
11818 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11819 (aarch64_uqdec<mode>): New insn.
11821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11822 Alan Hayward <alan.hayward@arm.com>
11823 David Sherwood <david.sherwood@arm.com>
11825 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11826 (reduc_xor_scal_optab): New optabs.
11827 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11828 (reduc_xor_scal_@var{m}): Document.
11829 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11830 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11831 internal functions.
11832 * fold-const-call.c (fold_const_call): Handle them.
11833 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11834 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11835 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11836 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11837 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11838 (UNSPEC_XORV): New unspecs.
11839 (optab): Add entries for them.
11840 (BITWISEV): New int iterator.
11841 (bit_reduc_op): New int attributes.
11843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11844 Alan Hayward <alan.hayward@arm.com>
11845 David Sherwood <david.sherwood@arm.com>
11847 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11848 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11849 * optabs.def (vec_shl_insert_optab): New optab.
11850 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11851 (duplicate_and_interleave): Likewise.
11852 * tree-vect-loop.c: Include internal-fn.h.
11853 (neutral_op_for_slp_reduction): New function, split out from
11854 get_initial_defs_for_reduction.
11855 (get_initial_def_for_reduction): Handle option 2 for variable-length
11856 vectors by loading the neutral value into a vector and then shifting
11857 the initial value into element 0.
11858 (get_initial_defs_for_reduction): Replace the code argument with
11859 the neutral value calculated by neutral_op_for_slp_reduction.
11860 Use gimple_build_vector for constant-length vectors.
11861 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11862 but the first group_size elements have a neutral value.
11863 Use duplicate_and_interleave otherwise.
11864 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11865 Update call to get_initial_defs_for_reduction. Handle SLP
11866 reductions for variable-length vectors by creating one vector
11867 result for each scalar result, with the elements associated
11868 with other scalar results stubbed out with the neutral value.
11869 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11870 Require IFN_VEC_SHL_INSERT for double reductions on
11871 variable-length vectors, or SLP reductions that have
11872 a neutral value. Require can_duplicate_and_interleave_p
11873 support for variable-length unchained SLP reductions if there
11874 is no neutral value, such as for MIN/MAX reductions. Also require
11875 the number of vector elements to be a multiple of the number of
11876 SLP statements when doing variable-length unchained SLP reductions.
11877 Update call to vect_create_epilog_for_reduction.
11878 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11879 and remove initial values.
11880 (duplicate_and_interleave): Make public.
11881 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11882 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11885 Alan Hayward <alan.hayward@arm.com>
11886 David Sherwood <david.sherwood@arm.com>
11888 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11889 (can_duplicate_and_interleave_p): New function.
11890 (vect_get_and_check_slp_defs): Take the vector of statements
11891 rather than just the current one. Remove excess parentheses.
11892 Restriction rejectinon of vect_constant_def and vect_external_def
11893 for variable-length vectors to boolean types, or types for which
11894 can_duplicate_and_interleave_p is false.
11895 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11896 (duplicate_and_interleave): New function.
11897 (vect_get_constant_vectors): Use gimple_build_vector for
11898 constant-length vectors and suitable variable-length constant
11899 vectors. Use duplicate_and_interleave for other variable-length
11900 vectors. Don't defer the update when inserting new statements.
11902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11903 Alan Hayward <alan.hayward@arm.com>
11904 David Sherwood <david.sherwood@arm.com>
11906 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11907 min_profitable_iters doesn't go negative.
11909 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11910 Alan Hayward <alan.hayward@arm.com>
11911 David Sherwood <david.sherwood@arm.com>
11913 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11914 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11915 * optabs.def (vec_mask_load_lanes_optab): New optab.
11916 (vec_mask_store_lanes_optab): Likewise.
11917 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11918 (MASK_STORE_LANES): Likewise.
11919 * internal-fn.c (mask_load_lanes_direct): New macro.
11920 (mask_store_lanes_direct): Likewise.
11921 (expand_mask_load_optab_fn): Handle masked operations.
11922 (expand_mask_load_lanes_optab_fn): New macro.
11923 (expand_mask_store_optab_fn): Handle masked operations.
11924 (expand_mask_store_lanes_optab_fn): New macro.
11925 (direct_mask_load_lanes_optab_supported_p): Likewise.
11926 (direct_mask_store_lanes_optab_supported_p): Likewise.
11927 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11929 (vect_load_lanes_supported): Likewise.
11930 * tree-vect-data-refs.c (strip_conversion): New function.
11931 (can_group_stmts_p): Likewise.
11932 (vect_analyze_data_ref_accesses): Use it instead of checking
11933 for a pair of assignments.
11934 (vect_store_lanes_supported): Take a masked_p parameter.
11935 (vect_load_lanes_supported): Likewise.
11936 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11937 vect_store_lanes_supported and vect_load_lanes_supported.
11938 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11939 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11940 parameter. Don't allow gaps for masked accesses.
11941 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11942 and vect_load_lanes_supported.
11943 (get_load_store_type): Take a masked_p parameter and update
11944 call to get_group_load_store_type.
11945 (vectorizable_store): Update call to get_load_store_type.
11946 Handle IFN_MASK_STORE_LANES.
11947 (vectorizable_load): Update call to get_load_store_type.
11948 Handle IFN_MASK_LOAD_LANES.
11950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11951 Alan Hayward <alan.hayward@arm.com>
11952 David Sherwood <david.sherwood@arm.com>
11954 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11956 * config/aarch64/aarch64-protos.h
11957 (aarch64_sve_struct_memory_operand_p): Declare.
11958 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11959 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11960 (VPRED, vpred): Handle SVE structure modes.
11961 * config/aarch64/constraints.md (Utx): New constraint.
11962 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11963 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11964 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11965 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11966 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11967 structure modes. Split into pieces after RA.
11968 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11969 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11971 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11972 SVE structure modes.
11973 (aarch64_classify_address): Likewise.
11974 (sizetochar): Move earlier in file.
11975 (aarch64_print_operand): Handle SVE register lists.
11976 (aarch64_array_mode): New function.
11977 (aarch64_sve_struct_memory_operand_p): Likewise.
11978 (TARGET_ARRAY_MODE): Redefine.
11980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11981 Alan Hayward <alan.hayward@arm.com>
11982 David Sherwood <david.sherwood@arm.com>
11984 * target.def (array_mode): New target hook.
11985 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11986 * doc/tm.texi: Regenerate.
11987 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11988 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11989 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11990 targetm.array_mode.
11991 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11994 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11995 Alan Hayward <alan.hayward@arm.com>
11996 David Sherwood <david.sherwood@arm.com>
11998 * fold-const.c (fold_binary_loc): Check the argument types
11999 rather than the result type when testing for a vector operation.
12001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12003 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
12004 * doc/tm.texi: Regenerate.
12006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12007 Alan Hayward <alan.hayward@arm.com>
12008 David Sherwood <david.sherwood@arm.com>
12010 * doc/invoke.texi (-msve-vector-bits=): Document new option.
12011 (sve): Document new AArch64 extension.
12012 * doc/md.texi (w): Extend the description of the AArch64
12013 constraint to include SVE vectors.
12014 (Upl, Upa): Document new AArch64 predicate constraints.
12015 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
12017 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
12018 (msve-vector-bits=): New option.
12019 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
12020 SVE when these are disabled.
12021 (sve): New extension.
12022 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
12023 modes. Adjust their number of units based on aarch64_sve_vg.
12024 (MAX_BITSIZE_MODE_ANY_MODE): Define.
12025 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
12026 aarch64_addr_query_type.
12027 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
12028 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
12029 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
12030 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
12031 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
12032 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
12033 (aarch64_simd_imm_zero_p): Delete.
12034 (aarch64_check_zero_based_sve_index_immediate): Declare.
12035 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12036 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12037 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12038 (aarch64_sve_float_mul_immediate_p): Likewise.
12039 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12040 rather than an rtx.
12041 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
12042 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
12043 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
12044 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
12045 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
12046 (aarch64_regmode_natural_size): Likewise.
12047 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
12048 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
12050 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
12051 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
12052 for VG and the SVE predicate registers.
12053 (V_ALIASES): Add a "z"-prefixed alias.
12054 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
12055 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
12056 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
12057 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
12058 (REG_CLASS_NAMES): Add entries for them.
12059 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
12060 and the predicate registers.
12061 (aarch64_sve_vg): Declare.
12062 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
12063 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
12064 (REGMODE_NATURAL_SIZE): Define.
12065 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
12067 * config/aarch64/aarch64.c: Include cfgrtl.h.
12068 (simd_immediate_info): Add a constructor for series vectors,
12069 and an associated step field.
12070 (aarch64_sve_vg): New variable.
12071 (aarch64_dbx_register_number): Handle VG and the predicate registers.
12072 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
12073 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
12074 (VEC_ANY_DATA, VEC_STRUCT): New constants.
12075 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
12076 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
12077 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
12078 (aarch64_get_mask_mode): New functions.
12079 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
12080 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
12081 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
12082 predicate modes and predicate registers. Explicitly restrict
12083 GPRs to modes of 16 bytes or smaller. Only allow FP registers
12084 to store a vector mode if it is recognized by
12085 aarch64_classify_vector_mode.
12086 (aarch64_regmode_natural_size): New function.
12087 (aarch64_hard_regno_caller_save_mode): Return the original mode
12089 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12090 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12091 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12092 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12094 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12095 does not overlap dest if the function is frame-related. Handle
12097 (aarch64_split_add_offset): New function.
12098 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12099 them aarch64_add_offset.
12100 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12101 and update call to aarch64_sub_sp.
12102 (aarch64_add_cfa_expression): New function.
12103 (aarch64_expand_prologue): Pass extra temporary registers to the
12104 functions above. Handle the case in which we need to emit new
12105 DW_CFA_expressions for registers that were originally saved
12106 relative to the stack pointer, but now have to be expressed
12107 relative to the frame pointer.
12108 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12110 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12111 IP0 and IP1 values for SVE frames.
12112 (aarch64_expand_vec_series): New function.
12113 (aarch64_expand_sve_widened_duplicate): Likewise.
12114 (aarch64_expand_sve_const_vector): Likewise.
12115 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12116 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12117 into the register, rather than emitting a SET directly.
12118 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12119 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12120 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12121 (offset_9bit_signed_scaled_p): New functions.
12122 (aarch64_replicate_bitmask_imm): New function.
12123 (aarch64_bitmask_imm): Use it.
12124 (aarch64_cannot_force_const_mem): Reject expressions involving
12125 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12126 (aarch64_classify_index): Handle SVE indices, by requiring
12127 a plain register index with a scale that matches the element size.
12128 (aarch64_classify_address): Handle SVE addresses. Assert that
12129 the mode of the address is VOIDmode or an integer mode.
12130 Update call to aarch64_classify_symbol.
12131 (aarch64_classify_symbolic_expression): Update call to
12132 aarch64_classify_symbol.
12133 (aarch64_const_vec_all_in_range_p): New function.
12134 (aarch64_print_vector_float_operand): Likewise.
12135 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12136 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12137 and the FP immediates 1.0 and 0.5.
12138 (aarch64_print_address_internal): Handle SVE addresses.
12139 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12140 (aarch64_regno_regclass): Handle predicate registers.
12141 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12143 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12144 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12145 (aarch64_convert_sve_vector_bits): New function.
12146 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12147 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12148 rather than an rtx.
12149 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12150 Handle SVE vector and predicate modes. Accept VL-based constants
12151 that need only one temporary register, and VL offsets that require
12152 no temporary registers.
12153 (aarch64_conditional_register_usage): Mark the predicate registers
12154 as fixed if SVE isn't available.
12155 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12156 Return true for SVE vector and predicate modes.
12157 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12158 rather than an unsigned int. Handle SVE modes.
12159 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12161 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12163 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12164 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12165 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12166 (aarch64_sve_float_mul_immediate_p): New functions.
12167 (aarch64_sve_valid_immediate): New function.
12168 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12169 Explicitly reject structure modes. Check for INDEX constants.
12170 Handle PTRUE and PFALSE constants.
12171 (aarch64_check_zero_based_sve_index_immediate): New function.
12172 (aarch64_simd_imm_zero_p): Delete.
12173 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12174 vector modes. Accept constants in the range of CNT[BHWD].
12175 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12176 ask for an Advanced SIMD mode.
12177 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12178 (aarch64_simd_vector_alignment): Handle SVE predicates.
12179 (aarch64_vectorize_preferred_vector_alignment): New function.
12180 (aarch64_simd_vector_alignment_reachable): Use it instead of
12182 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12183 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12185 (MAX_VECT_LEN): Delete.
12186 (expand_vec_perm_d): Add a vec_flags field.
12187 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12188 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12189 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12191 (aarch64_evpc_rev): Rename to...
12192 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12193 (aarch64_evpc_rev_global): New function.
12194 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12195 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12197 (aarch64_evpc_sve_tbl): New function.
12198 (aarch64_expand_vec_perm_const_1): Update after rename of
12199 aarch64_evpc_rev. Handle SVE permutes too, trying
12200 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12201 than aarch64_evpc_tbl.
12202 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12203 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12204 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12205 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12206 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12207 (aarch64_expand_sve_vcond): New functions.
12208 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12209 of aarch64_vector_mode_p.
12210 (aarch64_dwarf_poly_indeterminate_value): New function.
12211 (aarch64_compute_pressure_classes): Likewise.
12212 (aarch64_can_change_mode_class): Likewise.
12213 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12214 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12215 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12216 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12217 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12218 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12219 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12220 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12222 (Dn, Dl, Dr): Accept const as well as const_vector.
12223 (Dz): Likewise. Compare against CONST0_RTX.
12224 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12225 of "vector" where appropriate.
12226 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12227 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12228 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12229 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12230 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12231 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12232 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12233 (v_int_equiv): Extend to SVE modes.
12234 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12236 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12237 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12238 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12239 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12240 (SVE_COND_FP_CMP): New int iterators.
12241 (perm_hilo): Handle the new unpack unspecs.
12242 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12244 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12245 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12246 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12247 (aarch64_equality_operator, aarch64_constant_vector_operand)
12248 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12249 (aarch64_sve_nonimmediate_operand): Likewise.
12250 (aarch64_sve_general_operand): Likewise.
12251 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12252 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12253 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12254 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12255 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12256 (aarch64_sve_float_arith_immediate): Likewise.
12257 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12258 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12259 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12260 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12261 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12262 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12263 (aarch64_sve_float_arith_operand): Likewise.
12264 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12265 (aarch64_sve_float_mul_operand): Likewise.
12266 (aarch64_sve_vec_perm_operand): Likewise.
12267 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12268 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12269 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12270 as well as const_vector.
12271 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12272 in file. Use CONST0_RTX and CONSTM1_RTX.
12273 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12274 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12275 Use aarch64_simd_imm_zero.
12276 * config/aarch64/aarch64-sve.md: New file.
12277 * config/aarch64/aarch64.md: Include it.
12278 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12279 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12280 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12281 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12282 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12283 (sve): New attribute.
12284 (enabled): Disable instructions with the sve attribute unless
12286 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12287 aarch64_expand_mov_immediate.
12288 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12289 CNT[BHSD] immediates.
12290 (movti): Split CONST_POLY_INT moves into two halves.
12291 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12292 Split additions that need a temporary here if the destination
12293 is the stack pointer.
12294 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12295 (*add<mode>3_poly_1): New instruction.
12296 (set_clobber_cc): New expander.
12298 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12300 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12301 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12302 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12303 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12304 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12305 Change innermode from fixed_mode_size to machine_mode.
12306 (simplify_subreg): Update call accordingly. Handle a constant-sized
12307 subreg of a variable-length CONST_VECTOR.
12309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12310 Alan Hayward <alan.hayward@arm.com>
12311 David Sherwood <david.sherwood@arm.com>
12313 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12314 (add_offset_to_base): New function, split out from...
12315 (create_mem_ref): ...here. When handling a scale other than 1,
12316 check first whether the address is valid without the offset.
12317 Add it into the base if so, leaving the index and scale as-is.
12319 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12322 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12323 fold_for_warn before checking if arg2 is INTEGER_CST.
12325 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12327 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12328 (store_multiple_operation): Delete.
12329 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12330 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12331 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12332 guarded by TARGET_STRING.
12333 (rs6000_output_load_multiple): Delete.
12334 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12335 OPTION_MASK_STRING / TARGET_STRING handling.
12336 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12337 (const rs6000_opt_masks) <"string">: Change mask to 0.
12338 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12339 (MASK_STRING): Delete.
12340 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12342 (load_multiple): Delete.
12349 (store_multiple): Delete.
12356 (movmemsi_8reg): Delete.
12357 (corresponding unnamed define_insn): Delete.
12358 (movmemsi_6reg): Delete.
12359 (corresponding unnamed define_insn): Delete.
12360 (movmemsi_4reg): Delete.
12361 (corresponding unnamed define_insn): Delete.
12362 (movmemsi_2reg): Delete.
12363 (corresponding unnamed define_insn): Delete.
12364 (movmemsi_1reg): Delete.
12365 (corresponding unnamed define_insn): Delete.
12366 * config/rs6000/rs6000.opt (mno-string): New.
12367 (mstring): Replace by deprecation warning stub.
12368 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12370 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12372 * regrename.c (regrename_do_replace): If replacing the same
12373 reg multiple times, try to reuse last created gen_raw_REG.
12376 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12377 main to workaround a bug in GDB.
12379 2018-01-12 Tom de Vries <tom@codesourcery.com>
12382 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12384 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12386 PR rtl-optimization/80481
12387 * ira-color.c (get_cap_member): New function.
12388 (allocnos_conflict_by_live_ranges_p): Use it.
12389 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12390 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12392 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12395 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12396 (*saddl_se_1): Ditto.
12397 (*ssubsi_1): Ditto.
12398 (*ssubl_se_1): Ditto.
12400 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12402 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12403 rather than wi::to_widest for DR_INITs.
12404 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12405 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12406 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12408 (vect_analyze_group_access_1): Note that here.
12410 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12412 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12413 polynomial type sizes.
12415 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12417 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12418 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12419 (gimple_add_tmp_var): Likewise.
12421 2018-01-12 Martin Liska <mliska@suse.cz>
12423 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12424 (gimple_alloc_sizes): Likewise.
12425 (dump_gimple_statistics): Use PRIu64 in printf format.
12426 * gimple.h: Change uint64_t to int.
12428 2018-01-12 Martin Liska <mliska@suse.cz>
12430 * tree-core.h: Use uint64_t instead of int.
12431 * tree.c (tree_node_counts): Likewise.
12432 (tree_node_sizes): Likewise.
12433 (dump_tree_statistics): Use PRIu64 in printf format.
12435 2018-01-12 Martin Liska <mliska@suse.cz>
12437 * Makefile.in: As qsort_chk is implemented in vec.c, add
12438 vec.o to linkage of gencfn-macros.
12439 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12440 passing the info to record_node_allocation_statistics.
12441 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12443 * ggc-common.c (struct ggc_usage): Add operator== and use
12444 it in operator< and compare function.
12445 * mem-stats.h (struct mem_usage): Likewise.
12446 * vec.c (struct vec_usage): Remove operator< and compare
12447 function. Can be simply inherited.
12449 2018-01-12 Martin Jambor <mjambor@suse.cz>
12452 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12453 * tree-ssa-math-opts.c: Include domwalk.h.
12454 (convert_mult_to_fma_1): New function.
12455 (fma_transformation_info): New type.
12456 (fma_deferring_state): Likewise.
12457 (cancel_fma_deferring): New function.
12458 (result_of_phi): Likewise.
12459 (last_fma_candidate_feeds_initial_phi): Likewise.
12460 (convert_mult_to_fma): Added deferring logic, split actual
12461 transformation to convert_mult_to_fma_1.
12462 (math_opts_dom_walker): New type.
12463 (math_opts_dom_walker::after_dom_children): New method, body moved
12464 here from pass_optimize_widening_mul::execute, added deferring logic
12466 (pass_optimize_widening_mul::execute): Moved most of code to
12467 math_opts_dom_walker::after_dom_children.
12468 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12469 * config/i386/i386.c (ix86_option_override_internal): Added
12470 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12472 2018-01-12 Richard Biener <rguenther@suse.de>
12475 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12476 inline instance vars.
12478 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12481 * config/rx/rx.c (rx_is_restricted_memory_address):
12482 Handle SUBREG case.
12484 2018-01-12 Richard Biener <rguenther@suse.de>
12486 PR tree-optimization/80846
12487 * target.def (split_reduction): New target hook.
12488 * targhooks.c (default_split_reduction): New function.
12489 * targhooks.h (default_split_reduction): Declare.
12490 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12491 target requests first reduce vectors by combining low and high
12493 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12494 (get_vectype_for_scalar_type_and_size): Export.
12495 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12496 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12497 * doc/tm.texi: Regenerate.
12498 * config/i386/i386.c (ix86_split_reduction): Implement
12499 TARGET_VECTORIZE_SPLIT_REDUCTION.
12501 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12504 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12505 in PIC mode except for TARGET_VXWORKS_RTP.
12506 * config/sparc/sparc.c: Include cfgrtl.h.
12507 (TARGET_INIT_PIC_REG): Define.
12508 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12509 (sparc_pic_register_p): New predicate.
12510 (sparc_legitimate_address_p): Use it.
12511 (sparc_legitimize_pic_address): Likewise.
12512 (sparc_delegitimize_address): Likewise.
12513 (sparc_mode_dependent_address_p): Likewise.
12514 (gen_load_pcrel_sym): Remove 4th parameter.
12515 (load_got_register): Adjust call to above. Remove obsolete stuff.
12516 (sparc_expand_prologue): Do not call load_got_register here.
12517 (sparc_flat_expand_prologue): Likewise.
12518 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12519 (sparc_use_pseudo_pic_reg): New function.
12520 (sparc_init_pic_reg): Likewise.
12521 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12522 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12524 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12526 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12527 Add item for branch_cost.
12529 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12531 PR rtl-optimization/83565
12532 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12533 not extend the result to a larger mode for rotate operations.
12534 (num_sign_bit_copies1): Likewise.
12536 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12539 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12541 Use values-Xc.o for -pedantic.
12542 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12544 2018-01-12 Martin Liska <mliska@suse.cz>
12547 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12549 (possible_polymorphic_call_targets): Use it.
12550 (ipa_devirt): Likewise.
12552 2018-01-12 Martin Liska <mliska@suse.cz>
12554 * profile-count.h (enum profile_quality): Use 0 as invalid
12555 enum value of profile_quality.
12557 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12559 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12560 -mext-string options.
12562 2018-01-12 Richard Biener <rguenther@suse.de>
12564 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12565 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12566 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12568 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12570 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12572 * configure.ac (--with-long-double-format): Add support for the
12573 configuration option to change the default long double format on
12575 * config.gcc (powerpc*-linux*-*): Likewise.
12576 * configure: Regenerate.
12577 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12578 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12579 used without modification.
12581 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12583 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12584 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12585 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12586 MISC_BUILTIN_SPEC_BARRIER.
12587 (rs6000_init_builtins): Likewise.
12588 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12590 (speculation_barrier): New define_insn.
12591 * doc/extend.texi: Document __builtin_speculation_barrier.
12593 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12596 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12597 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12598 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12600 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12601 integral modes instead of "ss" and "sd".
12602 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12603 vectors with 32-bit and 64-bit elements.
12604 (vecdupssescalarmodesuffix): New mode attribute.
12605 (vec_dup<mode>): Use it.
12607 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12610 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12611 frame if argument is passed on stack.
12613 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12616 * ree.c (combine_reaching_defs): Optimize also
12617 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12618 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12620 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12622 PR middle-end/83189
12623 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12625 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12627 PR middle-end/83718
12628 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12629 after they are computed.
12631 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12633 PR tree-optimization/83695
12634 * gimple-loop-linterchange.cc
12635 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12636 reset cached scev information after interchange.
12637 (pass_linterchange::execute): Remove call to scev_reset_htab.
12639 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12641 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12642 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12643 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12644 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12645 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12646 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12647 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12648 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12649 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12650 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12651 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12652 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12653 (V_lane_reg): Likewise.
12654 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12656 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12657 (vfmal_lane_low<mode>_intrinsic,
12658 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12659 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12660 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12661 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12662 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12663 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12665 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12667 * config/arm/arm-cpus.in (fp16fml): New feature.
12668 (ALL_SIMD): Add fp16fml.
12669 (armv8.2-a): Add fp16fml as an option.
12670 (armv8.3-a): Likewise.
12671 (armv8.4-a): Add fp16fml as part of fp16.
12672 * config/arm/arm.h (TARGET_FP16FML): Define.
12673 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12675 * config/arm/arm-modes.def (V2HF): Define.
12676 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12677 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12678 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12679 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12680 vfmsl_low, vfmsl_high): New set of builtins.
12681 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12682 (vfml_op): New code attribute.
12683 (VFMLHALVES): New int iterator.
12684 (VFML, VFMLSEL): New mode attributes.
12685 (V_reg): Define mapping for V2HF.
12686 (V_hi, V_lo): New mode attributes.
12687 (VF_constraint): Likewise.
12688 (vfml_half, vfml_half_selector): New int attributes.
12689 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12691 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12692 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12694 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12695 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12696 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12697 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12699 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12700 Document new effective target and option set.
12702 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12704 * config/arm/arm-cpus.in (armv8_4): New feature.
12705 (ARMv8_4a): New fgroup.
12706 (armv8.4-a): New arch.
12707 * config/arm/arm-tables.opt: Regenerate.
12708 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12709 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12710 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12711 Add matching rules for -march=armv8.4-a and extensions.
12712 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12714 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12717 * config/rx/rx.md (BW): New mode attribute.
12718 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12720 2018-01-11 Richard Biener <rguenther@suse.de>
12722 PR tree-optimization/83435
12723 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12724 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12725 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12727 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12728 Alan Hayward <alan.hayward@arm.com>
12729 David Sherwood <david.sherwood@arm.com>
12731 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12733 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12734 (aarch64_print_address_internal): Use it to check for a zero offset.
12736 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12737 Alan Hayward <alan.hayward@arm.com>
12738 David Sherwood <david.sherwood@arm.com>
12740 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12741 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12742 Return a poly_int64 rather than a HOST_WIDE_INT.
12743 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12744 rather than a HOST_WIDE_INT.
12745 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12746 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12747 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12748 final_offset from HOST_WIDE_INT to poly_int64.
12749 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12750 to_constant when getting the number of units in an Advanced SIMD
12752 (aarch64_builtin_vectorized_function): Check for a constant number
12754 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12756 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12757 attribute instead of GET_MODE_NUNITS.
12758 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12759 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12760 GET_MODE_SIZE for fixed-size registers.
12761 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12762 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12763 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12764 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12765 (aarch64_print_operand, aarch64_print_address_internal)
12766 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12767 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12768 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12769 Handle polynomial GET_MODE_SIZE.
12770 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12771 wider than SImode without modification.
12772 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12773 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12774 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12775 passing and returning SVE modes.
12776 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12777 rather than GEN_INT.
12778 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12779 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12780 (aarch64_allocate_and_probe_stack_space): Likewise.
12781 (aarch64_layout_frame): Cope with polynomial offsets.
12782 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12783 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12784 polynomial offsets.
12785 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12786 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12787 poly_int64 rather than a HOST_WIDE_INT.
12788 (aarch64_get_separate_components, aarch64_process_components)
12789 (aarch64_expand_prologue, aarch64_expand_epilogue)
12790 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12791 (aarch64_anchor_offset): New function, split out from...
12792 (aarch64_legitimize_address): ...here.
12793 (aarch64_builtin_vectorization_cost): Handle polynomial
12794 TYPE_VECTOR_SUBPARTS.
12795 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12797 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12798 number of elements from the PARALLEL rather than the mode.
12799 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12800 rather than GET_MODE_BITSIZE.
12801 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12802 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12803 (aarch64_expand_vec_perm_const_1): Handle polynomial
12804 d->perm.length () and d->perm elements.
12805 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12806 Apply to_constant to d->perm elements.
12807 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12808 polynomial CONST_VECTOR_NUNITS.
12809 (aarch64_move_pointer): Take amount as a poly_int64 rather
12811 (aarch64_progress_pointer): Avoid temporary variable.
12812 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12813 the mode attribute instead of GET_MODE.
12815 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12816 Alan Hayward <alan.hayward@arm.com>
12817 David Sherwood <david.sherwood@arm.com>
12819 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12820 x exists before using it.
12821 (aarch64_add_constant_internal): Rename to...
12822 (aarch64_add_offset_1): ...this. Replace regnum with separate
12823 src and dest rtxes. Handle the case in which they're different,
12824 including when the offset is zero. Replace scratchreg with an rtx.
12825 Use 2 additions if there is no spare register into which we can
12826 move a 16-bit constant.
12827 (aarch64_add_constant): Delete.
12828 (aarch64_add_offset): Replace reg with separate src and dest
12829 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12830 Use aarch64_add_offset_1.
12831 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12832 an rtx rather than an int. Take the delta as a poly_int64
12833 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12834 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12835 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12836 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12837 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12838 and aarch64_add_sp.
12839 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12840 aarch64_add_constant.
12842 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12844 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12845 Use scalar_float_mode.
12847 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12849 * config/aarch64/aarch64-simd.md
12850 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12851 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12852 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12853 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12854 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12855 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12856 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12857 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12858 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12859 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12861 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12864 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12865 targ_options->x_arm_arch_string is non NULL.
12867 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12869 * config/aarch64/aarch64.h
12870 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12872 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12875 * expmed.c (emit_store_flag_force): Swap if const op0
12876 and change VOIDmode to mode of op0.
12878 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12880 PR rtl-optimization/83761
12881 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12882 than bytes to mode_for_size.
12884 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12886 PR middle-end/83189
12887 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12888 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12891 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12893 PR middle-end/83575
12894 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12895 when in layout mode.
12896 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12897 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12900 2018-01-10 Michael Collison <michael.collison@arm.com>
12902 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12903 * config/aarch64/aarch64-option-extension.def: Add
12904 AARCH64_OPT_EXTENSION of 'fp16fml'.
12905 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12906 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12907 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12908 * config/aarch64/constraints.md (Ui7): New constraint.
12909 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12910 (VFMLA_SEL_W): Ditto.
12913 (VFMLA16_LOW): New int iterator.
12914 (VFMLA16_HIGH): Ditto.
12915 (UNSPEC_FMLAL): New unspec.
12916 (UNSPEC_FMLSL): Ditto.
12917 (UNSPEC_FMLAL2): Ditto.
12918 (UNSPEC_FMLSL2): Ditto.
12919 (f16mac): New code attribute.
12920 * config/aarch64/aarch64-simd-builtins.def
12921 (aarch64_fmlal_lowv2sf): Ditto.
12922 (aarch64_fmlsl_lowv2sf): Ditto.
12923 (aarch64_fmlalq_lowv4sf): Ditto.
12924 (aarch64_fmlslq_lowv4sf): Ditto.
12925 (aarch64_fmlal_highv2sf): Ditto.
12926 (aarch64_fmlsl_highv2sf): Ditto.
12927 (aarch64_fmlalq_highv4sf): Ditto.
12928 (aarch64_fmlslq_highv4sf): Ditto.
12929 (aarch64_fmlal_lane_lowv2sf): Ditto.
12930 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12931 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12932 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12933 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12934 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12935 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12936 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12937 (aarch64_fmlal_lane_highv2sf): Ditto.
12938 (aarch64_fmlsl_lane_highv2sf): Ditto.
12939 (aarch64_fmlal_laneq_highv2sf): Ditto.
12940 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12941 (aarch64_fmlalq_lane_highv4sf): Ditto.
12942 (aarch64_fmlsl_lane_highv4sf): Ditto.
12943 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12944 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12945 * config/aarch64/aarch64-simd.md:
12946 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12947 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12948 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12949 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12950 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12951 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12952 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12953 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12954 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12955 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12956 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12957 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12958 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12959 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12960 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12961 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12962 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12963 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12964 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12965 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12966 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12967 (vfmlsl_low_u32): Ditto.
12968 (vfmlalq_low_u32): Ditto.
12969 (vfmlslq_low_u32): Ditto.
12970 (vfmlal_high_u32): Ditto.
12971 (vfmlsl_high_u32): Ditto.
12972 (vfmlalq_high_u32): Ditto.
12973 (vfmlslq_high_u32): Ditto.
12974 (vfmlal_lane_low_u32): Ditto.
12975 (vfmlsl_lane_low_u32): Ditto.
12976 (vfmlal_laneq_low_u32): Ditto.
12977 (vfmlsl_laneq_low_u32): Ditto.
12978 (vfmlalq_lane_low_u32): Ditto.
12979 (vfmlslq_lane_low_u32): Ditto.
12980 (vfmlalq_laneq_low_u32): Ditto.
12981 (vfmlslq_laneq_low_u32): Ditto.
12982 (vfmlal_lane_high_u32): Ditto.
12983 (vfmlsl_lane_high_u32): Ditto.
12984 (vfmlal_laneq_high_u32): Ditto.
12985 (vfmlsl_laneq_high_u32): Ditto.
12986 (vfmlalq_lane_high_u32): Ditto.
12987 (vfmlslq_lane_high_u32): Ditto.
12988 (vfmlalq_laneq_high_u32): Ditto.
12989 (vfmlslq_laneq_high_u32): Ditto.
12990 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12991 (AARCH64_FL_FOR_ARCH8_4): New.
12992 (AARCH64_ISA_F16FML): New ISA flag.
12993 (TARGET_F16FML): New feature flag for fp16fml.
12994 (doc/invoke.texi): Document new fp16fml option.
12996 2018-01-10 Michael Collison <michael.collison@arm.com>
12998 * config/aarch64/aarch64-builtins.c:
12999 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
13000 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13001 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
13002 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
13003 (AARCH64_ISA_SHA3): New ISA flag.
13004 (TARGET_SHA3): New feature flag for sha3.
13005 * config/aarch64/iterators.md (sha512_op): New int attribute.
13006 (CRYPTO_SHA512): New int iterator.
13007 (UNSPEC_SHA512H): New unspec.
13008 (UNSPEC_SHA512H2): Ditto.
13009 (UNSPEC_SHA512SU0): Ditto.
13010 (UNSPEC_SHA512SU1): Ditto.
13011 * config/aarch64/aarch64-simd-builtins.def
13012 (aarch64_crypto_sha512hqv2di): New builtin.
13013 (aarch64_crypto_sha512h2qv2di): Ditto.
13014 (aarch64_crypto_sha512su0qv2di): Ditto.
13015 (aarch64_crypto_sha512su1qv2di): Ditto.
13016 (aarch64_eor3qv8hi): Ditto.
13017 (aarch64_rax1qv2di): Ditto.
13018 (aarch64_xarqv2di): Ditto.
13019 (aarch64_bcaxqv8hi): Ditto.
13020 * config/aarch64/aarch64-simd.md:
13021 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
13022 (aarch64_crypto_sha512su0qv2di): Ditto.
13023 (aarch64_crypto_sha512su1qv2di): Ditto.
13024 (aarch64_eor3qv8hi): Ditto.
13025 (aarch64_rax1qv2di): Ditto.
13026 (aarch64_xarqv2di): Ditto.
13027 (aarch64_bcaxqv8hi): Ditto.
13028 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
13029 (vsha512h2q_u64): Ditto.
13030 (vsha512su0q_u64): Ditto.
13031 (vsha512su1q_u64): Ditto.
13032 (veor3q_u16): Ditto.
13033 (vrax1q_u64): Ditto.
13034 (vxarq_u64): Ditto.
13035 (vbcaxq_u16): Ditto.
13036 * config/arm/types.md (crypto_sha512): New type attribute.
13037 (crypto_sha3): Ditto.
13038 (doc/invoke.texi): Document new sha3 option.
13040 2018-01-10 Michael Collison <michael.collison@arm.com>
13042 * config/aarch64/aarch64-builtins.c:
13043 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
13044 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13045 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
13046 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
13047 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
13048 (AARCH64_ISA_SM4): New ISA flag.
13049 (TARGET_SM4): New feature flag for sm4.
13050 * config/aarch64/aarch64-simd-builtins.def
13051 (aarch64_sm3ss1qv4si): Ditto.
13052 (aarch64_sm3tt1aq4si): Ditto.
13053 (aarch64_sm3tt1bq4si): Ditto.
13054 (aarch64_sm3tt2aq4si): Ditto.
13055 (aarch64_sm3tt2bq4si): Ditto.
13056 (aarch64_sm3partw1qv4si): Ditto.
13057 (aarch64_sm3partw2qv4si): Ditto.
13058 (aarch64_sm4eqv4si): Ditto.
13059 (aarch64_sm4ekeyqv4si): Ditto.
13060 * config/aarch64/aarch64-simd.md:
13061 (aarch64_sm3ss1qv4si): Ditto.
13062 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
13063 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
13064 (aarch64_sm4eqv4si): Ditto.
13065 (aarch64_sm4ekeyqv4si): Ditto.
13066 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
13067 (sm3part_op): Ditto.
13068 (CRYPTO_SM3TT): Ditto.
13069 (CRYPTO_SM3PART): Ditto.
13070 (UNSPEC_SM3SS1): New unspec.
13071 (UNSPEC_SM3TT1A): Ditto.
13072 (UNSPEC_SM3TT1B): Ditto.
13073 (UNSPEC_SM3TT2A): Ditto.
13074 (UNSPEC_SM3TT2B): Ditto.
13075 (UNSPEC_SM3PARTW1): Ditto.
13076 (UNSPEC_SM3PARTW2): Ditto.
13077 (UNSPEC_SM4E): Ditto.
13078 (UNSPEC_SM4EKEY): Ditto.
13079 * config/aarch64/constraints.md (Ui2): New constraint.
13080 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
13081 * config/arm/types.md (crypto_sm3): New type attribute.
13082 (crypto_sm4): Ditto.
13083 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
13084 (vsm3tt1aq_u32): Ditto.
13085 (vsm3tt1bq_u32): Ditto.
13086 (vsm3tt2aq_u32): Ditto.
13087 (vsm3tt2bq_u32): Ditto.
13088 (vsm3partw1q_u32): Ditto.
13089 (vsm3partw2q_u32): Ditto.
13090 (vsm4eq_u32): Ditto.
13091 (vsm4ekeyq_u32): Ditto.
13092 (doc/invoke.texi): Document new sm4 option.
13094 2018-01-10 Michael Collison <michael.collison@arm.com>
13096 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13097 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13098 (AARCH64_FL_FOR_ARCH8_4): New.
13099 (AARCH64_FL_V8_4): New flag.
13100 (doc/invoke.texi): Document new armv8.4-a option.
13102 2018-01-10 Michael Collison <michael.collison@arm.com>
13104 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13105 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13106 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13107 * config/aarch64/aarch64-option-extension.def: Add
13108 AARCH64_OPT_EXTENSION of 'sha2'.
13109 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13110 (crypto): Disable sha2 and aes if crypto disabled.
13111 (crypto): Enable aes and sha2 if enabled.
13112 (simd): Disable sha2 and aes if simd disabled.
13113 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13115 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13116 (TARGET_SHA2): New feature flag for sha2.
13117 (TARGET_AES): New feature flag for aes.
13118 * config/aarch64/aarch64-simd.md:
13119 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13120 conditional on TARGET_AES.
13121 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13122 (aarch64_crypto_sha1hsi): Make pattern conditional
13124 (aarch64_crypto_sha1hv4si): Ditto.
13125 (aarch64_be_crypto_sha1hv4si): Ditto.
13126 (aarch64_crypto_sha1su1v4si): Ditto.
13127 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13128 (aarch64_crypto_sha1su0v4si): Ditto.
13129 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13130 (aarch64_crypto_sha256su0v4si): Ditto.
13131 (aarch64_crypto_sha256su1v4si): Ditto.
13132 (doc/invoke.texi): Document new aes and sha2 options.
13134 2018-01-10 Martin Sebor <msebor@redhat.com>
13136 PR tree-optimization/83781
13137 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13140 2018-01-11 Martin Sebor <msebor@gmail.com>
13141 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13143 PR tree-optimization/83501
13144 PR tree-optimization/81703
13146 * tree-ssa-strlen.c (get_string_cst): Rename...
13147 (get_string_len): ...to this. Handle global constants.
13148 (handle_char_store): Adjust.
13150 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13151 Jim Wilson <jimw@sifive.com>
13153 * config/riscv/riscv-protos.h (riscv_output_return): New.
13154 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13155 (riscv_attribute_table, riscv_output_return),
13156 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13157 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13158 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13160 (riscv_expand_prologue): Add early return for naked function.
13161 (riscv_expand_epilogue): Likewise.
13162 (riscv_function_ok_for_sibcall): Return false for naked function.
13163 (riscv_set_current_function): New.
13164 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13165 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13166 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13167 * doc/extend.texi (RISC-V Function Attributes): New.
13169 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13171 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13172 check for 128-bit long double before checking TCmode.
13173 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13174 128-bit long doubles before checking TFmode or TCmode.
13175 (FLOAT128_IBM_P): Likewise.
13177 2018-01-10 Martin Sebor <msebor@redhat.com>
13179 PR tree-optimization/83671
13180 * builtins.c (c_strlen): Unconditionally return zero for the empty
13182 Use -Warray-bounds for warnings.
13183 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13184 for non-constant array indices with COMPONENT_REF, arrays of
13185 arrays, and pointers to arrays.
13186 (gimple_fold_builtin_strlen): Determine and set length range for
13187 non-constant character arrays.
13189 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13191 PR middle-end/81897
13192 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13195 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13197 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13199 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13202 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13203 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13204 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13205 indexed_or_indirect_operand predicate.
13206 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13207 (*vsx_le_perm_load_v8hi): Likewise.
13208 (*vsx_le_perm_load_v16qi): Likewise.
13209 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13210 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13211 (*vsx_le_perm_store_v8hi): Likewise.
13212 (*vsx_le_perm_store_v16qi): Likewise.
13213 (eight unnamed splitters): Likewise.
13215 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13217 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13218 * config/rs6000/emmintrin.h: Likewise.
13219 * config/rs6000/mmintrin.h: Likewise.
13220 * config/rs6000/xmmintrin.h: Likewise.
13222 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13225 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13227 * tree.c (tree_nop_conversion): Return true for location wrapper
13229 (maybe_wrap_with_location): New function.
13230 (selftest::check_strip_nops): New function.
13231 (selftest::test_location_wrappers): New function.
13232 (selftest::tree_c_tests): Call it.
13233 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13234 (maybe_wrap_with_location): New decl.
13235 (EXPR_LOCATION_WRAPPER_P): New macro.
13236 (location_wrapper_p): New inline function.
13237 (tree_strip_any_location_wrapper): New inline function.
13239 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13242 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13243 stack_realign_offset for the largest alignment of stack slot
13245 (ix86_find_max_used_stack_alignment): New function.
13246 (ix86_finalize_stack_frame_flags): Use it. Set
13247 max_used_stack_alignment if we don't realign stack.
13248 * config/i386/i386.h (machine_function): Add
13249 max_used_stack_alignment.
13251 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13253 * config/arm/arm.opt (-mbranch-cost): New option.
13254 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13257 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13260 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13261 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13263 2018-01-10 Richard Biener <rguenther@suse.de>
13266 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13267 early out so it also covers the case where we have a non-NULL
13270 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13272 PR tree-optimization/83753
13273 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13274 for non-strided grouped accesses if the number of elements is 1.
13276 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13279 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13280 * i386.h (TARGET_USE_GATHER): Define.
13281 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13283 2018-01-10 Martin Liska <mliska@suse.cz>
13286 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13287 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13289 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13290 CLEANUP_NO_PARTITIONING is not set.
13292 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13294 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13295 for vectors, as a partial revert of r254296.
13296 * rtl.h (const_vec_p): Delete.
13297 (const_vec_duplicate_p): Don't test for vector CONSTs.
13298 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13299 * expmed.c (make_tree): Likewise.
13302 * common.md (E, F): Use CONSTANT_P instead of checking for
13304 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13305 checking for CONST_VECTOR.
13307 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13309 PR middle-end/83575
13310 * predict.c (force_edge_cold): Handle in more sane way edges
13311 with no prediction.
13313 2018-01-09 Carl Love <cel@us.ibm.com>
13315 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13317 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13318 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13319 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13320 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13321 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13322 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13323 * config/rs6000/rs6000-protos.h: Add extern defition for
13324 rs6000_generate_float2_double_code.
13325 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13327 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13328 (float2_v2df): Add define_expand.
13330 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13333 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13334 op_mode in the force_to_mode call.
13336 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13338 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13339 instead of checking each element individually.
13340 (aarch64_evpc_uzp): Likewise.
13341 (aarch64_evpc_zip): Likewise.
13342 (aarch64_evpc_ext): Likewise.
13343 (aarch64_evpc_rev): Likewise.
13344 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13345 instead of checking each element individually. Return true without
13347 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13348 whether all selected elements come from the same input, instead of
13349 checking each element individually. Remove calls to gen_rtx_REG,
13350 start_sequence and end_sequence and instead assert that no rtl is
13353 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13355 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13356 order of HIGH and CONST checks.
13358 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13360 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13361 if the destination isn't an SSA_NAME.
13363 2018-01-09 Richard Biener <rguenther@suse.de>
13365 PR tree-optimization/83668
13366 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13368 (canonicalize_loop_form): ... here, renamed from ...
13369 (canonicalize_loop_closed_ssa_form): ... this and amended to
13370 swap successor edges for loop exit blocks to make us use
13371 the RPO order we need for initial schedule generation.
13373 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13375 PR tree-optimization/64811
13376 * match.pd: When optimizing comparisons with Inf, avoid
13377 introducing or losing exceptions from comparisons with NaN.
13379 2018-01-09 Martin Liska <mliska@suse.cz>
13382 * asan.c (shadow_mem_size): Add gcc_assert.
13384 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13386 Don't save registers in main().
13389 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13390 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13391 * config/avr/avr.c (avr_set_current_function): Don't error if
13392 naked, OS_task or OS_main are specified at the same time.
13393 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13395 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13397 * common/config/avr/avr-common.c (avr_option_optimization_table):
13398 Switch on -mmain-is-OS_task for optimizing compilations.
13400 2018-01-09 Richard Biener <rguenther@suse.de>
13402 PR tree-optimization/83572
13403 * graphite.c: Include cfganal.h.
13404 (graphite_transform_loops): Connect infinite loops to exit
13405 and remove fake edges at the end.
13407 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13409 * ipa-inline.c (edge_badness): Revert accidental checkin.
13411 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13414 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13415 symbols; not inline clones.
13417 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13420 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13421 hard registers. Formatting fixes.
13423 PR preprocessor/83722
13424 * gcc.c (try_generate_repro): Pass
13425 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13426 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13429 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13430 Kito Cheng <kito.cheng@gmail.com>
13432 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13433 (riscv_leaf_function_p): Delete.
13434 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13436 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13438 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13440 (do_ifelse): New function.
13441 (do_isel): New function.
13442 (do_sub3): New function.
13443 (do_add3): New function.
13444 (do_load_mask_compare): New function.
13445 (do_overlap_load_compare): New function.
13446 (expand_compare_loop): New function.
13447 (expand_block_compare): Call expand_compare_loop() when appropriate.
13448 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13449 option description.
13450 (-mblock-compare-inline-loop-limit): New option.
13452 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13455 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13456 Reverse order of second and third operands in first alternative.
13457 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13458 of first and second elements in UNSPEC_VPERMR vector.
13459 (altivec_expand_vec_perm_le): Likewise.
13461 2018-01-08 Jeff Law <law@redhat.com>
13463 PR rtl-optimizatin/81308
13464 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13465 (process_switch): If group_case_labels makes a change, then set
13467 (pass_convert_switch::execute): If a switch is converted, then
13468 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13470 PR rtl-optimization/81308
13471 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13474 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13476 PR target/83663 - Revert r255946
13477 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13478 generation for cases where splatting a value is not useful.
13479 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13480 across a vec_duplicate and a paradoxical subreg forming a vector
13481 mode to a vec_concat.
13483 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13485 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13486 -march=armv8.3-a variants.
13487 * config/arm/t-multilib: Likewise.
13488 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13490 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13492 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13494 (cceq_ior_compare_complement): Give it a name so I can use it, and
13495 change boolean_or_operator predicate to boolean_operator so it can
13496 be used to generate a crand.
13497 (eqne): New code iterator.
13498 (bd/bd_neg): New code_attrs.
13499 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13500 a single define_insn.
13501 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13502 decrement (bdnzt/bdnzf/bdzt/bdzf).
13503 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13504 with the new names of the branch decrement patterns, and added the
13505 names of the branch decrement conditional patterns.
13507 2018-01-08 Richard Biener <rguenther@suse.de>
13509 PR tree-optimization/83563
13510 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13513 2018-01-08 Richard Biener <rguenther@suse.de>
13515 PR middle-end/83713
13516 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13518 2018-01-08 Richard Biener <rguenther@suse.de>
13520 PR tree-optimization/83685
13521 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13522 references to abnormals.
13524 2018-01-08 Richard Biener <rguenther@suse.de>
13527 * dwarf2out.c (output_indirect_strings): Handle empty
13528 skeleton_debug_str_hash.
13529 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13531 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13533 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13534 (emit_store_direct): Likewise.
13535 (arc_trampoline_adjust_address): Likewise.
13536 (arc_asm_trampoline_template): New function.
13537 (arc_initialize_trampoline): Use asm_trampoline_template.
13538 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13539 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13540 * config/arc/arc.md (flush_icache): Delete pattern.
13542 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13544 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13545 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13548 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13551 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13552 by not USED_FOR_TARGET.
13553 (make_pass_resolve_sw_modes): Likewise.
13555 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13557 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13560 2018-01-08 Richard Biener <rguenther@suse.de>
13562 PR middle-end/83580
13563 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13565 2018-01-08 Richard Biener <rguenther@suse.de>
13567 PR middle-end/83517
13568 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13570 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13572 PR middle-end/81897
13573 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13574 basic blocks with a small number of successors.
13575 (convert_control_dep_chain_into_preds): Improve handling of
13577 (dump_predicates): Split apart into...
13578 (dump_pred_chain): ...here...
13579 (dump_pred_info): ...and here.
13580 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13581 (can_chain_union_be_invalidated_p): Improve check for invalidation
13583 (uninit_uses_cannot_happen): Avoid unnecessary if
13584 convert_control_dep_chain_into_preds yielded nothing.
13586 2018-01-06 Martin Sebor <msebor@redhat.com>
13588 PR tree-optimization/83640
13589 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13590 subtracting negative offset from size.
13591 (builtin_access::overlap): Adjust offset bounds of the access to fall
13592 within the size of the object if possible.
13594 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13596 PR rtl-optimization/83699
13597 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13598 extract_bit_field_as_subreg to cases in which the extracted
13599 value is also a vector.
13601 * lra-constraints.c (process_alt_operands): Test for the equivalence
13602 substitutions when detecting a possible reload cycle.
13604 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13607 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13608 by default if flag_selective_schedling{,2}. Formatting fixes.
13610 PR rtl-optimization/83682
13611 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13612 if it has non-VECTOR_MODE element mode.
13613 (vec_duplicate_p): Likewise.
13615 PR middle-end/83694
13616 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13617 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13619 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13622 * config/i386/i386-builtin.def
13623 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13624 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13625 Require also OPTION_MASK_ISA_AVX512F in addition to
13626 OPTION_MASK_ISA_GFNI.
13627 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13628 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13629 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13630 to OPTION_MASK_ISA_GFNI.
13631 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13632 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13633 OPTION_MASK_ISA_AVX512BW.
13634 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13635 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13636 addition to OPTION_MASK_ISA_GFNI.
13637 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13638 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13639 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13640 to OPTION_MASK_ISA_GFNI.
13641 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13642 a requirement for all ISAs rather than any of them with a few
13644 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13646 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13647 bitmasks to be enabled with 3 exceptions, instead of requiring any
13648 enabled ISA with lots of exceptions.
13649 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13650 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13651 Change avx512bw in isa attribute to avx512f.
13652 * config/i386/sgxintrin.h: Add license boilerplate.
13653 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13654 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13655 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13656 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13658 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13659 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13660 temporarily sse2 rather than sse if not enabled already.
13663 * config/i386/sse.md (VI248_VLBW): Rename to ...
13664 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13665 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13666 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13667 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13668 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13669 mode iterator instead of VI248_VLBW.
13671 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13673 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13674 (record_modified): Skip clobbers; add debug output.
13675 (param_change_prob): Use sreal frequencies.
13677 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13679 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13680 punt for user-aligned variables.
13682 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13684 * tree-chrec.c (chrec_contains_symbols): Return true for
13687 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13690 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13691 of (x|y) == x for BICS pattern.
13693 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13695 PR tree-optimization/83605
13696 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13697 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13700 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13702 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13703 * config/epiphany/rtems.h: New file.
13705 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13706 Uros Bizjak <ubizjak@gmail.com>
13709 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13710 QIreg_operand instead of register_operand predicate.
13711 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13712 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13713 comments instead of -fmitigate[-_]rop.
13715 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13718 * cgraphunit.c (symbol_table::compile): Switch to text_section
13719 before calling assembly_start debug hook.
13720 * run-rtl-passes.c (run_rtl_passes): Likewise.
13723 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13725 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13726 range_int_cst_p rather than !symbolic_range_p before calling
13727 extract_range_from_multiplicative_op_1.
13729 2018-01-04 Jeff Law <law@redhat.com>
13731 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13732 redundant test in assertion.
13734 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13736 * doc/rtl.texi: Document machine_mode wrapper classes.
13738 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13740 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13741 using tree_to_uhwi.
13743 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13745 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13746 the VEC_PERM_EXPR fold to fail.
13748 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13751 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13752 to switched_sections.
13754 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13757 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13758 test for d.testing.
13760 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13763 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13764 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13766 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13769 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13770 is BLKmode and bitpos not zero or mode change is needed.
13772 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13775 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13778 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13781 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13782 instead of MULT rtx. Update all corresponding splitters.
13783 (*saddl_se): Ditto.
13784 (*ssub<modesuffix>): Ditto.
13785 (*ssubl_se): Ditto.
13786 (*cmp_sadd_di): Update split patterns.
13787 (*cmp_sadd_si): Ditto.
13788 (*cmp_sadd_sidi): Ditto.
13789 (*cmp_ssub_di): Ditto.
13790 (*cmp_ssub_si): Ditto.
13791 (*cmp_ssub_sidi): Ditto.
13792 * config/alpha/predicates.md (const23_operand): New predicate.
13793 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13794 Look for ASHIFT, not MULT inner operand.
13795 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13797 2018-01-04 Martin Liska <mliska@suse.cz>
13799 PR gcov-profile/83669
13800 * gcov.c (output_intermediate_file): Add version to intermediate
13802 * doc/gcov.texi: Document new field 'version' in intermediate
13803 file format. Fix location of '-k' option of gcov command.
13805 2018-01-04 Martin Liska <mliska@suse.cz>
13808 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13810 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13812 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13814 2018-01-03 Martin Sebor <msebor@redhat.com>
13816 PR tree-optimization/83655
13817 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13818 checking calls with invalid arguments.
13820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13822 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13823 (vectorizable_mask_load_store): Delete.
13824 (vectorizable_call): Return false for masked loads and stores.
13825 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13826 instead of gimple_assign_rhs1.
13827 (vectorizable_load): Handle IFN_MASK_LOAD.
13828 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13832 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13834 (vectorizable_mask_load_store): ...here.
13835 (vectorizable_load): ...and here.
13837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13839 * tree-vect-stmts.c (vect_build_all_ones_mask)
13840 (vect_build_zero_merge_argument): New functions, split out from...
13841 (vectorizable_load): ...here.
13843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13845 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13847 (vectorizable_mask_load_store): ...here.
13848 (vectorizable_store): ...and here.
13850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13852 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13854 (vectorizable_mask_load_store): ...here.
13856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13858 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13859 (vect_model_store_cost): Take a vec_load_store_type instead of a
13861 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13862 (vect_model_store_cost): Take a vec_load_store_type instead of a
13864 (vectorizable_mask_load_store): Update accordingly.
13865 (vectorizable_store): Likewise.
13866 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13870 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13871 IFN_MASK_LOAD calls here rather than...
13872 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13875 Alan Hayward <alan.hayward@arm.com>
13876 David Sherwood <david.sherwood@arm.com>
13878 * expmed.c (extract_bit_field_1): For vector extracts,
13879 fall back to extract_bit_field_as_subreg if vec_extract
13882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13883 Alan Hayward <alan.hayward@arm.com>
13884 David Sherwood <david.sherwood@arm.com>
13886 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13887 they are variable or constant sized.
13888 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13889 slots for constant-sized data.
13891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13892 Alan Hayward <alan.hayward@arm.com>
13893 David Sherwood <david.sherwood@arm.com>
13895 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13896 handling COND_EXPRs with boolean comparisons, try to find a better
13897 basis for the mask type than the boolean itself.
13899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13901 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13902 is calculated and how it can be overridden.
13903 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13904 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13906 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13910 Alan Hayward <alan.hayward@arm.com>
13911 David Sherwood <david.sherwood@arm.com>
13913 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13914 Remove the mode argument.
13915 (aarch64_simd_valid_immediate): Remove the mode and inverse
13917 * config/aarch64/iterators.md (bitsize): New iterator.
13918 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13919 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13920 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13921 aarch64_simd_valid_immediate.
13922 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13923 (aarch64_reg_or_bic_imm): Likewise.
13924 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13925 with an insn_type enum and msl with a modifier_type enum.
13926 Replace element_width with a scalar_mode. Change the shift
13927 to unsigned int. Add constructors for scalar_float_mode and
13928 scalar_int_mode elements.
13929 (aarch64_vect_float_const_representable_p): Delete.
13930 (aarch64_can_const_movi_rtx_p)
13931 (aarch64_simd_scalar_immediate_valid_for_move)
13932 (aarch64_simd_make_constant): Update call to
13933 aarch64_simd_valid_immediate.
13934 (aarch64_advsimd_valid_immediate_hs): New function.
13935 (aarch64_advsimd_valid_immediate): Likewise.
13936 (aarch64_simd_valid_immediate): Remove mode and inverse
13937 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13938 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13939 and aarch64_float_const_representable_p on the result.
13940 (aarch64_output_simd_mov_immediate): Remove mode argument.
13941 Update call to aarch64_simd_valid_immediate and use of
13942 simd_immediate_info.
13943 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13951 (mode_nunits): Likewise CONST_MODE_NUNITS.
13952 * machmode.def (ADJUST_NUNITS): Document.
13953 * genmodes.c (mode_data::need_nunits_adj): New field.
13954 (blank_mode): Update accordingly.
13955 (adj_nunits): New variable.
13956 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13958 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13959 listed in adj_nunits.
13960 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13961 listed in adj_nunits. Don't emit case statements for such modes.
13962 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13963 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13964 nothing if adj_nunits is nonnull.
13965 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13966 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13967 (emit_mode_fbit): Update use of print_maybe_const_decl.
13968 (emit_move_size): Likewise. Treat the array as non-const
13970 (emit_mode_adjustments): Handle adj_nunits.
13972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13974 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13975 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13976 (VECTOR_MODES): Use it.
13977 (make_vector_modes): Take the prefix as an argument.
13979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13980 Alan Hayward <alan.hayward@arm.com>
13981 David Sherwood <david.sherwood@arm.com>
13983 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13984 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13985 for MODE_VECTOR_BOOL.
13986 * machmode.def (VECTOR_BOOL_MODE): Document.
13987 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13988 (make_vector_bool_mode): New function.
13989 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13991 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13992 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13994 * stor-layout.c (int_mode_for_mode): Likewise.
13995 * tree.c (build_vector_type_for_mode): Likewise.
13996 * varasm.c (output_constant_pool_2): Likewise.
13997 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13998 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13999 for MODE_VECTOR_BOOL.
14000 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
14001 of mode class checks.
14002 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
14003 instead of a list of mode class checks.
14004 (expand_vector_scalar_condition): Likewise.
14005 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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 * machmode.h (mode_size): Change from unsigned short to
14013 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
14014 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14015 or if measurement_type is not polynomial.
14016 (fixed_size_mode::includes_p): Check for constant-sized modes.
14017 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
14018 return a poly_uint16 rather than an unsigned short.
14019 (emit_mode_size): Change the type of mode_size from unsigned short
14020 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
14021 (emit_mode_adjustments): Cope with polynomial vector sizes.
14022 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14024 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14026 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
14027 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
14028 * caller-save.c (setup_save_areas): Likewise.
14029 (replace_reg_with_saved_mem): Likewise.
14030 * calls.c (emit_library_call_value_1): Likewise.
14031 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
14032 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
14033 (gen_lowpart_for_combine): Likewise.
14034 * convert.c (convert_to_integer_1): Likewise.
14035 * cse.c (equiv_constant, cse_insn): Likewise.
14036 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
14037 (cselib_subst_to_values): Likewise.
14038 * dce.c (word_dce_process_block): Likewise.
14039 * df-problems.c (df_word_lr_mark_ref): Likewise.
14040 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
14041 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
14042 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
14043 (rtl_for_decl_location): Likewise.
14044 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
14045 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
14046 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
14047 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
14048 (expand_expr_real_1): Likewise.
14049 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
14050 (pad_below): Likewise.
14051 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14052 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
14053 * ira.c (get_subreg_tracking_sizes): Likewise.
14054 * ira-build.c (ira_create_allocno_objects): Likewise.
14055 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
14056 (ira_sort_regnos_for_alter_reg): Likewise.
14057 * ira-costs.c (record_operand_costs): Likewise.
14058 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
14059 (resolve_simple_move): Likewise.
14060 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
14061 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
14062 (lra_constraints): Likewise.
14063 (CONST_POOL_OK_P): Reject variable-sized modes.
14064 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
14065 (add_pseudo_to_slot, lra_spill): Likewise.
14066 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14067 * optabs-query.c (get_best_extraction_insn): Likewise.
14068 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14069 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
14070 (expand_mult_highpart, valid_multiword_target_p): Likewise.
14071 * recog.c (offsettable_address_addr_space_p): Likewise.
14072 * regcprop.c (maybe_mode_change): Likewise.
14073 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
14074 * regrename.c (build_def_use): Likewise.
14075 * regstat.c (dump_reg_info): Likewise.
14076 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
14077 (find_reloads, find_reloads_subreg_address): Likewise.
14078 * reload1.c (eliminate_regs_1): Likewise.
14079 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
14080 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
14081 (simplify_binary_operation_1, simplify_subreg): Likewise.
14082 * targhooks.c (default_function_arg_padding): Likewise.
14083 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
14084 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
14085 (verify_gimple_assign_ternary): Likewise.
14086 * tree-inline.c (estimate_move_cost): Likewise.
14087 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14088 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14089 (get_address_cost_ainc): Likewise.
14090 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14091 (vect_supportable_dr_alignment): Likewise.
14092 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14093 (vectorizable_reduction): Likewise.
14094 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14095 (vectorizable_operation, vectorizable_load): Likewise.
14096 * tree.c (build_same_sized_truth_vector_type): Likewise.
14097 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14098 * var-tracking.c (emit_note_insn_var_location): Likewise.
14099 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14100 (ADDR_VEC_ALIGN): Likewise.
14102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14103 Alan Hayward <alan.hayward@arm.com>
14104 David Sherwood <david.sherwood@arm.com>
14106 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14108 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14109 or if measurement_type is polynomial.
14110 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14111 * combine.c (make_extraction): Likewise.
14112 * dse.c (find_shift_sequence): Likewise.
14113 * dwarf2out.c (mem_loc_descriptor): Likewise.
14114 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14115 (extract_bit_field, extract_low_bits): Likewise.
14116 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14117 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14118 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14119 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14120 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14121 * reload.c (find_reloads): Likewise.
14122 * reload1.c (alter_reg): Likewise.
14123 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14124 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14125 * tree-if-conv.c (predicate_mem_writes): Likewise.
14126 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14127 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14128 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14129 * valtrack.c (dead_debug_insert_temp): Likewise.
14130 * varasm.c (mergeable_constant_section): Likewise.
14131 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14134 Alan Hayward <alan.hayward@arm.com>
14135 David Sherwood <david.sherwood@arm.com>
14137 * expr.c (expand_assignment): Cope with polynomial mode sizes
14138 when assigning to a CONCAT.
14140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14141 Alan Hayward <alan.hayward@arm.com>
14142 David Sherwood <david.sherwood@arm.com>
14144 * machmode.h (mode_precision): Change from unsigned short to
14146 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14148 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14149 or if measurement_type is not polynomial.
14150 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14151 in which the mode is already known to be a scalar_int_mode.
14152 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14153 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14155 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14156 for GET_MODE_PRECISION.
14157 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14158 for GET_MODE_PRECISION.
14159 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14161 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14162 (expand_field_assignment, make_extraction): Likewise.
14163 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14164 (get_last_value): Likewise.
14165 * convert.c (convert_to_integer_1): Likewise.
14166 * cse.c (cse_insn): Likewise.
14167 * expr.c (expand_expr_real_1): Likewise.
14168 * lra-constraints.c (simplify_operand_subreg): Likewise.
14169 * optabs-query.c (can_atomic_load_p): Likewise.
14170 * optabs.c (expand_atomic_load): Likewise.
14171 (expand_atomic_store): Likewise.
14172 * ree.c (combine_reaching_defs): Likewise.
14173 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14174 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14175 * tree.h (type_has_mode_precision_p): Likewise.
14176 * ubsan.c (instrument_si_overflow): Likewise.
14178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14179 Alan Hayward <alan.hayward@arm.com>
14180 David Sherwood <david.sherwood@arm.com>
14182 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14183 polynomial numbers of units.
14184 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14185 (valid_vector_subparts_p): New function.
14186 (build_vector_type): Remove temporary shim and take the number
14187 of units as a poly_uint64 rather than an int.
14188 (build_opaque_vector_type): Take the number of units as a
14189 poly_uint64 rather than an int.
14190 * tree.c (build_vector_from_ctor): Handle polynomial
14191 TYPE_VECTOR_SUBPARTS.
14192 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14193 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14194 (build_vector_from_val): If the number of units is variable,
14195 use build_vec_duplicate_cst for constant operands and
14196 VEC_DUPLICATE_EXPR otherwise.
14197 (make_vector_type): Remove temporary is_constant ().
14198 (build_vector_type, build_opaque_vector_type): Take the number of
14199 units as a poly_uint64 rather than an int.
14200 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14202 * cfgexpand.c (expand_debug_expr): Likewise.
14203 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14204 (store_constructor, expand_expr_real_1): Likewise.
14205 (const_scalar_mask_from_tree): Likewise.
14206 * fold-const-call.c (fold_const_reduction): Likewise.
14207 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14208 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14209 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14210 (fold_relational_const): Likewise.
14211 (native_interpret_vector): Likewise. Change the size from an
14212 int to an unsigned int.
14213 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14214 TYPE_VECTOR_SUBPARTS.
14215 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14216 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14217 duplicating a non-constant operand into a variable-length vector.
14218 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14219 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14220 * ipa-icf.c (sem_variable::equals): Likewise.
14221 * match.pd: Likewise.
14222 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14223 * print-tree.c (print_node): Likewise.
14224 * stor-layout.c (layout_type): Likewise.
14225 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14226 * tree-cfg.c (verify_gimple_comparison): Likewise.
14227 (verify_gimple_assign_binary): Likewise.
14228 (verify_gimple_assign_ternary): Likewise.
14229 (verify_gimple_assign_single): Likewise.
14230 * tree-pretty-print.c (dump_generic_node): Likewise.
14231 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14232 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14233 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14234 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14235 (vect_shift_permute_load_chain): Likewise.
14236 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14237 (expand_vector_condition, optimize_vector_constructor): Likewise.
14238 (lower_vec_perm, get_compute_type): Likewise.
14239 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14240 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14241 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14242 (vect_recog_mask_conversion_pattern): Likewise.
14243 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14244 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14245 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14246 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14247 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14248 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14249 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14250 (supportable_widening_operation): Likewise.
14251 (supportable_narrowing_operation): Likewise.
14252 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14254 * varasm.c (output_constant): Likewise.
14256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14257 Alan Hayward <alan.hayward@arm.com>
14258 David Sherwood <david.sherwood@arm.com>
14260 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14261 so that both the length == 3 and length != 3 cases set up their
14262 own permute vectors. Add comments explaining why we know the
14263 number of elements is constant.
14264 (vect_permute_load_chain): Likewise.
14266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14267 Alan Hayward <alan.hayward@arm.com>
14268 David Sherwood <david.sherwood@arm.com>
14270 * machmode.h (mode_nunits): Change from unsigned char to
14272 (ONLY_FIXED_SIZE_MODES): New macro.
14273 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14274 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14275 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14277 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14278 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14279 or if measurement_type is not polynomial.
14280 * genmodes.c (ZERO_COEFFS): New macro.
14281 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14283 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14284 Use ZERO_COEFFS when emitting initializers.
14285 * data-streamer.h (bp_pack_poly_value): New function.
14286 (bp_unpack_poly_value): Likewise.
14287 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14288 for GET_MODE_NUNITS.
14289 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14290 for GET_MODE_NUNITS.
14291 * tree.c (make_vector_type): Remove temporary shim and make
14292 the real function take the number of units as a poly_uint64
14293 rather than an int.
14294 (build_vector_type_for_mode): Handle polynomial nunits.
14295 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14296 * emit-rtl.c (const_vec_series_p_1): Likewise.
14297 (gen_rtx_CONST_VECTOR): Likewise.
14298 * fold-const.c (test_vec_duplicate_folding): Likewise.
14299 * genrecog.c (validate_pattern): Likewise.
14300 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14301 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14302 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14303 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14304 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14305 * rtlanal.c (subreg_get_info): Likewise.
14306 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14307 (vect_grouped_load_supported): Likewise.
14308 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14309 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14310 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14311 (simplify_const_unary_operation, simplify_binary_operation_1)
14312 (simplify_const_binary_operation, simplify_ternary_operation)
14313 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14314 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14315 instead of CONST_VECTOR_NUNITS.
14316 * varasm.c (output_constant_pool_2): Likewise.
14317 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14318 explicit-encoded elements in the XVEC for variable-length vectors.
14320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14322 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14325 Alan Hayward <alan.hayward@arm.com>
14326 David Sherwood <david.sherwood@arm.com>
14328 * coretypes.h (fixed_size_mode): Declare.
14329 (fixed_size_mode_pod): New typedef.
14330 * builtins.h (target_builtins::x_apply_args_mode)
14331 (target_builtins::x_apply_result_mode): Change type to
14332 fixed_size_mode_pod.
14333 * builtins.c (apply_args_size, apply_result_size, result_vector)
14334 (expand_builtin_apply_args_1, expand_builtin_apply)
14335 (expand_builtin_return): Update accordingly.
14337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14339 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14340 * cselib.c (cselib_hash_rtx): Likewise.
14341 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14342 CONST_VECTOR encoding.
14344 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14345 Jeff Law <law@redhat.com>
14348 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14349 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14350 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14351 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14354 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14355 explicitly probe *sp in a noreturn function if there were any callee
14356 register saves or frame pointer is needed.
14358 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14361 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14362 BLKmode for ternary, binary or unary expressions.
14365 * var-tracking.c (delete_vta_debug_insn): New inline function.
14366 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14367 insns from get_insns () to NULL instead of each bb separately.
14368 Use delete_vta_debug_insn. No longer static.
14369 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14370 delete_vta_debug_insns callers.
14371 * rtl.h (delete_vta_debug_insns): Declare.
14372 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14373 instead of variable_tracking_main.
14375 2018-01-03 Martin Sebor <msebor@redhat.com>
14377 PR tree-optimization/83603
14378 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14379 arguments past the endof the argument list in functions declared
14380 without a prototype.
14381 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14382 Avoid checking when arguments are null.
14384 2018-01-03 Martin Sebor <msebor@redhat.com>
14387 * doc/extend.texi (attribute const): Fix a typo.
14388 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14389 issuing -Wsuggest-attribute for void functions.
14391 2018-01-03 Martin Sebor <msebor@redhat.com>
14393 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14394 offset_int::from instead of wide_int::to_shwi.
14395 (maybe_diag_overlap): Remove assertion.
14396 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14397 * gimple-ssa-sprintf.c (format_directive): Same.
14398 (parse_directive): Same.
14399 (sprintf_dom_walker::compute_format_length): Same.
14400 (try_substitute_return_value): Same.
14402 2018-01-03 Jeff Law <law@redhat.com>
14404 PR middle-end/83654
14405 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14406 non-constant residual for zero at runtime and avoid probing in
14407 that case. Reorganize code for trailing problem to mirror handling
14410 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14412 PR tree-optimization/83501
14413 * tree-ssa-strlen.c (get_string_cst): New.
14414 (handle_char_store): Call get_string_cst.
14416 2018-01-03 Martin Liska <mliska@suse.cz>
14418 PR tree-optimization/83593
14419 * tree-ssa-strlen.c: Include tree-cfg.h.
14420 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14421 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14422 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14424 (strlen_dom_walker::before_dom_children): Call
14425 gimple_purge_dead_eh_edges. Dump tranformation with details
14427 (strlen_dom_walker::before_dom_children): Update call by adding
14428 new argument cleanup_eh.
14429 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14431 2018-01-03 Martin Liska <mliska@suse.cz>
14434 * cif-code.def (VARIADIC_THUNK): New enum value.
14435 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14438 2018-01-03 Jan Beulich <jbeulich@suse.com>
14440 * sse.md (mov<mode>_internal): Tighten condition for when to use
14441 vmovdqu<ssescalarsize> for TI and OI modes.
14443 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14445 Update copyright years.
14447 2018-01-03 Martin Liska <mliska@suse.cz>
14450 * ipa-visibility.c (function_and_variable_visibility): Skip
14451 functions with noipa attribure.
14453 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14455 * gcc.c (process_command): Update copyright notice dates.
14456 * gcov-dump.c (print_version): Ditto.
14457 * gcov.c (print_version): Ditto.
14458 * gcov-tool.c (print_version): Ditto.
14459 * gengtype.c (create_file): Ditto.
14460 * doc/cpp.texi: Bump @copying's copyright year.
14461 * doc/cppinternals.texi: Ditto.
14462 * doc/gcc.texi: Ditto.
14463 * doc/gccint.texi: Ditto.
14464 * doc/gcov.texi: Ditto.
14465 * doc/install.texi: Ditto.
14466 * doc/invoke.texi: Ditto.
14468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14470 * vector-builder.h (vector_builder::m_full_nelts): Change from
14471 unsigned int to poly_uint64.
14472 (vector_builder::full_nelts): Update prototype accordingly.
14473 (vector_builder::new_vector): Likewise.
14474 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14475 (vector_builder::operator ==): Likewise.
14476 (vector_builder::finalize): Likewise.
14477 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14478 Take the number of elements as a poly_uint64 rather than an
14480 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14481 from unsigned int to poly_uint64.
14482 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14483 (vec_perm_indices::new_vector): Likewise.
14484 (vec_perm_indices::length): Likewise.
14485 (vec_perm_indices::nelts_per_input): Likewise.
14486 (vec_perm_indices::input_nelts): Likewise.
14487 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14488 number of elements per input as a poly_uint64 rather than an
14489 unsigned int. Use the original encoding for variable-length
14490 vectors, rather than clamping each individual element.
14491 For the second and subsequent elements in each pattern,
14492 clamp the step and base before clamping their sum.
14493 (vec_perm_indices::series_p): Handle polynomial element counts.
14494 (vec_perm_indices::all_in_range_p): Likewise.
14495 (vec_perm_indices_to_tree): Likewise.
14496 (vec_perm_indices_to_rtx): Likewise.
14497 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14498 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14499 (tree_vector_builder::new_binary_operation): Handle polynomial
14500 element counts. Return false if we need to know the number
14501 of elements at compile time.
14502 * fold-const.c (fold_vec_perm): Punt if the number of elements
14503 isn't known at compile time.
14505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14507 * vec-perm-indices.h (vec_perm_builder): Change element type
14508 from HOST_WIDE_INT to poly_int64.
14509 (vec_perm_indices::element_type): Update accordingly.
14510 (vec_perm_indices::clamp): Handle polynomial element_types.
14511 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14512 (vec_perm_indices::all_in_range_p): Likewise.
14513 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14515 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14516 polynomial vec_perm_indices element types.
14517 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14518 * fold-const.c (fold_vec_perm): Likewise.
14519 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14520 * tree-vect-generic.c (lower_vec_perm): Likewise.
14521 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14522 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14523 element type to HOST_WIDE_INT.
14525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14526 Alan Hayward <alan.hayward@arm.com>
14527 David Sherwood <david.sherwood@arm.com>
14529 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14530 rather than an int. Use plus_constant.
14531 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14532 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14535 Alan Hayward <alan.hayward@arm.com>
14536 David Sherwood <david.sherwood@arm.com>
14538 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14539 a HOST_WIDE_INT to a poly_int64.
14541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14542 Alan Hayward <alan.hayward@arm.com>
14543 David Sherwood <david.sherwood@arm.com>
14545 * calls.c (load_register_parameters): Cope with polynomial
14546 mode sizes. Require a constant size for BLKmode parameters
14547 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14548 forces a parameter to be padded at the lsb end in order to
14549 fill a complete number of words, require the parameter size
14550 to be ordered wrt UNITS_PER_WORD.
14552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14553 Alan Hayward <alan.hayward@arm.com>
14554 David Sherwood <david.sherwood@arm.com>
14556 * reload1.c (spill_stack_slot_width): Change element type
14557 from unsigned int to poly_uint64_pod.
14558 (alter_reg): Treat mode sizes as polynomial.
14560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * reload.c (complex_word_subreg_p): New function.
14565 (reload_inner_reg_of_subreg, push_reload): Use it.
14567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14568 Alan Hayward <alan.hayward@arm.com>
14569 David Sherwood <david.sherwood@arm.com>
14571 * lra-constraints.c (process_alt_operands): Reject matched
14572 operands whose sizes aren't ordered.
14573 (match_reload): Refer to this check here.
14575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14576 Alan Hayward <alan.hayward@arm.com>
14577 David Sherwood <david.sherwood@arm.com>
14579 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14580 that the mode size is in the set {1, 2, 4, 8, 16}.
14582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14583 Alan Hayward <alan.hayward@arm.com>
14584 David Sherwood <david.sherwood@arm.com>
14586 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14587 Use plus_constant instead of gen_rtx_PLUS.
14589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14590 Alan Hayward <alan.hayward@arm.com>
14591 David Sherwood <david.sherwood@arm.com>
14593 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14594 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14595 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14596 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14597 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14598 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14599 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14600 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14601 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14602 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14604 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14605 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14606 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14607 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14608 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14609 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14610 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14611 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14612 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14613 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14615 * expr.c (emit_move_resolve_push): Treat the input and result
14616 of PUSH_ROUNDING as a poly_int64.
14617 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14618 (emit_push_insn): Likewise.
14619 * lra-eliminations.c (mark_not_eliminable): Likewise.
14620 * recog.c (push_operand): Likewise.
14621 * reload1.c (elimination_effects): Likewise.
14622 * rtlanal.c (nonzero_bits1): Likewise.
14623 * calls.c (store_one_arg): Likewise. Require the padding to be
14624 known at compile time.
14626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14627 Alan Hayward <alan.hayward@arm.com>
14628 David Sherwood <david.sherwood@arm.com>
14630 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14631 Use plus_constant instead of gen_rtx_PLUS.
14633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14634 Alan Hayward <alan.hayward@arm.com>
14635 David Sherwood <david.sherwood@arm.com>
14637 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14638 rather than an int.
14640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14641 Alan Hayward <alan.hayward@arm.com>
14642 David Sherwood <david.sherwood@arm.com>
14644 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14645 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14646 via stack temporaries. Treat the mode size as polynomial too.
14648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14649 Alan Hayward <alan.hayward@arm.com>
14650 David Sherwood <david.sherwood@arm.com>
14652 * expr.c (expand_expr_real_2): When handling conversions involving
14653 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14654 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14655 as a poly_uint64 too.
14657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14658 Alan Hayward <alan.hayward@arm.com>
14659 David Sherwood <david.sherwood@arm.com>
14661 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14664 Alan Hayward <alan.hayward@arm.com>
14665 David Sherwood <david.sherwood@arm.com>
14667 * combine.c (can_change_dest_mode): Handle polynomial
14668 REGMODE_NATURAL_SIZE.
14669 * expmed.c (store_bit_field_1): Likewise.
14670 * expr.c (store_constructor): Likewise.
14671 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14672 and polynomial REGMODE_NATURAL_SIZE.
14673 (gen_lowpart_common): Likewise.
14674 * reginfo.c (record_subregs_of_mode): Likewise.
14675 * rtlanal.c (read_modify_subreg_p): Likewise.
14677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14678 Alan Hayward <alan.hayward@arm.com>
14679 David Sherwood <david.sherwood@arm.com>
14681 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14682 numbers of elements.
14684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14685 Alan Hayward <alan.hayward@arm.com>
14686 David Sherwood <david.sherwood@arm.com>
14688 * match.pd: Cope with polynomial numbers of vector elements.
14690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14691 Alan Hayward <alan.hayward@arm.com>
14692 David Sherwood <david.sherwood@arm.com>
14694 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14695 in a POINTER_PLUS_EXPR.
14697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14698 Alan Hayward <alan.hayward@arm.com>
14699 David Sherwood <david.sherwood@arm.com>
14701 * omp-simd-clone.c (simd_clone_subparts): New function.
14702 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14703 (ipa_simd_modify_function_body): Likewise.
14705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14706 Alan Hayward <alan.hayward@arm.com>
14707 David Sherwood <david.sherwood@arm.com>
14709 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14710 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14711 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14712 (expand_vector_condition, vector_element): Likewise.
14713 (subparts_gt): New function.
14714 (get_compute_type): Use subparts_gt.
14715 (count_type_subparts): Delete.
14716 (expand_vector_operations_1): Use subparts_gt instead of
14717 count_type_subparts.
14719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14720 Alan Hayward <alan.hayward@arm.com>
14721 David Sherwood <david.sherwood@arm.com>
14723 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14724 (vect_compile_time_alias): ...this new function. Do the calculation
14725 on poly_ints rather than trees.
14726 (vect_prune_runtime_alias_test_list): Update call accordingly.
14728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14729 Alan Hayward <alan.hayward@arm.com>
14730 David Sherwood <david.sherwood@arm.com>
14732 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14734 (vect_schedule_slp_instance): Likewise.
14736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14737 Alan Hayward <alan.hayward@arm.com>
14738 David Sherwood <david.sherwood@arm.com>
14740 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14741 constant and extern definitions for variable-length vectors.
14742 (vect_get_constant_vectors): Note that the number of units
14743 is known to be constant.
14745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14746 Alan Hayward <alan.hayward@arm.com>
14747 David Sherwood <david.sherwood@arm.com>
14749 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14750 of units as polynomial. Choose between WIDE and NARROW based
14753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14754 Alan Hayward <alan.hayward@arm.com>
14755 David Sherwood <david.sherwood@arm.com>
14757 * tree-vect-stmts.c (simd_clone_subparts): New function.
14758 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14761 Alan Hayward <alan.hayward@arm.com>
14762 David Sherwood <david.sherwood@arm.com>
14764 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14765 vectors as polynomial. Use build_index_vector for
14766 IFN_GOMP_SIMD_LANE.
14768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14769 Alan Hayward <alan.hayward@arm.com>
14770 David Sherwood <david.sherwood@arm.com>
14772 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14773 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14774 for variable-length vectors.
14775 (vectorizable_mask_load_store): Treat the number of units as
14776 polynomial, asserting that it is constant if the condition has
14777 already been enforced.
14778 (vectorizable_store, vectorizable_load): Likewise.
14780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14781 Alan Hayward <alan.hayward@arm.com>
14782 David Sherwood <david.sherwood@arm.com>
14784 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14785 of units as polynomial. Punt if we can't tell at compile time
14786 which vector contains the final result.
14788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14789 Alan Hayward <alan.hayward@arm.com>
14790 David Sherwood <david.sherwood@arm.com>
14792 * tree-vect-loop.c (vectorizable_induction): Treat the number
14793 of units as polynomial. Punt on SLP inductions. Use an integer
14794 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14795 cast of such a series for variable-length floating-point
14798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14799 Alan Hayward <alan.hayward@arm.com>
14800 David Sherwood <david.sherwood@arm.com>
14802 * tree.h (build_index_vector): Declare.
14803 * tree.c (build_index_vector): New function.
14804 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14805 of units as polynomial, forcibly converting it to a constant if
14806 vectorizable_reduction has already enforced the condition.
14807 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14808 to create a {1,2,3,...} vector.
14809 (vectorizable_reduction): Treat the number of units as polynomial.
14810 Choose vectype_in based on the largest scalar element size rather
14811 than the smallest number of units. Enforce the restrictions
14814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14815 Alan Hayward <alan.hayward@arm.com>
14816 David Sherwood <david.sherwood@arm.com>
14818 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14819 number of units as polynomial.
14821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14822 Alan Hayward <alan.hayward@arm.com>
14823 David Sherwood <david.sherwood@arm.com>
14825 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14826 * target.def (autovectorize_vector_sizes): Return the vector sizes
14827 by pointer, using vector_sizes rather than a bitmask.
14828 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14829 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14830 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14832 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14833 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14834 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14835 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14836 * omp-general.c (omp_max_vf): Likewise.
14837 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14838 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14839 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14840 * tree-vect-slp.c (vect_slp_bb): Likewise.
14841 * doc/tm.texi: Regenerate.
14842 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14844 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14845 the vector size as a poly_uint64 rather than an unsigned int.
14846 (current_vector_size): Change from an unsigned int to a poly_uint64.
14847 (get_vectype_for_scalar_type): Update accordingly.
14848 * tree.h (build_truth_vector_type): Take the size and number of
14849 units as a poly_uint64 rather than an unsigned int.
14850 (build_vector_type): Add a temporary overload that takes
14851 the number of units as a poly_uint64 rather than an unsigned int.
14852 * tree.c (make_vector_type): Likewise.
14853 (build_truth_vector_type): Take the number of units as a poly_uint64
14854 rather than an unsigned int.
14856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14857 Alan Hayward <alan.hayward@arm.com>
14858 David Sherwood <david.sherwood@arm.com>
14860 * target.def (get_mask_mode): Take the number of units and length
14861 as poly_uint64s rather than unsigned ints.
14862 * targhooks.h (default_get_mask_mode): Update accordingly.
14863 * targhooks.c (default_get_mask_mode): Likewise.
14864 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14865 * doc/tm.texi: Regenerate.
14867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14868 Alan Hayward <alan.hayward@arm.com>
14869 David Sherwood <david.sherwood@arm.com>
14871 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14872 * omp-general.c (omp_max_vf): Likewise.
14873 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14874 (expand_omp_simd): Handle polynomial safelen.
14875 * omp-low.c (omplow_simd_context): Add a default constructor.
14876 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14877 (lower_rec_simd_input_clauses): Update accordingly.
14878 (lower_rec_input_clauses): Likewise.
14880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14881 Alan Hayward <alan.hayward@arm.com>
14882 David Sherwood <david.sherwood@arm.com>
14884 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14885 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14886 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14887 (vect_analyze_slp_cost): Likewise.
14888 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14889 (vect_model_load_cost): Likewise.
14891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14892 Alan Hayward <alan.hayward@arm.com>
14893 David Sherwood <david.sherwood@arm.com>
14895 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14896 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14897 from an unsigned int * to a poly_uint64_pod *.
14898 (calculate_unrolling_factor): New function.
14899 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14902 Alan Hayward <alan.hayward@arm.com>
14903 David Sherwood <david.sherwood@arm.com>
14905 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14906 from an unsigned int to a poly_uint64.
14907 (_loop_vec_info::slp_unrolling_factor): Likewise.
14908 (_loop_vec_info::vectorization_factor): Change from an int
14910 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14911 (vect_get_num_vectors): New function.
14912 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14913 (vect_get_num_copies): Use vect_get_num_vectors.
14914 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14915 to an unsigned int *.
14916 (vect_analyze_data_refs): Change min_vf from an int * to a
14918 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14919 than an unsigned HOST_WIDE_INT.
14920 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14921 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14922 to an unsigned int *.
14923 (vect_analyze_data_ref_dependences): Likewise.
14924 (vect_compute_data_ref_alignment): Handle polynomial vf.
14925 (vect_enhance_data_refs_alignment): Likewise.
14926 (vect_prune_runtime_alias_test_list): Likewise.
14927 (vect_shift_permute_load_chain): Likewise.
14928 (vect_supportable_dr_alignment): Likewise.
14929 (dependence_distance_ge_vf): Take the vectorization factor as a
14930 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14931 (vect_analyze_data_refs): Change min_vf from an int * to a
14933 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14934 vfm1 as a poly_uint64 rather than an int. Make the same change
14935 for the returned bound_scalar.
14936 (vect_gen_vector_loop_niters): Handle polynomial vf.
14937 (vect_do_peeling): Likewise. Update call to
14938 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14939 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14941 * tree-vect-loop.c (vect_determine_vectorization_factor)
14942 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14943 (vect_get_known_peeling_cost): Likewise.
14944 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14945 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14946 (vect_transform_loop): Likewise. Use the lowest possible VF when
14947 updating the upper bounds of the loop.
14948 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14949 rather than an int.
14950 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14951 polynomial unroll factors.
14952 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14953 (vect_make_slp_decision): Likewise.
14954 (vect_supported_load_permutation_p): Likewise, and polynomial
14956 (vect_analyze_slp_cost): Handle polynomial vf.
14957 (vect_slp_analyze_node_operations): Likewise.
14958 (vect_slp_analyze_bb_1): Likewise.
14959 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14960 than an unsigned HOST_WIDE_INT.
14961 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14962 (vectorizable_load): Handle polynomial vf.
14963 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14965 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14968 Alan Hayward <alan.hayward@arm.com>
14969 David Sherwood <david.sherwood@arm.com>
14971 * match.pd: Handle bit operations involving three constants
14972 and try to fold one pair.
14974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14976 * tree-vect-loop-manip.c: Include gimple-fold.h.
14977 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14978 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14979 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14980 Add a path that uses a step of VF instead of 1, but disable it
14982 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14983 and niters_no_overflow parameters. Update calls to
14984 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14985 Create a new SSA name if the latter choses to use a ste other
14986 than zero, and return it via niters_vector_mult_vf_var.
14987 * tree-vect-loop.c (vect_transform_loop): Update calls to
14988 vect_do_peeling, vect_gen_vector_loop_niters and
14989 slpeel_make_loop_iterate_ntimes.
14990 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14991 (vect_gen_vector_loop_niters): Update declarations after above changes.
14993 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14995 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14996 128-bit round to integer instructions.
14997 (ceil<mode>2): Likewise.
14998 (btrunc<mode>2): Likewise.
14999 (round<mode>2): Likewise.
15001 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15003 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
15004 unaligned VSX load/store on P8/P9.
15005 (expand_block_clear): Allow the use of unaligned VSX
15006 load/store on P8/P9.
15008 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15010 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
15012 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
15013 swap associated with both a load and a store.
15015 2018-01-02 Andrew Waterman <andrew@sifive.com>
15017 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
15018 * config/riscv/riscv.md (clear_cache): Use it.
15020 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
15022 * web.c: Remove out-of-date comment.
15024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15026 * expr.c (fixup_args_size_notes): Check that any existing
15027 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
15028 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
15029 (emit_single_push_insn): ...here.
15031 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15033 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
15034 (const_vector_encoded_nelts): New function.
15035 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
15036 (const_vector_int_elt, const_vector_elt): Declare.
15037 * emit-rtl.c (const_vector_int_elt_1): New function.
15038 (const_vector_elt): Likewise.
15039 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
15040 of CONST_VECTOR_ELT.
15042 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15044 * expr.c: Include rtx-vector-builder.h.
15045 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
15046 directly on the tree encoding.
15047 (const_vector_from_tree): Likewise.
15048 * optabs.c: Include rtx-vector-builder.h.
15049 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
15050 sequence of "u" values.
15051 * vec-perm-indices.c: Include rtx-vector-builder.h.
15052 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
15053 directly on the vec_perm_indices encoding.
15055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15057 * doc/rtl.texi (const_vector): Describe new encoding scheme.
15058 * Makefile.in (OBJS): Add rtx-vector-builder.o.
15059 * rtx-vector-builder.h: New file.
15060 * rtx-vector-builder.c: Likewise.
15061 * rtl.h (rtx_def::u2): Add a const_vector field.
15062 (CONST_VECTOR_NPATTERNS): New macro.
15063 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
15064 (CONST_VECTOR_DUPLICATE_P): Likewise.
15065 (CONST_VECTOR_STEPPED_P): Likewise.
15066 (CONST_VECTOR_ENCODED_ELT): Likewise.
15067 (const_vec_duplicate_p): Check for a duplicated vector encoding.
15068 (unwrap_const_vec_duplicate): Likewise.
15069 (const_vec_series_p): Check for a non-duplicated vector encoding.
15070 Say that the function only returns true for integer vectors.
15071 * emit-rtl.c: Include rtx-vector-builder.h.
15072 (gen_const_vec_duplicate_1): Delete.
15073 (gen_const_vector): Call gen_const_vec_duplicate instead of
15074 gen_const_vec_duplicate_1.
15075 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
15076 (gen_const_vec_duplicate): Use rtx_vector_builder.
15077 (gen_const_vec_series): Likewise.
15078 (gen_rtx_CONST_VECTOR): Likewise.
15079 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
15080 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15081 Build a new vector rather than modifying a CONST_VECTOR in-place.
15082 (handle_special_swappables): Update call accordingly.
15083 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
15084 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
15085 Build a new vector rather than modifying a CONST_VECTOR in-place.
15086 (handle_special_swappables): Update call accordingly.
15088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15090 * simplify-rtx.c (simplify_const_binary_operation): Use
15091 CONST_VECTOR_ELT instead of XVECEXP.
15093 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15095 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15096 the selector elements to be different from the data elements
15097 if the selector is a VECTOR_CST.
15098 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15099 ssizetype for the selector.
15101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15103 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15104 before testing each element individually.
15105 * tree-vect-generic.c (lower_vec_perm): Likewise.
15107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15109 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15110 * selftest-run-tests.c (selftest::run_tests): Call it.
15111 * vector-builder.h (vector_builder::operator ==): New function.
15112 (vector_builder::operator !=): Likewise.
15113 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15114 (vec_perm_indices::all_from_input_p): New function.
15115 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15116 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15117 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15118 instead of reading the VECTOR_CST directly. Detect whether both
15119 vector inputs are the same before constructing the vec_perm_indices,
15120 and update the number of inputs argument accordingly. Use the
15121 utility functions added above. Only construct sel2 if we need to.
15123 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15125 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15126 the broadcast of the low byte.
15127 (expand_mult_highpart): Use an explicit encoding for the permutes.
15128 * optabs-query.c (can_mult_highpart_p): Likewise.
15129 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15130 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15131 (vectorizable_bswap): Likewise.
15132 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15133 explicit encoding for the power-of-2 permutes.
15134 (vect_permute_store_chain): Likewise.
15135 (vect_grouped_load_supported): Likewise.
15136 (vect_permute_load_chain): Likewise.
15138 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15140 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15141 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15142 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15143 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15144 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15145 (vect_gen_perm_mask_any): Likewise.
15147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15149 * int-vector-builder.h: New file.
15150 * vec-perm-indices.h: Include int-vector-builder.h.
15151 (vec_perm_indices): Redefine as an int_vector_builder.
15152 (auto_vec_perm_indices): Delete.
15153 (vec_perm_builder): Redefine as a stand-alone class.
15154 (vec_perm_indices::vec_perm_indices): New function.
15155 (vec_perm_indices::clamp): Likewise.
15156 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15157 (vec_perm_indices::new_vector): New function.
15158 (vec_perm_indices::new_expanded_vector): Update for new
15159 vec_perm_indices class.
15160 (vec_perm_indices::rotate_inputs): New function.
15161 (vec_perm_indices::all_in_range_p): Operate directly on the
15162 encoded form, without computing elided elements.
15163 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15164 encoding. Update for new vec_perm_indices class.
15165 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15166 the given vec_perm_builder.
15167 (expand_vec_perm_var): Update vec_perm_builder constructor.
15168 (expand_mult_highpart): Use vec_perm_builder instead of
15169 auto_vec_perm_indices.
15170 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15171 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15172 or double series encoding as appropriate.
15173 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15174 vec_perm_indices instead of auto_vec_perm_indices.
15175 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15176 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15177 (vect_permute_store_chain): Likewise.
15178 (vect_grouped_load_supported): Likewise.
15179 (vect_permute_load_chain): Likewise.
15180 (vect_shift_permute_load_chain): Likewise.
15181 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15182 (vect_transform_slp_perm_load): Likewise.
15183 (vect_schedule_slp_instance): Likewise.
15184 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15185 (vectorizable_mask_load_store): Likewise.
15186 (vectorizable_bswap): Likewise.
15187 (vectorizable_store): Likewise.
15188 (vectorizable_load): Likewise.
15189 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15190 vec_perm_indices instead of auto_vec_perm_indices. Use
15191 tree_to_vec_perm_builder to read the vector from a tree.
15192 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15193 vec_perm_builder instead of a vec_perm_indices.
15194 (have_whole_vector_shift): Use vec_perm_builder and
15195 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15196 truncation to calc_vec_perm_mask_for_shift.
15197 (vect_create_epilog_for_reduction): Likewise.
15198 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15199 from auto_vec_perm_indices to vec_perm_indices.
15200 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15201 instead of changing individual elements.
15202 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15203 the vector in d.perm.
15204 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15205 from auto_vec_perm_indices to vec_perm_indices.
15206 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15207 instead of changing individual elements.
15208 (arm_vectorize_vec_perm_const): Use new_vector to install
15209 the vector in d.perm.
15210 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15211 Update vec_perm_builder constructor.
15212 (rs6000_expand_interleave): Likewise.
15213 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15214 (rs6000_expand_interleave): Likewise.
15216 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15218 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15219 to qimode could truncate the indices.
15220 * optabs.c (expand_vec_perm_var): Likewise.
15222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15224 * Makefile.in (OBJS): Add vec-perm-indices.o.
15225 * vec-perm-indices.h: New file.
15226 * vec-perm-indices.c: Likewise.
15227 * target.h (vec_perm_indices): Replace with a forward class
15229 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15230 * optabs.h: Include vec-perm-indices.h.
15231 (expand_vec_perm): Delete.
15232 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15233 (expand_vec_perm_const): Declare.
15234 * target.def (vec_perm_const_ok): Replace with...
15235 (vec_perm_const): ...this new hook.
15236 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15237 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15238 * doc/tm.texi: Regenerate.
15239 * optabs.def (vec_perm_const): Delete.
15240 * doc/md.texi (vec_perm_const): Likewise.
15241 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15242 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15243 expand_vec_perm for constant permutation vectors. Assert that
15244 the mode of variable permutation vectors is the integer equivalent
15245 of the mode that is being permuted.
15246 * optabs-query.h (selector_fits_mode_p): Declare.
15247 * optabs-query.c: Include vec-perm-indices.h.
15248 (selector_fits_mode_p): New function.
15249 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15250 is defined, instead of checking whether the vec_perm_const_optab
15251 exists. Use targetm.vectorize.vec_perm_const instead of
15252 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15253 fit in the vector mode before using a variable permute.
15254 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15255 vec_perm_indices instead of an rtx.
15256 (expand_vec_perm): Replace with...
15257 (expand_vec_perm_const): ...this new function. Take the selector
15258 as a vec_perm_indices rather than an rtx. Also take the mode of
15259 the selector. Update call to shift_amt_for_vec_perm_mask.
15260 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15261 Use vec_perm_indices::new_expanded_vector to expand the original
15262 selector into bytes. Check whether the indices fit in the vector
15263 mode before using a variable permute.
15264 (expand_vec_perm_var): Make global.
15265 (expand_mult_highpart): Use expand_vec_perm_const.
15266 * fold-const.c: Includes vec-perm-indices.h.
15267 * tree-ssa-forwprop.c: Likewise.
15268 * tree-vect-data-refs.c: Likewise.
15269 * tree-vect-generic.c: Likewise.
15270 * tree-vect-loop.c: Likewise.
15271 * tree-vect-slp.c: Likewise.
15272 * tree-vect-stmts.c: Likewise.
15273 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15275 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15276 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15277 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15278 (aarch64_vectorize_vec_perm_const): ...this new function.
15279 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15280 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15281 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15282 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15283 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15284 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15285 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15287 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15288 check for NEON modes.
15289 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15290 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15291 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15292 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15294 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15295 the old VEC_PERM_CONST conditions.
15296 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15297 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15298 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15299 (ia64_vectorize_vec_perm_const_ok): Merge into...
15300 (ia64_vectorize_vec_perm_const): ...this new function.
15301 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15302 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15303 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15304 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15305 * config/mips/mips.c (mips_expand_vec_perm_const)
15306 (mips_vectorize_vec_perm_const_ok): Merge into...
15307 (mips_vectorize_vec_perm_const): ...this new function.
15308 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15309 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15310 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15311 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15312 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15313 (rs6000_expand_vec_perm_const): Delete.
15314 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15316 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15317 (altivec_expand_vec_perm_const_le): Take each operand individually.
15318 Operate on constant selectors rather than rtxes.
15319 (altivec_expand_vec_perm_const): Likewise. Update call to
15320 altivec_expand_vec_perm_const_le.
15321 (rs6000_expand_vec_perm_const): Delete.
15322 (rs6000_vectorize_vec_perm_const_ok): Delete.
15323 (rs6000_vectorize_vec_perm_const): New function.
15324 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15325 an element count and rtx array.
15326 (rs6000_expand_extract_even): Update call accordingly.
15327 (rs6000_expand_interleave): Likewise.
15328 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15329 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15330 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15331 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15332 (rs6000_expand_vec_perm_const): Delete.
15333 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15334 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15335 (altivec_expand_vec_perm_const_le): Take each operand individually.
15336 Operate on constant selectors rather than rtxes.
15337 (altivec_expand_vec_perm_const): Likewise. Update call to
15338 altivec_expand_vec_perm_const_le.
15339 (rs6000_expand_vec_perm_const): Delete.
15340 (rs6000_vectorize_vec_perm_const_ok): Delete.
15341 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15342 reference to the SPE evmerge intructions.
15343 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15344 an element count and rtx array.
15345 (rs6000_expand_extract_even): Update call accordingly.
15346 (rs6000_expand_interleave): Likewise.
15347 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15348 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15350 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15354 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15355 vector mode and that that mode matches the mode of the data
15357 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15358 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15359 directly using expand_vec_perm_1 when forcing selectors into
15361 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15365 * optabs-query.h (can_vec_perm_p): Delete.
15366 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15367 * optabs-query.c (can_vec_perm_p): Split into...
15368 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15369 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15370 particular selector is valid.
15371 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15372 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15373 (vect_grouped_load_supported): Likewise.
15374 (vect_shift_permute_load_chain): Likewise.
15375 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15376 (vect_transform_slp_perm_load): Likewise.
15377 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15378 (vectorizable_bswap): Likewise.
15379 (vect_gen_perm_mask_checked): Likewise.
15380 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15381 implementations of variable permutation vectors into account
15382 when deciding which selector to use.
15383 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15384 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15385 with a false third argument.
15386 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15387 to test whether the constant selector is valid and can_vec_perm_var_p
15388 to test whether a variable selector is valid.
15390 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15392 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15393 * optabs-query.c (can_vec_perm_p): Likewise.
15394 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15395 instead of vec_perm_indices.
15396 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15397 (vect_gen_perm_mask_checked): Likewise,
15398 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15399 (vect_gen_perm_mask_checked): Likewise,
15401 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15403 * optabs-query.h (qimode_for_vec_perm): Declare.
15404 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15405 (qimode_for_vec_perm): ...this new function.
15406 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15408 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15410 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15411 does not have a conditional at the top.
15413 2018-01-02 Richard Biener <rguenther@suse.de>
15415 * ipa-inline.c (big_speedup_p): Fix expression.
15417 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15420 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15423 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15427 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15428 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15429 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15430 cond_taken_branch_cost 3->4.
15432 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15434 PR tree-optimization/83581
15435 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15436 TODO_cleanup_cfg if any changes have been made.
15438 PR middle-end/83608
15439 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15440 convert_modes if target mode has the right side, but different mode
15443 PR middle-end/83609
15444 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15445 last argument when extracting from CONCAT. If either from_real or
15446 from_imag is NULL, use expansion through memory. If result is not
15447 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15448 the parts directly to inner mode, if even that fails, use expansion
15451 PR middle-end/83623
15452 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15453 check for bswap in mode rather than HImode and use that in expand_unop
15456 Copyright (C) 2018 Free Software Foundation, Inc.
15458 Copying and distribution of this file, with or without modification,
15459 are permitted in any medium without royalty provided the copyright
15460 notice and this notice are preserved.