1 2020-03-13 Segher Boessenkool <segher@kernel.crashing.org>
3 PR rtl-optimization/94148
4 PR rtl-optimization/94042
5 * df-core.c (BB_LAST_CHANGE_AGE): Delete.
6 (df_worklist_propagate_forward): New parameter last_change_age, use
7 that instead of bb->aux.
8 (df_worklist_propagate_backward): Ditto.
9 (df_worklist_dataflow_doublequeue): Use a local array last_change_age.
11 2020-03-13 Richard Biener <rguenther@suse.de>
13 PR tree-optimization/94163
14 * tree-ssa-pre.c (create_expression_by_pieces): Check
15 whether alignment would be zero.
17 2020-03-13 Martin Liska <mliska@suse.cz>
20 * lto-wrapper.c (run_gcc): Use concat for appending
21 to collect_gcc_options.
23 2020-03-13 Jakub Jelinek <jakub@redhat.com>
26 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use gen_int_mode
29 2020-03-13 H.J. Lu <hongjiu.lu@intel.com>
32 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DF.
33 * config/i386/i386.md (*movdf_internal): Call ix86_output_ssemov
34 for TYPE_SSEMOV. Remove TARGET_AVX512F, TARGET_PREFER_AVX256,
35 TARGET_AVX512VL and ext_sse_reg_operand check.
37 2020-03-13 Bu Le <bule1@huawei.com>
40 * config/aarch64/aarch64.opt (-param=aarch64-float-recp-precision=)
41 (-param=aarch64-double-recp-precision=): New options.
42 * doc/invoke.texi: Document them.
43 * config/aarch64/aarch64.c (aarch64_emit_approx_div): Use them
44 instead of hard-coding the choice of 1 for float and 2 for double.
46 2019-03-13 Eric Botcazou <ebotcazou@adacore.com>
48 PR rtl-optimization/94119
49 * resource.h (clear_hashed_info_until_next_barrier): Declare.
50 * resource.c (clear_hashed_info_until_next_barrier): New function.
51 * reorg.c (add_to_delay_list): Fix formatting.
52 (relax_delay_slots): Call clear_hashed_info_until_next_barrier on
53 the next instruction after removing a BARRIER.
55 2019-03-13 Eric Botcazou <ebotcazou@adacore.com>
58 * expmed.c (store_integral_bit_field): For fields larger than a word,
59 call extract_bit_field on the value if the mode is BLKmode. Remove
60 specific path for big-endian targets and tidy things up a little bit.
62 2020-03-12 Richard Sandiford <richard.sandiford@arm.com>
64 PR rtl-optimization/90275
65 * cse.c (cse_insn): Delete no-op register moves too.
67 2020-03-12 Darius Galis <darius.galis@cyberthorstudios.com>
69 * config/rx/rx.md (CTRLREG_CPEN): Remove.
70 * config/rx/rx.c (rx_print_operand): Remove CTRLREG_CPEN support.
72 2020-03-12 Richard Biener <rguenther@suse.de>
74 PR tree-optimization/94103
75 * tree-ssa-sccvn.c (visit_reference_op_load): Avoid type
76 punning when the mode precision is not sufficient.
78 2020-03-12 H.J. Lu <hongjiu.lu@intel.com>
81 * config/i386/i386.c (ix86_output_ssemov): Handle MODE_DI,
82 MODE_V1DF and MODE_V2SF.
83 * config/i386/mmx.md (MMXMODE:*mov<mode>_internal): Call
84 ix86_output_ssemov for TYPE_SSEMOV. Remove ext_sse_reg_operand
87 2020-03-12 Jakub Jelinek <jakub@redhat.com>
89 * doc/tm.texi.in (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Change
90 ASM_OUTPUT_ALIGNED_DECL in description to ASM_OUTPUT_ALIGNED_LOCAL
91 and ASM_OUTPUT_DECL to ASM_OUTPUT_LOCAL.
92 * doc/tm.texi: Regenerated.
94 PR tree-optimization/94130
95 * tree-ssa-dse.c: Include gimplify.h.
96 (increment_start_addr): If stmt has lhs, drop the lhs from call and
97 set it after the call to the original value of the first argument.
99 (decrement_count): Formatting fix.
101 2020-03-11 Delia Burduv <delia.burduv@arm.com>
103 * config/arm/arm-builtins.c
104 (arm_init_simd_builtin_scalar_types): New.
105 * config/arm/arm_neon.h (vld2_bf16): Used new builtin type.
106 (vld2q_bf16): Used new builtin type.
107 (vld3_bf16): Used new builtin type.
108 (vld3q_bf16): Used new builtin type.
109 (vld4_bf16): Used new builtin type.
110 (vld4q_bf16): Used new builtin type.
111 (vld2_dup_bf16): Used new builtin type.
112 (vld2q_dup_bf16): Used new builtin type.
113 (vld3_dup_bf16): Used new builtin type.
114 (vld3q_dup_bf16): Used new builtin type.
115 (vld4_dup_bf16): Used new builtin type.
116 (vld4q_dup_bf16): Used new builtin type.
118 2020-03-11 Jakub Jelinek <jakub@redhat.com>
121 * config/pdp11/pdp11.c (pdp11_asm_output_var): Call switch_to_section
122 at the start to switch to data section. Don't print extra newline if
123 .globl directive has not been emitted.
125 2020-03-11 Richard Biener <rguenther@suse.de>
127 * match.pd ((T *)(ptr - ptr-cst) -> &MEM[ptr + -ptr-cst]):
130 2020-03-11 Eric Botcazou <ebotcazou@adacore.com>
133 * tree.c (variably_modified_type_p) <RECORD_TYPE>: Recurse into fields
134 whose type is a qualified union.
136 2020-03-11 Jakub Jelinek <jakub@redhat.com>
139 * config/aarch64/aarch64.c (aarch64_add_offset_1): Use absu_hwi
140 instead of abs_hwi, change moffset type to unsigned HOST_WIDE_INT.
143 * value-prof.c (dump_histogram_value): Use abs_hwi instead of
145 (get_nth_most_common_value): Use abs_hwi instead of abs.
148 * dfp.c (decimal_to_binary): Only use decimal128ToString if from->cl
149 is rvc_normal, otherwise use real_to_decimal to print the number to
152 PR tree-optimization/94114
153 * tree-loop-distribution.c (generate_memset_builtin): Call
154 rewrite_to_non_trapping_overflow even on mem.
155 (generate_memcpy_builtin): Call rewrite_to_non_trapping_overflow even
158 2020-03-10 Jeff Law <law@redhat.com>
160 * config/bfin/bfin.md (movsi_insv): Add length attribute.
162 2020-03-10 Jiufu Guo <guojiufu@linux.ibm.com>
165 * gcc/config/rs6000/rs6000.c (rs6000_emit_p9_fp_minmax): Check
166 NAN and SIGNED_ZEROR for smax/smin.
168 2020-03-10 Will Schmidt <will_schmidt@vnet.ibm.com>
171 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
172 clause to handle P9V_BUILTIN_VEC_LXVL with const arguments.
174 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
176 * loop-iv.c (find_simple_exit): Make it static.
177 * cfgloop.h: Remove the corresponding prototype.
179 2020-03-10 Roman Zhuykov <zhroma@ispras.ru>
181 * ddg.c (create_ddg): Fix intendation.
182 (set_recurrence_length): Likewise.
183 (create_ddg_all_sccs): Likewise.
185 2020-03-10 Jakub Jelinek <jakub@redhat.com>
188 * config/i386/i386.md (*testqi_ext_3): Call ix86_match_ccmode with
189 CCZmode instead of CCNOmode if operands[2] has DImode and pos + len
192 2020-03-09 Jason Merrill <jason@redhat.com>
194 * gdbinit.in (pgs): Fix typo in documentation.
196 2020-03-09 Vladimir Makarov <vmakarov@redhat.com>
200 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
202 PR rtl-optimization/93564
203 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
204 do not honor reg alloc order.
206 2020-03-09 Andrew Pinski <apinski@marvell.com>
209 * doc/extend.texi (x86 Operand Modifiers): Fix column
212 2020-03-09 Martin Liska <mliska@suse.cz>
215 * config/rs6000/rs6000.c (rs6000_option_override_internal):
216 Remove set of str_align_loops and str_align_jumps as these
217 should be set in previous 2 conditions in the function.
219 2020-03-09 Jakub Jelinek <jakub@redhat.com>
221 PR rtl-optimization/94045
222 * params.opt (-param=max-find-base-term-values=): New option.
223 * alias.c (find_base_term): Add cut-off for number of visited VALUEs
224 in a single toplevel find_base_term call.
226 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
229 * config/aarch64/aarch64-builtins.c (TYPES_TERNOPU_LANE): Add define.
230 * config/aarch64/aarch64-simd.md
231 (aarch64_vec_<su>mult_lane<Qlane>): Add new insn for widening lane mul.
232 (aarch64_vec_<su>mlal_lane<Qlane>): Likewise.
233 * config/aarch64/aarch64-simd-builtins.def: Add intrinsics.
234 * config/aarch64/arm_neon.h:
235 (vmlal_lane_s16): Expand using intrinsics rather than inline asm.
236 (vmlal_lane_u16): Likewise.
237 (vmlal_lane_s32): Likewise.
238 (vmlal_lane_u32): Likewise.
239 (vmlal_laneq_s16): Likewise.
240 (vmlal_laneq_u16): Likewise.
241 (vmlal_laneq_s32): Likewise.
242 (vmlal_laneq_u32): Likewise.
243 (vmull_lane_s16): Likewise.
244 (vmull_lane_u16): Likewise.
245 (vmull_lane_s32): Likewise.
246 (vmull_lane_u32): Likewise.
247 (vmull_laneq_s16): Likewise.
248 (vmull_laneq_u16): Likewise.
249 (vmull_laneq_s32): Likewise.
250 (vmull_laneq_u32): Likewise.
251 * config/aarch64/iterators.md (Vcondtype): New iterator for lane mul.
254 2020-03-06 Wilco Dijkstra <wdijkstr@arm.com>
256 * aarch64/aarch64-simd.md (aarch64_mla_elt<mode>): Correct lane syntax.
257 (aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
258 (aarch64_mls_elt<mode>): Likewise.
259 (aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
260 (aarch64_fma4_elt<mode>): Likewise.
261 (aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
262 (aarch64_fma4_elt_to_64v2df): Likewise.
263 (aarch64_fnma4_elt<mode>): Likewise.
264 (aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
265 (aarch64_fnma4_elt_to_64v2df): Likewise.
267 2020-03-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
269 * config/aarch64/aarch64-sve2.md (@aarch64_sve_<sve_int_op><mode>:
270 Specify movprfx attribute.
271 (@aarch64_sve_<sve_int_op>_lane_<mode>): Likewise.
273 2020-03-06 David Edelsohn <dje.gcc@gmail.com>
276 * config/rs6000/aix61.h (TARGET_NO_SUM_IN_TOC): Set to 1 for
278 (TARGET_NO_FP_IN_TOC): Same.
279 * config/rs6000/aix71.h: Same.
280 * config/rs6000/aix72.h: Same.
282 2020-03-06 Andrew Pinski <apinski@marvell.com>
283 Jeff Law <law@redhat.com>
285 PR rtl-optimization/93996
286 * haifa-sched.c (remove_notes): Be more careful when adding
289 2020-03-06 Delia Burduv <delia.burduv@arm.com>
291 * config/arm/arm_neon.h (vld2_bf16): New.
297 (vld2_dup_bf16): New.
298 (vld2q_dup_bf16): New.
299 (vld3_dup_bf16): New.
300 (vld3q_dup_bf16): New.
301 (vld4_dup_bf16): New.
302 (vld4q_dup_bf16): New.
303 * config/arm/arm_neon_builtins.def
304 (vld2): Changed to VAR13 and added v4bf, v8bf
305 (vld2_dup): Changed to VAR8 and added v4bf, v8bf
306 (vld3): Changed to VAR13 and added v4bf, v8bf
307 (vld3_dup): Changed to VAR8 and added v4bf, v8bf
308 (vld4): Changed to VAR13 and added v4bf, v8bf
309 (vld4_dup): Changed to VAR8 and added v4bf, v8bf
310 * config/arm/iterators.md (VDXBF2): New iterator.
311 *config/arm/neon.md (neon_vld2): Use new iterators.
312 (neon_vld2_dup<mode): Use new iterators.
313 (neon_vld3<mode>): Likewise.
314 (neon_vld3qa<mode>): Likewise.
315 (neon_vld3qb<mode>): Likewise.
316 (neon_vld3_dup<mode>): Likewise.
317 (neon_vld4<mode>): Likewise.
318 (neon_vld4qa<mode>): Likewise.
319 (neon_vld4qb<mode>): Likewise.
320 (neon_vld4_dup<mode>): Likewise.
321 (neon_vld2_dupv8bf): New.
322 (neon_vld3_dupv8bf): Likewise.
323 (neon_vld4_dupv8bf): Likewise.
325 2020-03-06 Delia Burduv <delia.burduv@arm.com>
327 * config/arm/arm_neon.h (bfloat16x4x2_t): New typedef.
328 (bfloat16x8x2_t): New typedef.
329 (bfloat16x4x3_t): New typedef.
330 (bfloat16x8x3_t): New typedef.
331 (bfloat16x4x4_t): New typedef.
332 (bfloat16x8x4_t): New typedef.
339 * config/arm/arm-builtins.c (v2bf_UP): Define.
341 (arm_init_simd_builtin_types): Init Bfloat16x2_t eltype.
342 * config/arm/arm-modes.def (V2BF): New mode.
343 * config/arm/arm-simd-builtin-types.def
344 (Bfloat16x2_t): New entry.
345 * config/arm/arm_neon_builtins.def
346 (vst2): Changed to VAR13 and added v4bf, v8bf
347 (vst3): Changed to VAR13 and added v4bf, v8bf
348 (vst4): Changed to VAR13 and added v4bf, v8bf
349 * config/arm/iterators.md (VDXBF): New iterator.
350 (VQ2BF): New iterator.
351 *config/arm/neon.md (neon_vst2<mode>): Used new iterators.
352 (neon_vst2<mode>): Used new iterators.
353 (neon_vst3<mode>): Used new iterators.
354 (neon_vst3<mode>): Used new iterators.
355 (neon_vst3qa<mode>): Used new iterators.
356 (neon_vst3qb<mode>): Used new iterators.
357 (neon_vst4<mode>): Used new iterators.
358 (neon_vst4<mode>): Used new iterators.
359 (neon_vst4qa<mode>): Used new iterators.
360 (neon_vst4qb<mode>): Used new iterators.
362 2020-03-06 Delia Burduv <delia.burduv@arm.com>
364 * config/aarch64/aarch64-simd-builtins.def
365 (bfcvtn): New built-in function.
366 (bfcvtn_q): New built-in function.
367 (bfcvtn2): New built-in function.
368 (bfcvt): New built-in function.
369 * config/aarch64/aarch64-simd.md
370 (aarch64_bfcvtn<q><mode>): New pattern.
371 (aarch64_bfcvtn2v8bf): New pattern.
372 (aarch64_bfcvtbf): New pattern.
373 * config/aarch64/arm_bf16.h (float32_t): New typedef.
374 (vcvth_bf16_f32): New intrinsic.
375 * config/aarch64/arm_bf16.h (vcvt_bf16_f32): New intrinsic.
376 (vcvtq_low_bf16_f32): New intrinsic.
377 (vcvtq_high_bf16_f32): New intrinsic.
378 * config/aarch64/iterators.md (V4SF_TO_BF): New mode iterator.
379 (UNSPEC_BFCVTN): New UNSPEC.
380 (UNSPEC_BFCVTN2): New UNSPEC.
381 (UNSPEC_BFCVT): New UNSPEC.
382 * config/arm/types.md (bf_cvt): New type.
384 2020-03-06 Andreas Krebbel <krebbel@linux.ibm.com>
386 * config/s390/s390.md ("tabort"): Get rid of two consecutive
387 blanks in format string.
389 2020-03-05 H.J. Lu <hongjiu.lu@intel.com>
393 * config/i386/i386-protos.h (ix86_output_ssemov): New prototype.
394 * config/i386/i386.c (ix86_get_ssemov): New function.
395 (ix86_output_ssemov): Likewise.
396 * config/i386/sse.md (VMOVE:mov<mode>_internal): Call
397 ix86_output_ssemov for TYPE_SSEMOV. Remove TARGET_AVX512VL
399 (*movxi_internal_avx512f): Call ix86_output_ssemov for TYPE_SSEMOV.
400 (*movoi_internal_avx): Call ix86_output_ssemov for TYPE_SSEMOV.
401 Remove ext_sse_reg_operand and TARGET_AVX512VL check.
402 (*movti_internal): Likewise.
403 (*movtf_internal): Call ix86_output_ssemov for TYPE_SSEMOV.
405 2020-03-05 Jeff Law <law@redhat.com>
407 PR tree-optimization/91890
408 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Remove LOC argument.
409 Use gimple_or_expr_nonartificial_location.
410 (check_bounds_overlap): Drop LOC argument to maybe_diag_access_bounds.
411 Use gimple_or_expr_nonartificial_location.
412 * gimple.c (gimple_or_expr_nonartificial_location): New function.
413 * gimple.h (gimple_or_expr_nonartificial_location): Declare it.
414 * tree-ssa-strlen.c (maybe_warn_overflow): Use
415 gimple_or_expr_nonartificial_location.
416 (maybe_diag_stxncpy_trunc, handle_builtin_stxncpy_strncat): Likewise.
417 (maybe_warn_pointless_strcmp): Likewise.
419 2020-03-05 Jakub Jelinek <jakub@redhat.com>
422 * config/i386/avx2intrin.h (_mm_mask_i32gather_ps): Fix first cast of
423 SRC and MASK arguments to __m128 from __m128d.
424 (_mm256_mask_i32gather_ps): Fix first cast of MASK argument to __m256
426 (_mm_mask_i64gather_ps): Fix first cast of MASK argument to __m128
428 * config/i386/xopintrin.h (_mm_permute2_pd): Fix first cast of C
429 argument to __m128i from __m128d.
430 (_mm256_permute2_pd): Fix first cast of C argument to __m256i from
432 (_mm_permute2_ps): Fix first cast of C argument to __m128i from __m128.
433 (_mm256_permute2_ps): Fix first cast of C argument to __m256i from
436 2020-03-05 Delia Burduv <delia.burduv@arm.com>
438 * config/arm/arm_neon.h (vbfmmlaq_f32): New.
439 (vbfmlalbq_f32): New.
440 (vbfmlaltq_f32): New.
441 (vbfmlalbq_lane_f32): New.
442 (vbfmlaltq_lane_f32): New.
443 (vbfmlalbq_laneq_f32): New.
444 (vbfmlaltq_laneq_f32): New.
445 * config/arm/arm_neon_builtins.def (vmmla): New.
452 * config/arm/iterators.md (BF_MA): New int iterator.
453 (bt): New int attribute.
454 (VQXBF): Copy of VQX with V8BF.
455 * config/arm/neon.md (neon_vmmlav8bf): New insn.
456 (neon_vfma<bt>v8bf): New insn.
457 (neon_vfma<bt>_lanev8bf): New insn.
458 (neon_vfma<bt>_laneqv8bf): New expand.
459 (neon_vget_high<mode>): Changed iterator to VQXBF.
460 * config/arm/unspecs.md (UNSPEC_BFMMLA): New UNSPEC.
461 (UNSPEC_BFMAB): New UNSPEC.
462 (UNSPEC_BFMAT): New UNSPEC.
464 2020-03-05 Jakub Jelinek <jakub@redhat.com>
467 * tree-pretty-print.h (pretty_print_string): Declare.
468 * tree-pretty-print.c (pretty_print_string): Remove forward
469 declaration, no longer static. Change nbytes parameter type
470 from unsigned to size_t.
471 * print-rtl.c (print_value) <case CONST_STRING>: Use
472 pretty_print_string and for shrink way too long strings.
474 2020-03-05 Richard Biener <rguenther@suse.de>
475 Jakub Jelinek <jakub@redhat.com>
477 PR tree-optimization/93582
478 * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
479 last operand as signed when looking for memset offset. Formatting
482 2020-03-04 Andrew Pinski <apinski@marvell.com>
485 * value-prof.c (dump_histogram_value): Use std::abs.
487 2020-03-04 Martin Sebor <msebor@redhat.com>
489 PR tree-optimization/93986
490 * tree-ssa-strlen.c (maybe_warn_overflow): Convert all wide_int
491 operands to the same precision widest_int to avoid ICEs.
493 2020-03-04 Bill Schmidt <wschmidt@linux.ibm.com>
496 * rs6000-cpus.def (OTHER_ALTIVEC_MASKS): New #define.
497 * rs6000.c (rs6000_disable_incompatible_switches): Add table entry
498 for OPTION_MASK_ALTIVEC.
500 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
502 * config.gcc: Include the glibc-stdint.h header for zTPF.
504 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
506 * config/s390/s390.c (s390_secondary_memory_needed): Disallow
507 direct FPR-GPR copies.
508 (s390_register_info_gprtofpr): Disallow GPR content to be saved in
511 2020-03-04 Andreas Krebbel <krebbel@linux.ibm.com>
513 * config/s390/s390.c (s390_emit_prologue): Specify the 2 new
514 operands to the prologue_tpf expander.
515 (s390_emit_epilogue): Likewise.
516 (s390_option_override_internal): Do error checking and setup for
518 * config/s390/tpf.h (TPF_TRACE_PROLOGUE_CHECK)
519 (TPF_TRACE_EPILOGUE_CHECK, TPF_TRACE_PROLOGUE_TARGET)
520 (TPF_TRACE_EPILOGUE_TARGET, TPF_TRACE_PROLOGUE_SKIP_TARGET)
521 (TPF_TRACE_EPILOGUE_SKIP_TARGET): New macro definitions.
522 * config/s390/tpf.md ("prologue_tpf", "epilogue_tpf"): Add two new
523 operands for the check flag and the branch target.
524 * config/s390/tpf.opt ("mtpf-trace-hook-prologue-check")
525 ("mtpf-trace-hook-prologue-target")
526 ("mtpf-trace-hook-epilogue-check")
527 ("mtpf-trace-hook-epilogue-target", "mtpf-trace-skip"): New
529 * doc/invoke.texi: Document -mtpf-trace-skip option. The other
530 options are for debugging purposes and will not be documented
533 2020-03-04 Jakub Jelinek <jakub@redhat.com>
536 * tree-inline.c (copy_decl_to_var): Copy DECL_BY_REFERENCE flag.
538 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Add offseti
539 argument. Change pd argument so that it can be modified. Turn
540 constant non-CONSTRUCTOR store into non-constant if it is too large.
541 Adjust offset and size of CONSTRUCTOR or non-constant store to avoid
543 (vn_walk_cb_data::vn_walk_cb_data, vn_reference_lookup_3): Adjust
546 2020-02-04 Richard Biener <rguenther@suse.de>
548 PR tree-optimization/93964
549 * graphite-isl-ast-to-gimple.c
550 (gcc_expression_from_isl_ast_expr_id): Add intermediate
551 conversion for pointer to integer converts.
552 * graphite-scop-detection.c (assign_parameter_index_in_region):
555 2020-03-04 Martin Liska <mliska@suse.cz>
559 * doc/invoke.texi: Clarify --help=language and --help=common
562 2020-03-04 Jakub Jelinek <jakub@redhat.com>
564 PR tree-optimization/94001
565 * tree-tailcall.c (process_assignment): Before comparing op1 to
566 *ass_var, verify *ass_var is non-NULL.
568 2020-03-04 Kito Cheng <kito.cheng@sifive.com>
571 * config/riscv/riscv.c (riscv_emit_float_compare): Using NE to compare
574 2020-03-03 Dennis Zhang <dennis.zhang@arm.com>
576 * config/arm/arm_bf16.h (vcvtah_f32_bf16, vcvth_bf16_f32): New.
577 * config/arm/arm_neon.h (vcvt_f32_bf16, vcvtq_low_f32_bf16): New.
578 (vcvtq_high_f32_bf16, vcvt_bf16_f32): New.
579 (vcvtq_low_bf16_f32, vcvtq_high_bf16_f32): New.
580 * config/arm/arm_neon_builtins.def (vbfcvt, vbfcvt_high): New entries.
581 (vbfcvtv4sf, vbfcvtv4sf_high): Likewise.
582 * config/arm/iterators.md (VBFCVT, VBFCVTM): New mode iterators.
583 (V_bf_low, V_bf_cvt_m): New mode attributes.
584 * config/arm/neon.md (neon_vbfcvtv4sf<VBFCVT:mode>): New.
585 (neon_vbfcvtv4sf_highv8bf, neon_vbfcvtsf): New.
586 (neon_vbfcvt<VBFCVT:mode>, neon_vbfcvt_highv8bf): New.
587 (neon_vbfcvtbf_cvtmode<mode>, neon_vbfcvtbf): New
588 * config/arm/unspecs.md (UNSPEC_BFCVT, UNSPEC_BFCVT_HIG): New.
590 2020-03-03 Jakub Jelinek <jakub@redhat.com>
592 PR tree-optimization/93582
593 * tree-ssa-sccvn.h (vn_reference_lookup): Add mask argument.
594 * tree-ssa-sccvn.c (struct vn_walk_cb_data): Add mask and masked_result
595 members, initialize them in the constructor and if mask is non-NULL,
596 artificially push_partial_def {} for the portions of the mask that
598 (vn_walk_cb_data::finish): If mask is non-NULL, set masked_result to
599 val and return (void *)-1. Formatting fix.
600 (vn_reference_lookup_pieces): Adjust vn_walk_cb_data initialization.
602 (vn_reference_lookup): Add mask argument. If non-NULL, don't call
603 fully_constant_vn_reference_p nor vn_reference_lookup_1 and return
605 (visit_nary_op): Handle BIT_AND_EXPR of a memory load and INTEGER_CST
607 (visit_stmt): Formatting fix.
609 2020-03-03 Richard Biener <rguenther@suse.de>
611 PR tree-optimization/93946
612 * alias.h (refs_same_for_tbaa_p): Declare.
613 * alias.c (refs_same_for_tbaa_p): New function.
614 * tree-ssa-alias.c (ao_ref_alias_set): For a NULL ref return
616 * tree-ssa-scopedtables.h
617 (avail_exprs_stack::lookup_avail_expr): Add output argument
618 giving access to the hashtable entry.
619 * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
621 * tree-ssa-dom.c: Include alias.h.
622 (dom_opt_dom_walker::optimize_stmt): Validate TBAA state before
623 removing redundant store.
624 * tree-ssa-sccvn.h (vn_reference_s::base_set): New member.
625 (ao_ref_init_from_vn_reference): Adjust prototype.
626 (vn_reference_lookup_pieces): Likewise.
627 (vn_reference_insert_pieces): Likewise.
628 * tree-ssa-sccvn.c: Track base alias set in addition to alias
630 (eliminate_dom_walker::eliminate_stmt): Also check base alias
631 set when removing redundant stores.
632 (visit_reference_op_store): Likewise.
633 * dse.c (record_store): Adjust valdity check for redundant
636 2020-03-03 Jakub Jelinek <jakub@redhat.com>
639 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
641 PR rtl-optimization/94002
642 * explow.c (plus_constant): Punt if cst has VOIDmode and
643 get_pool_mode is different from mode.
645 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
647 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
648 address has an offset which fits the scalling constraint for a
649 load/store operation.
650 (legitimate_scaled_address_p): Update use
651 leigitimate_small_data_address_p.
652 (arc_print_operand): Likewise.
653 (arc_legitimate_address_p): Likewise.
654 (legitimate_small_data_address_p): Likewise.
656 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
658 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
659 (fnmasf4_fpu): Likewise.
661 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
663 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
666 (adddi3_i): Remove pattern.
667 (subdi3_i): Likewise.
669 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
671 * config/arc/arc.md (eh_return): Add length info.
673 2020-03-02 David Malcolm <dmalcolm@redhat.com>
675 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
677 2020-03-02 David Malcolm <dmalcolm@redhat.com>
679 * doc/invoke.texi (Static Analyzer Options): Add
680 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
683 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
686 * config/i386/i386.md (movstrict<mode>): Allow only
687 registers with VALID_INT_MODE_P modes.
689 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
691 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
692 (reduc_insn): Use 'U' and 'B' operand codes.
693 (reduc_<reduc_op>_scal_<mode>): Allow all types.
694 (reduc_<reduc_op>_scal_v64di): Delete.
695 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
696 (*plus_carry_dpp_shr_v64si): Change to ...
697 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
698 (mov_from_lane63_v64di): Change to ...
699 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
700 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
701 Support UNSPEC_MOV_DPP_SHR output formats.
702 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
703 Add "use_extends" reductions.
704 (print_operand_address): Add 'I' and 'U' codes.
705 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
707 2020-03-02 Martin Liska <mliska@suse.cz>
709 * lto-wrapper.c: Fix typo in comment about
710 C++ standard version.
712 2020-03-01 Martin Sebor <msebor@redhat.com>
715 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
717 2020-03-01 Martin Sebor <msebor@redhat.com>
720 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
721 of a pointer in the outermost ADDR_EXPRs.
723 2020-02-28 Jeff Law <law@redhat.com>
725 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
726 * config/v850/v850.c (v850_asm_trampoline_template): Update
729 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
732 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
735 2020-02-28 Martin Liska <mliska@suse.cz>
738 * configure.ac: Improve detection of ld_date by requiring
739 either two dashes or none.
740 * configure: Regenerate.
742 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
744 PR rtl-optimization/93564
745 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
746 do not honor reg alloc order.
748 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
751 * config/aarch64/aarch64.c (aarch64_override_options): Fix
752 misleading warning string.
754 2020-02-27 Martin Sebor <msebor@redhat.com>
756 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
758 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
761 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
762 Split the insn into two parts. This insn only does variable
763 extract from a register.
764 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
765 variable extract from memory.
766 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
767 only does variable extract from a register.
768 (vsx_extract_v4sf_var_load): New insn, do variable extract from
770 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
771 into two parts. This insn only does variable extract from a
773 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
774 do variable extract from memory.
776 2020-02-27 Martin Jambor <mjambor@suse.cz>
777 Feng Xue <fxue@os.amperecomputing.com>
780 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
781 new function calls_same_node_or_its_all_contexts_clone_p.
782 (cgraph_edge_brings_value_p): Use it.
783 (cgraph_edge_brings_value_p): Likewise.
784 (self_recursive_pass_through_p): Return false if caller is a clone.
785 (self_recursive_agg_pass_through_p): Likewise.
787 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
790 * alias.c (ends_tbaa_access_path_p): Break out from ...
791 (component_uses_parent_alias_set_from): ... here.
792 * alias.h (ends_tbaa_access_path_p): Declare.
793 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
794 handle trailing arrays past end of tbaa access path.
795 (aliasing_component_refs_p): ... here; likewise.
796 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
797 path; disambiguate also past end of it.
798 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
801 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
803 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
804 beginning of the file.
805 (vcreate_bf16, vcombine_bf16): New.
806 (vdup_n_bf16, vdupq_n_bf16): New.
807 (vdup_lane_bf16, vdup_laneq_bf16): New.
808 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
809 (vduph_lane_bf16, vduph_laneq_bf16): New.
810 (vset_lane_bf16, vsetq_lane_bf16): New.
811 (vget_lane_bf16, vgetq_lane_bf16): New.
812 (vget_high_bf16, vget_low_bf16): New.
813 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
814 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
815 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
816 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
817 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
818 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
819 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
820 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
821 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
822 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
823 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
824 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
825 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
826 (vreinterpretq_bf16_p128): New.
827 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
828 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
829 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
830 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
831 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
832 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
833 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
834 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
835 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
836 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
837 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
838 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
839 (vreinterpretq_p128_bf16): New.
840 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
842 (V_elem_l): Likewise.
845 (V_DOUBLE): Likewise.
846 (VDQX): Add V4BF and V8BF.
847 (V_two_elem, V_three_elem, V_four_elem): Likewise.
850 (V_double_vector_mode): Likewise.
851 (V_cmp_result): Likewise.
852 (V_uf_sclr): Likewise.
853 (V_sz_elem): Likewise.
854 (Is_d_reg): Likewise.
855 (V_mode_nunits): Likewise.
856 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
858 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
860 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
861 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
862 (<expander><mode>3<exec>): Likewise.
863 (<expander><mode>3): New.
864 (v<expander><mode>3): New.
865 (<expander><mode>3): New.
866 (<expander><mode>3<exec>): Rename to ...
867 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
868 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
870 2020-02-27 Alexandre Oliva <oliva@adacore.com>
872 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
875 2020-02-27 Richard Biener <rguenther@suse.de>
877 PR tree-optimization/93508
878 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
879 non-_CHK variants. Valueize their length arguments.
881 2020-02-27 Richard Biener <rguenther@suse.de>
883 PR tree-optimization/93953
884 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
885 to the hash-map entry.
887 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
889 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
891 2020-02-27 Mark Williams <mwilliams@fb.com>
893 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
894 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
895 -ffile-prefix-map and -fmacro-prefix-map.
896 * lto-streamer-out.c: Include file-prefix-map.h.
897 (lto_output_location): Remap the file part of locations.
899 2020-02-27 Jakub Jelinek <jakub@redhat.com>
902 * gimplify.c (gimplify_init_constructor): Don't promote readonly
903 DECL_REGISTER variables to TREE_STATIC.
905 PR tree-optimization/93582
906 PR tree-optimization/93945
907 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
908 non-zero INTEGER_CST second argument and ref->offset or ref->size
909 not a multiple of BITS_PER_UNIT.
911 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
913 * doc/install.texi (Binaries): Update description of BullFreeware.
915 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
919 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
920 C++ Language Options, Warning Options, and Static Analyzer
921 Options lists. Document negative form of options enabled by
922 default. Move some things around to more accurately sort
923 warnings by category.
924 (C++ Dialect Options, Warning Options, Static Analyzer
925 Options): Document negative form of options when enabled by
926 default. Move some things around to more accurately sort
927 warnings by category. Add some missing index entries.
930 2020-02-26 Carl Love <cel@us.ibm.com>
933 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
934 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
935 for the vector unsigned short arguments. It is also listed as the
936 name of the built-in for arguments vector unsigned short,
937 vector unsigned int and vector unsigned long long built-ins. The
938 name of the builtins for these arguments should be:
939 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
940 __builtin_crypto_vpmsumd respectively.
942 2020-02-26 Richard Biener <rguenther@suse.de>
944 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
945 and load permutation.
947 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
950 * optabs-tree.c (supportable_convert_operation): Reject types with
953 2020-02-26 David Malcolm <dmalcolm@redhat.com>
955 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
957 2020-02-26 Jakub Jelinek <jakub@redhat.com>
959 PR tree-optimization/93820
960 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
961 argument to ALL_INTEGER_CST_P boolean.
962 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
963 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
964 adjacent INTEGER_CST store into merged_store->only_constants like
967 2020-02-25 Jakub Jelinek <jakub@redhat.com>
970 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
972 * cfghooks.c (verify_flow_info): Likewise.
973 * predict.c (combine_predictions_for_bb): Likewise.
974 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
975 sucessor -> successor.
976 (find_traces_1_round): Fix comment typo, destinarion -> destination.
977 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
979 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
980 message typo, sucessors -> successors.
982 2020-02-25 Martin Sebor <msebor@redhat.com>
984 * doc/extend.texi (attribute access): Correct an example.
986 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
988 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
990 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
992 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
993 (VD): Enable for V4BF.
995 (VQ): Enable for V8BF.
998 (VDBL, Vdbl): Add V4BF.
999 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
1000 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
1001 (bfloat16x8x2_t): Likewise.
1002 (bfloat16x4x3_t): Likewise.
1003 (bfloat16x8x3_t): Likewise.
1004 (bfloat16x4x4_t): Likewise.
1005 (bfloat16x8x4_t): Likewise.
1006 (vcombine_bf16): New.
1007 (vld1_bf16, vld1_bf16_x2): New.
1008 (vld1_bf16_x3, vld1_bf16_x4): New.
1009 (vld1q_bf16, vld1q_bf16_x2): New.
1010 (vld1q_bf16_x3, vld1q_bf16_x4): New.
1011 (vld1_lane_bf16): New.
1012 (vld1q_lane_bf16): New.
1013 (vld1_dup_bf16): New.
1014 (vld1q_dup_bf16): New.
1017 (vld2_dup_bf16): New.
1018 (vld2q_dup_bf16): New.
1021 (vld3_dup_bf16): New.
1022 (vld3q_dup_bf16): New.
1025 (vld4_dup_bf16): New.
1026 (vld4q_dup_bf16): New.
1027 (vst1_bf16, vst1_bf16_x2): New.
1028 (vst1_bf16_x3, vst1_bf16_x4): New.
1029 (vst1q_bf16, vst1q_bf16_x2): New.
1030 (vst1q_bf16_x3, vst1q_bf16_x4): New.
1031 (vst1_lane_bf16): New.
1032 (vst1q_lane_bf16): New.
1040 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
1042 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
1043 (VALL_F16): Likewise.
1044 (VALLDI_F16): Likewise.
1047 (vswap_width_name): Likewise.
1048 (VSWAP_WIDTH): Likewise.
1052 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
1053 (vget_lane_bf16, vgetq_lane_bf16): New.
1054 (vcreate_bf16): New.
1055 (vdup_n_bf16, vdupq_n_bf16): New.
1056 (vdup_lane_bf16, vdup_laneq_bf16): New.
1057 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
1058 (vduph_lane_bf16, vduph_laneq_bf16): New.
1059 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
1060 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
1061 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
1062 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
1063 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
1064 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
1065 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
1066 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
1067 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
1068 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
1069 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
1070 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
1071 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
1072 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
1073 (vreinterpretq_bf16_p128): New.
1074 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
1075 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
1076 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
1077 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
1078 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
1079 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
1080 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
1081 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
1082 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
1083 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
1084 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
1085 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
1086 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
1087 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
1088 (vreinterpretq_p128_bf16): New.
1090 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
1092 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
1093 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
1094 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
1095 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
1096 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
1097 * config/arm/iterators.md (VSF2BF): New attribute.
1098 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
1099 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
1100 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
1102 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
1104 * config/arm/arm.md (required_for_purecode): New attribute.
1105 (enabled): Handle required_for_purecode.
1106 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
1107 work with -mpure-code.
1109 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1111 PR rtl-optimization/93908
1112 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
1115 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
1117 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
1119 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
1121 * doc/install.texi (--enable-checking): Adjust wording.
1123 2020-02-25 Richard Biener <rguenther@suse.de>
1125 PR tree-optimization/93868
1126 * tree-vect-slp.c (slp_copy_subtree): New function.
1127 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
1128 re-arranging stmts in it.
1130 2020-02-25 Jakub Jelinek <jakub@redhat.com>
1133 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
1134 dummy function and remove it at the end.
1136 PR translation/93864
1137 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
1138 paramter -> parameter.
1139 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
1140 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
1142 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
1144 * doc/install.texi (--enable-checking): Properly document current
1146 (--enable-stage1-checking): Minor clarification about bootstrap.
1148 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1151 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
1152 -fanalyzer-checker=taint is also required.
1153 (-fanalyzer-checker=): Note that providing this option enables the
1154 given checker, and doing so may be required for checkers that are
1155 disabled by default.
1157 2020-02-24 David Malcolm <dmalcolm@redhat.com>
1159 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
1160 significant control flow events; add a "3" which shows all
1161 control flow events; the old "3" becomes "4".
1163 2020-02-24 Jakub Jelinek <jakub@redhat.com>
1165 PR tree-optimization/93582
1166 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
1167 pd.offset and pd.size to be counted in bits rather than bytes, add
1168 support for maxsizei that is not a multiple of BITS_PER_UNIT and
1169 handle bitfield stores and loads.
1170 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
1171 uncomparable quantities - bytes vs. bits. Allow push_partial_def
1172 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
1173 pd.offset/pd.size to be counted in bits rather than bytes.
1174 Formatting fix. Rename shadowed len variable to buflen.
1176 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1177 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
1180 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
1181 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
1182 * opts-common.c (parse_options_from_collect_gcc_options): New function.
1183 (prepend_xassembler_to_collect_as_options): Likewise.
1184 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
1185 (prepend_xassembler_to_collect_as_options): Likewise.
1186 * lto-opts.c (lto_write_options): Stream assembler options
1187 in COLLECT_AS_OPTIONS.
1188 * lto-wrapper.c (xassembler_options_error): New static variable.
1189 (get_options_from_collect_gcc_options): Move parsing options code to
1190 parse_options_from_collect_gcc_options and call it.
1191 (merge_and_complain): Validate -Xassembler options.
1192 (append_compiler_options): Handle OPT_Xassembler.
1193 (run_gcc): Append command line -Xassembler options to
1194 collect_gcc_options.
1195 * doc/invoke.texi: Add documentation about using Xassembler
1198 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
1200 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
1202 (riscv_rtx_costs): Update cost model for LTGT.
1204 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
1206 PR rtl-optimization/93564
1207 * ira-color.c (struct update_cost_queue_elem): New member start.
1208 (queue_update_cost, get_next_update_cost): Add new arg start.
1209 (allocnos_conflict_p): New function.
1210 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
1211 Add checking conflicts with allocnos_conflict_p.
1212 (update_costs_from_prefs, restore_costs_from_copies): Adjust
1213 update_costs_from_allocno calls.
1214 (update_conflict_hard_regno_costs): Add checking conflicts with
1215 allocnos_conflict_p. Adjust calls of queue_update_cost and
1216 get_next_update_cost.
1217 (assign_hard_reg): Adjust calls of queue_update_cost. Add
1219 (bucket_allocno_compare_func): Restore previous version.
1221 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
1223 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
1224 double-word size when handling aggregate return values.
1225 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
1226 that homogeneous SFmode and DFmode aggregates are passed and returned
1227 in general registers.
1229 2020-02-21 Jakub Jelinek <jakub@redhat.com>
1231 PR translation/93759
1232 * opts.c (print_filtered_help): Translate help before appending
1233 messages to it rather than after that.
1235 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1237 PR rtl-optimization/PR92989
1238 * lra-lives.c (process_bb_lives): Restore the original order
1239 of the bb liveness update. Call make_hard_regno_dead for each
1240 register clobbered at the start of an EH receiver.
1242 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
1245 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
1246 self-recursively generated.
1248 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
1251 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
1254 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
1256 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
1257 Document new target supports option.
1259 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
1261 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
1262 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
1263 * config/arm/iterators.md (MATMUL): New iterator.
1264 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
1265 (mmla_sfx): New attribute.
1266 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
1267 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
1268 (UNSPEC_MATMUL_US): New.
1270 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1272 * config/arm/arm.md: Prevent scalar shifts from being used when big
1275 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
1276 Richard Biener <rguenther@suse.de>
1278 PR tree-optimization/93586
1279 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
1280 after mismatched array refs; do not sure type size information to
1281 recover from unmatched referneces with !flag_strict_aliasing_p.
1283 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1285 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
1286 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
1287 (scatter_store<mode>): Rename to ...
1288 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
1289 (scatter<mode>_exec): Delete. Move contents ...
1290 (mask_scatter_store<mode>): ... here, and rename that to ...
1291 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
1292 Remove mode conversion.
1293 (mask_gather_load<mode>): Rename to ...
1294 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
1295 Remove mode conversion.
1296 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
1298 2020-02-21 Martin Jambor <mjambor@suse.cz>
1300 PR tree-optimization/93845
1301 * tree-sra.c (verify_sra_access_forest): Only test access size of
1304 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1306 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
1307 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
1308 (addv64di3_exec): Likewise.
1309 (subv64di3): Likewise.
1310 (subv64di3_exec): Likewise.
1311 (addv64di3_zext): Likewise.
1312 (addv64di3_zext_exec): Likewise.
1313 (addv64di3_zext_dup): Likewise.
1314 (addv64di3_zext_dup_exec): Likewise.
1315 (addv64di3_zext_dup2): Likewise.
1316 (addv64di3_zext_dup2_exec): Likewise.
1317 (addv64di3_sext_dup2): Likewise.
1318 (addv64di3_sext_dup2_exec): Likewise.
1319 (<expander>v64di3): Likewise.
1320 (<expander>v64di3_exec): Likewise.
1321 (*<reduc_op>_dpp_shr_v64di): Likewise.
1322 (*plus_carry_dpp_shr_v64di): Likewise.
1323 * config/gcn/gcn.md (adddi3): Likewise.
1324 (addptrdi3): Likewise.
1325 (<expander>di3): Likewise.
1327 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
1329 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
1331 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1333 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
1334 support. Use aarch64_emit_mult instead of emitting multiplication
1335 instructions directly.
1336 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
1337 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
1339 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1341 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
1342 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
1343 instead of emitting multiplication instructions directly.
1344 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
1345 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
1346 (@aarch64_frecps<mode>): New expanders.
1348 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1350 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
1351 on and produce uint64_ts rather than ints.
1352 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
1353 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
1355 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1357 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
1358 an unused xmsk register when handling approximate rsqrt.
1360 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
1362 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
1363 flag_finite_math_only condition.
1365 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
1368 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
1369 to destination operand for shufps alternative.
1370 (*vec_extractv2si_1): Ditto.
1372 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
1375 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
1378 2020-02-20 Martin Liska <mliska@suse.cz>
1380 PR translation/93831
1381 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
1383 2020-02-20 Martin Liska <mliska@suse.cz>
1385 PR translation/93830
1386 * common/config/avr/avr-common.c: Remote trailing "|".
1388 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1390 * collect2.c (maybe_run_lto_and_relink): Fix typo in
1393 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
1395 PR tree-optimization/93767
1396 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
1397 access-size bias from the offset calculations for negative strides.
1399 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
1401 * collect2.c (c_file, o_file): Make const again.
1402 (ldout,lderrout, dump_ld_file): Remove.
1403 (tool_cleanup): Avoid calling not signal-safe functions.
1404 (maybe_run_lto_and_relink): Avoid possible signal handler
1405 access to unintialzed memory (lto_o_files).
1406 (main): Avoid leaking temp files in $TMPDIR.
1407 Initialize c_file/o_file with concat, which avoids exposing
1408 uninitialized memory to signal handler, which calls unlink(!).
1409 Avoid calling maybe_unlink when the main function returns,
1410 since the atexit handler is already doing this.
1411 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
1413 2020-02-19 Martin Jambor <mjambor@suse.cz>
1415 PR tree-optimization/93776
1416 * tree-sra.c (create_access): Do not create zero size accesses.
1417 (get_access_for_expr): Do not search for zero sized accesses.
1419 2020-02-19 Martin Jambor <mjambor@suse.cz>
1421 PR tree-optimization/93667
1422 * tree-sra.c (scalarizable_type_p): Return false if record fields
1423 do not follow wach other.
1425 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1427 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
1428 rather than fmv.x.s/fmv.s.x.
1430 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
1432 * config/aarch64/aarch64-simd-builtins.def
1433 (intrinsic_vec_smult_lo_): New.
1434 (intrinsic_vec_umult_lo_): Likewise.
1435 (vec_widen_smult_hi_): Likewise.
1436 (vec_widen_umult_hi_): Likewise.
1437 * config/aarch64/aarch64-simd.md
1438 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
1439 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
1440 (vmull_high_s16): Likewise.
1441 (vmull_high_s32): Likewise.
1442 (vmull_high_u8): Likewise.
1443 (vmull_high_u16): Likewise.
1444 (vmull_high_u32): Likewise.
1445 (vmull_s8): Likewise.
1446 (vmull_s16): Likewise.
1447 (vmull_s32): Likewise.
1448 (vmull_u8): Likewise.
1449 (vmull_u16): Likewise.
1450 (vmull_u32): Likewise.
1452 2020-02-18 Martin Liska <mliska@suse.cz>
1454 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
1455 bootstrap by missing removal of invalid sanity check.
1457 2020-02-18 Martin Liska <mliska@suse.cz>
1460 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
1461 Always compare LHS of gimple_assign.
1463 2020-02-18 Martin Liska <mliska@suse.cz>
1466 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
1467 and return type of functions.
1468 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
1469 Drop MALLOC attribute for void functions.
1470 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
1471 malloc_state for a new VOID clone.
1473 2020-02-18 Martin Liska <mliska@suse.cz>
1476 * common.opt: Add -fprofile-reproducibility.
1477 * doc/invoke.texi: Document it.
1478 * value-prof.c (dump_histogram_value):
1479 Document and support behavior for counters[0]
1480 being a negative value.
1481 (get_nth_most_common_value): Handle negative
1482 counters[0] in respect to flag_profile_reproducible.
1484 2020-02-18 Jakub Jelinek <jakub@redhat.com>
1487 * cgraph.c (verify_speculative_call): Use speculative_id instead of
1488 speculative_uid in messages. Remove trailing whitespace from error
1489 message. Use num_speculative_call_targets instead of
1490 num_speculative_targets in a message.
1491 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
1492 edge messages and stmt instead of cal_stmt in reference message.
1494 PR tree-optimization/93780
1495 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
1496 before calling build_vector_type.
1497 (execute_update_addresses_taken): Likewise.
1500 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
1501 typo, functoin -> function.
1502 * tree.c (free_lang_data_in_decl): Fix comment typo,
1503 functoin -> function.
1504 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1506 2020-02-17 David Malcolm <dmalcolm@redhat.com>
1508 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
1510 (print_option_information): Don't call get_option_url if URLs
1513 2020-02-17 Alexandre Oliva <oliva@adacore.com>
1515 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
1516 handling of register_common-less targets.
1518 2020-02-17 Martin Liska <mliska@suse.cz>
1521 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
1523 2020-02-17 Martin Liska <mliska@suse.cz>
1525 PR translation/93755
1526 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1529 2020-02-17 Martin Liska <mliska@suse.cz>
1532 * config/rx/elf.opt: Fix typo.
1534 2020-02-17 Richard Biener <rguenther@suse.de>
1537 * opts-global.c (print_ignored_options): Use inform and
1540 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
1543 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
1545 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
1548 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
1549 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
1551 2020-02-15 Jason Merrill <jason@redhat.com>
1553 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
1555 2020-02-15 Jakub Jelinek <jakub@redhat.com>
1557 PR tree-optimization/93744
1558 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
1559 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
1560 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
1561 sure @2 in the first and @1 in the other patterns has no side-effects.
1563 2020-02-15 David Malcolm <dmalcolm@redhat.com>
1564 Bernd Edlinger <bernd.edlinger@hotmail.de>
1568 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
1569 * configure.ac (--with-diagnostics-urls): New configuration
1570 option, based on --with-diagnostics-color.
1571 (DIAGNOSTICS_URLS_DEFAULT): New define.
1572 * config.h: Regenerate.
1573 * configure: Regenerate.
1574 * diagnostic.c (diagnostic_urls_init): Handle -1 for
1575 DIAGNOSTICS_URLS_DEFAULT from configure-time
1576 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
1577 and TERM_URLS environment variable.
1578 * diagnostic-url.h (diagnostic_url_format): New enum type.
1579 (diagnostic_urls_enabled_p): rename to...
1580 (determine_url_format): ... this, and change return type.
1581 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
1582 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
1583 the linux console, and mingw.
1584 (diagnostic_urls_enabled_p): rename to...
1585 (determine_url_format): ... this, and adjust.
1586 * pretty-print.h (pretty_printer::show_urls): rename to...
1587 (pretty_printer::url_format): ... this, and change to enum.
1588 * pretty-print.c (pretty_printer::pretty_printer,
1589 pp_begin_url, pp_end_url, test_urls): Adjust.
1590 * doc/install.texi (--with-diagnostics-urls): Document the new
1591 configuration option.
1592 (--with-diagnostics-color): Document the existing interaction
1593 with GCC_COLORS better.
1594 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
1595 vindex reference. Update description of defaults based on the above.
1596 (-fdiagnostics-color): Update description of how -fdiagnostics-color
1597 interacts with GCC_COLORS.
1599 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
1602 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
1603 conjunction with TARGET_GNU_TLS in early return.
1605 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
1607 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
1608 the mode is not wider than UNITS_PER_WORD.
1610 2020-02-14 Martin Jambor <mjambor@suse.cz>
1612 PR tree-optimization/93516
1613 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
1614 access of the same type as the parent.
1615 (propagate_subaccesses_from_lhs): Likewise.
1617 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
1620 * config/i386/avx512vbmi2intrin.h
1621 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
1622 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
1623 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1624 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
1625 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
1626 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
1627 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1628 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
1629 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
1630 of lacking a closing parenthesis.
1631 * config/i386/avx512vbmi2vlintrin.h
1632 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
1633 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
1634 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1635 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1636 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1637 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1638 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1639 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1640 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1641 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1642 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1643 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1644 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1645 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1646 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1647 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1648 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1649 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1651 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1654 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1655 the target function entry.
1657 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1659 * common/config/arc/arc-common.c (arc_option_optimization_table):
1660 Disable if-conversion step when optimized for size.
1662 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1664 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1665 R12-R15 are always in ARCOMPACT16_REGS register class.
1666 * config/arc/arc.opt (mq-class): Deprecate.
1667 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1669 * doc/invoke.texi (mq-class): Update text.
1670 * common/config/arc/arc-common.c (arc_option_optimization_table):
1673 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1675 * config/arc/arc.c (arc_insn_cost): New function.
1676 (TARGET_INSN_COST): Define.
1677 * config/arc/arc.md (cost): New attribute.
1678 (add_n): Use arc_nonmemory_operand.
1679 (ashlsi3_insn): Likewise, also update constraints.
1680 (ashrsi3_insn): Likewise.
1681 (rotrsi3): Likewise.
1682 (add_shift): Likewise.
1683 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1685 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1687 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1689 (umulsidi_600): Likewise.
1691 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1694 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1695 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1696 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1697 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1698 pass __A to the builtin followed by __W instead of __A followed by
1700 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1701 _mm512_mask_popcnt_epi64): Likewise.
1702 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1703 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1704 _mm256_mask_popcnt_epi64): Likewise.
1706 PR tree-optimization/93582
1707 * fold-const.h (shift_bytes_in_array_left,
1708 shift_bytes_in_array_right): Declare.
1709 * fold-const.c (shift_bytes_in_array_left,
1710 shift_bytes_in_array_right): New function, moved from
1711 gimple-ssa-store-merging.c, no longer static.
1712 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1713 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1714 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1715 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1716 shift_bytes_in_array.
1717 (verify_shift_bytes_in_array): Rename to ...
1718 (verify_shift_bytes_in_array_left): ... this. Use
1719 shift_bytes_in_array_left instead of shift_bytes_in_array.
1720 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1721 instead of verify_shift_bytes_in_array.
1722 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1723 / native_interpret_expr where the store covers all needed bits,
1724 punt on PDP-endian, otherwise allow all involved offsets and sizes
1725 not to be byte-aligned.
1728 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1729 use const_0_to_255_operand predicate instead of immediate_operand.
1730 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1731 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1732 vgf2p8affineinvqb_<mode><mask_name>,
1733 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1734 const_0_to_255_operand predicated operands.
1736 2020-02-12 Jeff Law <law@redhat.com>
1738 * config/h8300/h8300.md (comparison shortening peepholes): Use
1739 a mode iterator to merge the HImode and SImode peepholes.
1741 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1744 * real.c (is_even): Make static. Function comment fix.
1745 (is_halfway_below): Make static, don't assert R is not inf/nan,
1746 instead return false for those. Small formatting fixes.
1748 2020-02-12 Martin Sebor <msebor@redhat.com>
1751 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1752 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1753 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1754 (strlen_check_and_optimize_call): Adjust callee name.
1756 2020-02-12 Jeff Law <law@redhat.com>
1758 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1759 (and (xor)) variant. Combine other two into single peephole.
1761 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1763 PR rtl-optimization/93565
1764 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1766 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1768 * config/aarch64/aarch64-simd.md
1769 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1770 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1771 generating separate ADDV and zero_extend patterns.
1772 * config/aarch64/iterators.md (VDQV_E): New iterator.
1774 2020-02-12 Jeff Law <law@redhat.com>
1776 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1777 expanders, splits, etc.
1778 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1779 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1780 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1781 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1782 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1783 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1785 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1787 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1790 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1791 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1792 TARGET_AVX512DQ from condition.
1793 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1794 instead of <mask_mode512bit_condition> in condition. If
1795 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1797 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1800 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1803 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1805 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1807 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1808 where strlen is more legible.
1809 (rs6000_builtin_vectorized_libmass): Ditto.
1810 (rs6000_print_options_internal): Ditto.
1812 2020-02-11 Martin Sebor <msebor@redhat.com>
1814 PR tree-optimization/93683
1815 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1817 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1819 * config/rs6000/predicates.md (cint34_operand): Rename the
1820 -mprefixed-addr option to be -mprefixed.
1821 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1822 the -mprefixed-addr option to be -mprefixed.
1823 (OTHER_FUTURE_MASKS): Likewise.
1824 (POWERPC_MASKS): Likewise.
1825 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1826 the -mprefixed-addr option to be -mprefixed. Change error
1827 messages to refer to -mprefixed.
1828 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1830 (rs6000_legitimate_offset_address_p): Likewise.
1831 (rs6000_mode_dependent_address): Likewise.
1832 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1833 "-mprefixed" for target attributes and pragmas.
1834 (address_to_insn_form): Rename the -mprefixed-addr option to be
1836 (rs6000_adjust_insn_length): Likewise.
1837 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1838 -mprefixed-addr option to be -mprefixed.
1839 (ASM_OUTPUT_OPCODE): Likewise.
1840 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1841 -mprefixed-addr option to be -mprefixed.
1842 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1843 -mprefixed-addr option to be prefixed. Change the option from
1844 being undocumented to being documented.
1845 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1846 -mprefixed option. Update the -mpcrel documentation to mention
1849 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1851 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1852 including FIRST_PSEUDO_REGISTER - 1.
1853 * ira-color.c (print_hard_reg_set): Ditto.
1855 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1857 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1858 (USTERNOP_QUALIFIERS): New define.
1859 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1860 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1861 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1862 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1863 * config/arm/arm_neon.h (vusdot_s32): New.
1864 (vusdot_lane_s32): New.
1865 (vusdotq_lane_s32): New.
1866 (vsudot_lane_s32): New.
1867 (vsudotq_lane_s32): New.
1868 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1869 * config/arm/iterators.md (DOTPROD_I8MM): New.
1870 (sup, opsuffix): Add <us/su>.
1871 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1872 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1874 2020-02-11 Richard Biener <rguenther@suse.de>
1876 PR tree-optimization/93661
1877 PR tree-optimization/93662
1878 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1880 * tree-sra.c (get_access_for_expr): Likewise.
1882 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1885 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1886 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1887 Change condition from TARGET_AVX2 to TARGET_AVX.
1889 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1892 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1893 argument of strncmp.
1895 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1897 Try to generate zero-based comparisons.
1898 * config/cris/cris.c (cris_reduce_compare): New function.
1899 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1900 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1901 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1903 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1906 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1907 in Thumb state and also as a destination in Arm state. Add T16
1910 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1912 * md.texi (Define Subst): Match closing paren in example.
1914 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1918 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1919 arguments of strncmp.
1921 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1924 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1925 but different source value.
1926 (adjust_callers_for_value_intersection): New function.
1927 (gather_edges_for_value): Adjust order of callers to let a
1928 non-self-recursive caller be the first element.
1929 (self_recursive_pass_through_p): Add a new parameter "simple", and
1930 check generalized self-recursive pass-through jump function.
1931 (self_recursive_agg_pass_through_p): Likewise.
1932 (find_more_scalar_values_for_callers_subset): Compute value from
1933 pass-through jump function for self-recursive.
1934 (intersect_with_plats): Cleanup previous implementation code for value
1935 itersection with self-recursive call edge.
1936 (intersect_with_agg_replacements): Likewise.
1937 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1938 function for self-recursive call edge. Cleanup previous implementation
1939 code for value intersection with self-recursive call edge.
1940 (decide_whether_version_node): Remove dead callers and adjust order
1941 to let a non-self-recursive caller be the first element.
1943 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1945 * recog.c: Move pass_split_before_sched2 code in front of
1946 pass_split_before_regstack.
1947 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1948 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1949 (rest_of_handle_split_before_sched2): Remove.
1950 (pass_split_before_sched2::execute): Unconditionally call
1952 (enable_split_before_sched2): New function.
1953 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1954 (pass_split_before_regstack::gate): Ditto.
1955 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1956 Update name check for renamed split4 pass.
1957 * config/sh/sh.c (register_sh_passes): Update pass insertion
1958 point for renamed split4 pass.
1960 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1962 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1963 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1964 copying them around between host and target.
1966 2020-02-08 Andrew Pinski <apinski@marvell.com>
1969 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1970 STRICT_ALIGNMENT also.
1972 2020-02-08 Jim Wilson <jimw@sifive.com>
1975 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1977 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1978 Jakub Jelinek <jakub@redhat.com>
1981 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1982 xmm16-xmm31 call-used even in 64-bit ms-abi.
1984 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1986 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1987 (simd_ummla, simd_usmmla): Likewise.
1988 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1989 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1990 (vusmmlaq_s32): New.
1992 2020-02-07 Richard Biener <rguenther@suse.de>
1995 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1996 skipping unreachable regions.
1997 (optimize_inline_calls): Skip folding stmts when we didn't
2000 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
2003 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
2004 Don't return aggregates with only SFmode and DFmode in SSE
2006 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
2008 2020-02-07 Jakub Jelinek <jakub@redhat.com>
2011 * config/rs6000/rs6000-logue.c
2012 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
2013 if it fails, move rs into end_addr and retry. Add
2014 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
2015 the insn pattern doesn't describe well what exactly happens to
2019 * config/i386/predicates.md (avx_identity_operand): Remove.
2020 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
2021 (avx_<castmode><avxsizesuffix>_<castmode>,
2022 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
2023 a VEC_CONCAT of the operand and UNSPEC_CAST.
2024 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
2025 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
2029 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
2030 recog_data.insn if distance_non_agu_define changed it.
2032 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
2035 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
2036 we only had X-FORM (reg+reg) addressing for vectors. Also before
2037 ISA 3.0, we only had X-FORM addressing for scalars in the
2038 traditional Altivec registers.
2040 2020-02-06 <zhongyunde@huawei.com>
2041 Vladimir Makarov <vmakarov@redhat.com>
2043 PR rtl-optimization/93561
2044 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
2045 hard register range.
2047 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2049 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
2052 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
2054 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
2055 where the low and the high 32 bits are equal to each other specially,
2056 with an rldimi instruction.
2058 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2060 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
2062 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
2064 * config/arm/arm-tables.opt: Regenerate.
2066 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2069 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
2070 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
2071 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
2073 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
2075 PR rtl-optimization/87763
2076 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
2078 2020-02-06 Delia Burduv <delia.burduv@arm.com>
2080 * config/aarch64/aarch64-simd-builtins.def
2081 (bfmlaq): New built-in function.
2082 (bfmlalb): New built-in function.
2083 (bfmlalt): New built-in function.
2084 (bfmlalb_lane): New built-in function.
2085 (bfmlalt_lane): New built-in function.
2086 * config/aarch64/aarch64-simd.md
2087 (aarch64_bfmmlaqv4sf): New pattern.
2088 (aarch64_bfmlal<bt>v4sf): New pattern.
2089 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
2090 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
2091 (vbfmlalbq_f32): New intrinsic.
2092 (vbfmlaltq_f32): New intrinsic.
2093 (vbfmlalbq_lane_f32): New intrinsic.
2094 (vbfmlaltq_lane_f32): New intrinsic.
2095 (vbfmlalbq_laneq_f32): New intrinsic.
2096 (vbfmlaltq_laneq_f32): New intrinsic.
2097 * config/aarch64/iterators.md (BF_MLA): New int iterator.
2098 (bt): New int attribute.
2100 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
2102 * config/i386/i386.md (*pushtf): Emit "#" instead of
2103 calling gcc_unreachable in insn output.
2106 (*pushsf_rex64): Ditto for alternatives other than 1.
2107 (*pushsf): Ditto for alternatives other than 1.
2109 2020-02-06 Martin Liska <mliska@suse.cz>
2111 PR gcov-profile/91971
2112 PR gcov-profile/93466
2113 * coverage.c (coverage_init): Revert mangling of
2114 path into filename. It can lead to huge filename length.
2115 Creation of subfolders seem more natural.
2117 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2120 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
2121 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
2122 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
2124 2020-02-06 Jakub Jelinek <jakub@redhat.com>
2127 * config/i386/predicates.md (avx_identity_operand): New predicate.
2128 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
2129 define_insn_and_split.
2132 * omp-low.c (use_pointer_for_field): For nested constructs, also
2133 look for map clauses on target construct.
2134 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
2135 taskreg_nesting_level.
2138 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
2139 shared clause, call omp_notice_variable on outer context if any.
2141 2020-02-05 Jason Merrill <jason@redhat.com>
2144 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
2145 non-zero address even if weak and not yet defined.
2147 2020-02-05 Martin Sebor <msebor@redhat.com>
2149 PR tree-optimization/92765
2150 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
2151 * tree-ssa-strlen.c (compute_string_length): Remove.
2152 (determine_min_objsize): Remove.
2153 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
2154 Avoid using type size as the upper bound on string length.
2155 (handle_builtin_string_cmp): Add an argument. Adjust.
2156 (strlen_check_and_optimize_call): Pass additional argument to
2157 handle_builtin_string_cmp.
2159 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
2161 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
2162 (*pushdi2_rex64 peephole2): Unconditionally split after
2164 (*ashl<mode>3_doubleword): Ditto.
2165 (*<shift_insn><mode>3_doubleword): Ditto.
2167 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
2170 * config/rs6000/rs6000.c (get_vector_offset): Fix
2172 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
2174 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
2176 2020-02-05 David Malcolm <dmalcolm@redhat.com>
2179 (Special Functions for Debugging the Analyzer): Update description
2180 of __analyzer_dump_exploded_nodes.
2182 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2185 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
2186 include sets and not clobbers in the vzeroupper pattern.
2187 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
2188 the parallel has 17 (64-bit) or 9 (32-bit) elts.
2189 (*avx_vzeroupper_1): New define_insn_and_split.
2192 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
2193 don't run when !optimize.
2194 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
2197 2020-02-05 Richard Biener <rguenther@suse.de>
2200 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
2201 checks before matching calls.
2203 2020-02-05 Jakub Jelinek <jakub@redhat.com>
2205 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
2206 function comment typo.
2209 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
2210 simd_clone_create failed when i == 0, adjust clone->nargs by
2213 2020-02-05 Martin Liska <mliska@suse.cz>
2216 * doc/invoke.texi: Document that one should
2217 not combine ASLR and -fpch.
2219 2020-02-04 Richard Biener <rguenther@suse.de>
2221 PR tree-optimization/93538
2222 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
2224 2020-02-04 Richard Biener <rguenther@suse.de>
2226 PR tree-optimization/91123
2227 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
2228 (vn_walk_cb_data::last_vuse): New member.
2229 (vn_walk_cb_data::saved_operands): Likewsie.
2230 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
2231 (vn_walk_cb_data::push_partial_def): Use finish.
2232 (vn_reference_lookup_2): Update last_vuse and use finish if
2233 we've saved operands.
2234 (vn_reference_lookup_3): Use finish and update calls to
2235 push_partial_defs everywhere. When translating through
2236 memcpy or aggregate copies save off operands and alias-set.
2237 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
2238 operation for redundant store removal.
2240 2020-02-04 Richard Biener <rguenther@suse.de>
2242 PR tree-optimization/92819
2243 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
2244 generating more stmts than before.
2246 2020-02-04 Martin Liska <mliska@suse.cz>
2248 * config/arm/arm.c (arm_gen_far_branch): Move the function
2249 outside of selftests.
2251 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2253 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
2254 function to adjust PC-relative vector addresses.
2255 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
2256 handle vectors with PC-relative addresses.
2258 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2260 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
2262 (hard_reg_and_mode_to_addr_mask): Delete.
2263 (rs6000_adjust_vec_address): If the original vector address
2264 was REG+REG or REG+OFFSET and the element is not zero, do the add
2265 of the elements in the original address before adding the offset
2266 for the vector element. Use address_to_insn_form to validate the
2267 address using the register being loaded, rather than guessing
2268 whether the address is a DS-FORM or DQ-FORM address.
2270 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2272 * config/rs6000/rs6000.c (get_vector_offset): New helper function
2273 to calculate the offset in memory from the start of a vector of a
2274 particular element. Add code to keep the element number in
2275 bounds if the element number is variable.
2276 (rs6000_adjust_vec_address): Move calculation of offset of the
2277 vector element to get_vector_offset.
2278 (rs6000_split_vec_extract_var): Do not do the initial AND of
2279 element here, move the code to get_vector_offset.
2281 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
2283 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
2286 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
2288 * config/rs6000/constraints.md: Improve documentation.
2290 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
2293 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
2294 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
2296 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
2298 * config.gcc: Remove "carrizo" support.
2299 * config/gcn/gcn-opts.h (processor_type): Likewise.
2300 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
2301 * config/gcn/gcn.opt (gpu_type): Likewise.
2302 * config/gcn/t-omp-device: Likewise.
2304 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2307 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
2308 * config/arm/arm.c (arm_gen_far_branch): New function
2310 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
2312 2020-02-03 Julian Brown <julian@codesourcery.com>
2313 Tobias Burnus <tobias@codesourcery.com>
2315 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
2317 2020-02-03 Jakub Jelinek <jakub@redhat.com>
2320 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
2321 valid RTL to sum up the lowest and second lowest bytes of the popcnt
2324 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
2326 PR rtl-optimization/91333
2327 * ira-color.c (struct allocno_color_data): Add member
2329 (init_allocno_threads): Set the member up.
2330 (bucket_allocno_compare_func): Add compare hard reg
2333 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
2335 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
2337 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2338 * config.in: Regenerated.
2339 * configure: Regenerated.
2340 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
2341 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
2342 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
2344 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
2346 * configure: Regenerate.
2348 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
2350 PR rtl-optimization/91333
2351 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
2352 reg preferences comparison up.
2354 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2356 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
2357 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
2358 aarch64-sve-builtins-base.h.
2359 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
2360 aarch64-sve-builtins-base.cc.
2361 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
2362 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2364 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
2365 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2366 (svcvtnt): New functions.
2367 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
2368 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
2369 (svcvtnt): New functions.
2370 (svcvt): Add a form that converts f32 to bf16.
2371 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
2372 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
2374 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
2375 Treat B as bfloat16_t.
2376 (ternary_bfloat_lane_base): New class.
2377 (ternary_bfloat_def): Likewise.
2378 (ternary_bfloat): New shape.
2379 (ternary_bfloat_lane_def): New class.
2380 (ternary_bfloat_lane): New shape.
2381 (ternary_bfloat_lanex2_def): New class.
2382 (ternary_bfloat_lanex2): New shape.
2383 (ternary_bfloat_opt_n_def): New class.
2384 (ternary_bfloat_opt_n): New shape.
2385 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
2386 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
2387 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
2388 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
2389 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2390 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
2391 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
2392 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
2393 the pattern off the narrow mode instead of the wider one.
2394 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
2395 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
2396 (sve_fp_op): Handle them.
2397 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
2398 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
2400 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
2402 * config/aarch64/arm_sve.h: Include arm_bf16.h.
2403 * config/aarch64/aarch64-modes.def (BF): Move definition before
2404 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
2405 (SVE_MODES): Handle BF modes.
2406 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
2408 (aarch64_full_sve_mode): Likewise.
2409 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
2411 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
2412 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
2413 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
2414 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
2416 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
2418 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
2419 (TYPES_all_data): Add bf16.
2420 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
2421 (register_tuple_type): Increase buffer size.
2422 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
2423 (bf16): New type suffix.
2424 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
2425 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
2426 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
2427 Change type from all_data to all_arith.
2428 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
2431 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
2432 Matthew Malcomson <matthew.malcomson@arm.com>
2433 Richard Sandiford <richard.sandiford@arm.com>
2435 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
2436 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
2437 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
2438 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
2439 __ARM_FEATURE_MATMUL_FP64.
2440 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
2441 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
2442 be disabled at the same time.
2443 (f32mm): New extension.
2444 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
2445 (AARCH64_FL_F64MM): Bump to the next bit up.
2446 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
2447 (TARGET_SVE_F64MM): New macros.
2448 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
2449 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
2450 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
2451 (UNSPEC_ZIP2Q): New unspeccs.
2452 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
2453 (optab, sur, perm_insn): Handle the new unspecs.
2454 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
2455 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
2456 TARGET_SVE_F64MM instead of separate tests.
2457 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
2458 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
2459 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
2460 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
2461 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
2462 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
2463 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
2464 (TYPES_s_signed): New macro.
2465 (TYPES_s_integer): Use it.
2466 (TYPES_d_float): New macro.
2467 (TYPES_d_data): Use it.
2468 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
2469 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
2470 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
2471 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
2472 (svmmla): New shape.
2473 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
2474 template parameters.
2475 (ternary_resize2_lane_base): Likewise.
2476 (ternary_resize2_base): New class.
2477 (ternary_qq_lane_base): Likewise.
2478 (ternary_intq_uintq_lane_def): Likewise.
2479 (ternary_intq_uintq_lane): New shape.
2480 (ternary_intq_uintq_opt_n_def): New class
2481 (ternary_intq_uintq_opt_n): New shape.
2482 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
2483 (ternary_uintq_intq_def): New class.
2484 (ternary_uintq_intq): New shape.
2485 (ternary_uintq_intq_lane_def): New class.
2486 (ternary_uintq_intq_lane): New shape.
2487 (ternary_uintq_intq_opt_n_def): New class.
2488 (ternary_uintq_intq_opt_n): New shape.
2489 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
2490 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
2491 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
2492 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
2494 (svdotprod_lane_impl): ...this new class.
2495 (svmmla_impl, svusdot_impl): New classes.
2496 (svdot_lane): Update to use svdotprod_lane_impl.
2497 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
2498 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
2500 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
2501 function, with no types defined.
2502 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
2503 AARCH64_FL_I8MM functions.
2504 (svmmla): New AARCH64_FL_F32MM function.
2505 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
2506 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
2507 AARCH64_FL_F64MM function.
2508 (REQUIRED_EXTENSIONS):
2510 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2512 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
2515 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
2517 * config/i386/i386.md (*movoi_internal_avx): Do not check for
2518 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
2519 (*movti_internal): Do not check for
2520 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2521 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
2522 just after check for TARGET_AVX.
2523 (*movdf_internal): Ditto.
2524 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
2525 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
2526 * config/i386/sse.md (mov<mode>_internal): Only check
2527 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
2528 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
2529 (<sse>_andnot<mode>3<mask_name>): Move check for
2530 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
2531 (<code><mode>3<mask_name>): Ditto.
2532 (*andnot<mode>3): Ditto.
2533 (*andnottf3): Ditto.
2534 (*<code><mode>3): Ditto.
2535 (*<code>tf3): Ditto.
2536 (*andnot<VI:mode>3): Remove
2537 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
2538 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
2539 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
2540 (sse4_1_blendv<ssemodesuffix>): Ditto.
2541 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
2542 Explain that tune applies to 128bit instructions only.
2544 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
2546 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
2547 to definition of hsa_kernel_description. Parse assembly to find SGPR
2548 and VGPR count of kernel and store in hsa_kernel_description.
2550 2020-01-31 Tamar Christina <tamar.christina@arm.com>
2552 PR rtl-optimization/91838
2553 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
2554 to truncate if allowed or reject combination.
2556 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
2558 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
2559 (find_inv_vars_cb): Likewise.
2561 2020-01-31 David Malcolm <dmalcolm@redhat.com>
2563 * calls.c (special_function_p): Split out the check for DECL_NAME
2564 being non-NULL and fndecl being extern at file scope into a
2565 new maybe_special_function_p and call it. Drop check for fndecl
2566 being non-NULL that was after a usage of DECL_NAME (fndecl).
2567 * tree.h (maybe_special_function_p): New inline function.
2569 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2571 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
2572 (mask_gather_load<mode>): ... here, and zero-initialize the
2574 (maskload<mode>di): Zero-initialize the destination.
2577 2020-01-30 David Malcolm <dmalcolm@redhat.com>
2580 * doc/analyzer.texi (Limitations): Note that constraints on
2581 floating-point values are currently ignored.
2583 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2586 * symtab.c (symtab_node::noninterposable_alias): If localalias
2587 already exists, but is not usable, append numbers after it until
2588 a unique name is found. Formatting fix.
2591 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
2594 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
2596 * config/gcn/gcn.c (print_operand): Handle LTGT.
2597 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
2599 2020-01-30 Richard Biener <rguenther@suse.de>
2601 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
2602 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
2604 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
2606 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
2607 without a DECL in .data.rel.ro.local.
2609 2020-01-30 Jakub Jelinek <jakub@redhat.com>
2612 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
2616 * config/i386/sse.md
2617 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
2618 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
2619 any_extend code iterator instead of always zero_extend.
2620 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
2621 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
2622 Use any_extend code iterator instead of always zero_extend.
2623 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
2624 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
2625 Use any_extend code iterator instead of always zero_extend.
2626 (*sse2_pmovmskb_ext): New define_insn.
2627 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
2630 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
2631 (*popcountsi2_zext_falsedep): New define_insn.
2633 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2635 * config.in: Regenerated.
2636 * configure: Regenerated.
2638 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2641 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2642 LLVM's assembler changed the default in version 9.
2644 2020-01-24 Jeff Law <law@redhat.com>
2646 PR tree-optimization/89689
2647 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2649 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2653 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2655 PR rtl-optimization/87763
2656 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2657 simplification to handle subregs as well as bare regs.
2658 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2660 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2663 * ira.c (ira): Revert use of simplified LRA algorithm.
2665 2020-01-29 Martin Jambor <mjambor@suse.cz>
2667 PR tree-optimization/92706
2668 * tree-sra.c (struct access): Fields first_link, last_link,
2669 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2670 next_rhs_queued and grp_rhs_queued respectively, new fields
2671 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2672 (struct assign_link): Field next renamed to next_rhs, new field
2673 next_lhs. Updated comment.
2674 (work_queue_head): Renamed to rhs_work_queue_head.
2675 (lhs_work_queue_head): New variable.
2676 (add_link_to_lhs): New function.
2677 (relink_to_new_repr): Also relink LHS lists.
2678 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2679 (add_access_to_lhs_work_queue): New function.
2680 (pop_access_from_work_queue): Renamed to
2681 pop_access_from_rhs_work_queue.
2682 (pop_access_from_lhs_work_queue): New function.
2683 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2685 (child_would_conflict_in_lacc): Renamed to
2686 child_would_conflict_in_acc. Adjusted parameter names.
2687 (create_artificial_child_access): New parameter set_grp_read, use it.
2688 (subtree_mark_written_and_enqueue): Renamed to
2689 subtree_mark_written_and_rhs_enqueue.
2690 (propagate_subaccesses_across_link): Renamed to
2691 propagate_subaccesses_from_rhs.
2692 (propagate_subaccesses_from_lhs): New function.
2693 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2696 2020-01-29 Martin Jambor <mjambor@suse.cz>
2698 PR tree-optimization/92706
2699 * tree-sra.c (struct access): Adjust comment of
2700 grp_total_scalarization.
2701 (find_access_in_subtree): Look for single children spanning an entire
2703 (scalarizable_type_p): Allow register accesses, adjust callers.
2704 (completely_scalarize): Remove function.
2705 (scalarize_elem): Likewise.
2706 (create_total_scalarization_access): Likewise.
2707 (sort_and_splice_var_accesses): Do not track total scalarization
2709 (analyze_access_subtree): New parameter totally, adjust to new meaning
2710 of grp_total_scalarization.
2711 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2712 (can_totally_scalarize_forest_p): New function.
2713 (create_total_scalarization_access): Likewise.
2714 (create_total_access_and_reshape): Likewise.
2715 (total_should_skip_creating_access): Likewise.
2716 (totally_scalarize_subtree): Likewise.
2717 (analyze_all_variable_accesses): Perform total scalarization after
2718 subaccess propagation using the new functions above.
2719 (initialize_constant_pool_replacements): Output initializers by
2720 traversing the access tree.
2722 2020-01-29 Martin Jambor <mjambor@suse.cz>
2724 * tree-sra.c (verify_sra_access_forest): New function.
2725 (verify_all_sra_access_forests): Likewise.
2726 (create_artificial_child_access): Set parent.
2727 (analyze_all_variable_accesses): Call the verifier.
2729 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2731 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2732 if called on indirect edge.
2733 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2734 speculative call if needed.
2736 2020-01-29 Richard Biener <rguenther@suse.de>
2738 PR tree-optimization/93428
2739 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2740 permutation when the load node is created.
2741 (vect_analyze_slp_instance): Re-use it here.
2743 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2745 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2747 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2749 PR rtl-optimization/93272
2750 * ira-lives.c (process_out_of_region_eh_regs): New function.
2751 (process_bb_node_lives): Call it.
2753 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2755 * coverage.c (read_counts_file): Make error message lowercase.
2757 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2759 * profile-count.c (profile_quality_display_names): Fix ordering.
2761 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2764 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2765 hash only when edge is first within the sequence.
2766 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2767 (symbol_table::create_edge): Do not set target_prob.
2768 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2770 (cgraph_edge::make_speculative): Drop target_prob parameter.
2771 (cgraph_edge::speculative_call_info): Remove.
2772 (cgraph_edge::first_speculative_call_target): New member function.
2773 (update_call_stmt_hash_for_removing_direct_edge): New function.
2774 (cgraph_edge::resolve_speculation): Rewrite to new API.
2775 (cgraph_edge::speculative_call_for_target): New member function.
2776 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2777 multiple speculation targets.
2778 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2780 (verify_speculative_call): Verify that targets form an interval.
2781 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2782 (cgraph_edge::first_speculative_call_target): New member function.
2783 (cgraph_edge::next_speculative_call_target): New member function.
2784 (cgraph_edge::speculative_call_target_ref): New member function.
2785 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2786 (cgraph_edge): Remove target_prob.
2787 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2788 Fix handling of speculative calls.
2789 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2790 * ipa-fnsummary.c (analyze_function_body): Likewise.
2791 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2792 * ipa-profile.c (dump_histogram): Fix formating.
2793 (ipa_profile_generate_summary): Watch for overflows.
2794 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2795 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2796 (update_indirect_edges_after_inlining): Update to new API.
2797 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2799 * profile-count.h: (profile_probability::adjusted): New.
2800 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2801 updating of profile.
2802 * value-prof.c (gimple_ic_transform): Rename to ...
2803 (dump_ic_profile): ... this one; update dumping.
2804 (stream_in_histogram_value): Fix formating.
2805 (gimple_value_profile_transformations): Update.
2807 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2810 * config/i386/i386.md (*movoi_internal_avx): Remove
2811 TARGET_SSE_TYPELESS_STORES check.
2812 (*movti_internal): Prefer TARGET_AVX over
2813 TARGET_SSE_TYPELESS_STORES.
2814 (*movtf_internal): Likewise.
2815 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2816 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2817 from TARGET_SSE_TYPELESS_STORES.
2819 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2821 * diagnostic-core.h (warning_at): Rename overload to...
2822 (warning_meta): ...this.
2823 (emit_diagnostic_valist): Delete decl of overload taking
2824 diagnostic_metadata.
2825 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2826 (warning_at): Rename overload taking diagnostic_metadata to...
2827 (warning_meta): ...this.
2829 2020-01-28 Richard Biener <rguenther@suse.de>
2831 PR tree-optimization/93439
2832 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2833 * tree-cfg.c (move_sese_region_to_fn): ... here.
2834 (verify_types_in_gimple_reference): Verify used cliques are
2837 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2840 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2841 argument of a pointer to struct gcc_options and pass it to
2842 parse_mtune_ctrl_str.
2843 (ix86_function_specific_restore): Pass opts to
2844 set_ix86_tune_features.
2845 (ix86_option_override_internal): Likewise.
2846 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2847 gcc_options and use it for x_ix86_tune_ctrl_string.
2849 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2851 PR rtl-optimization/87763
2852 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2853 simplification to handle subregs as well as bare regs.
2854 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2856 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2858 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2859 for reduction chains that (now) include a call.
2861 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2863 PR tree-optimization/92822
2864 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2865 out the don't-care elements of a vector whose significant elements
2866 are duplicates, make the don't-care elements duplicates too.
2868 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2870 PR tree-optimization/93434
2871 * tree-predcom.c (split_data_refs_to_components): Record which
2872 components have had aliasing loads removed. Prevent store-store
2873 commoning for all such components.
2875 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2878 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2879 -1 or is_vshift is true, use new_vector with number of elts npatterns
2880 rather than new_unary_operation.
2882 PR tree-optimization/93454
2883 * gimple-fold.c (fold_array_ctor_reference): Perform
2884 elt_size.to_uhwi () just once, instead of calling it in every
2885 iteration. Punt if that value is above size of the temporary
2886 buffer. Decrease third native_encode_expr argument when
2887 bufoff + elt_sz is above size of buf.
2889 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2891 * config/mips/mips.c (mips_declare_object_name)
2892 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2894 2020-01-27 Martin Liska <mliska@suse.cz>
2896 PR gcov-profile/93403
2897 * tree-profile.c (gimple_init_gcov_profiler): Generate
2898 both __gcov_indirect_call_profiler_v4 and
2899 __gcov_indirect_call_profiler_v4_atomic.
2901 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2904 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2906 (@aarch64_split_simd_mov<mode>): Use it.
2907 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2908 Leave the vec_extract patterns to handle 2-element vectors.
2909 (aarch64_simd_mov_from_<mode>high): Likewise.
2910 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2911 (vec_extractv2dfv1df): Likewise.
2913 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2915 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2916 jump conditions for *compare_condjump<GPI:mode>.
2918 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2921 * digraph.cc (test_edge::test_edge): Specify template for base
2924 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2926 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2928 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2930 * config/arc/arc-protos.h (gen_mlo): Remove.
2931 (gen_mhi): Likewise.
2932 * config/arc/arc.c (AUX_MULHI): Define.
2933 (arc_must_save_reister): Special handling for r58/59.
2934 (arc_compute_frame_size): Consider mlo/mhi registers.
2935 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2937 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2938 mlo/mhi name selection.
2939 (arc_restore_callee_saves): Don't early restore blink when ISR.
2940 (arc_expand_prologue): Add mlo/mhi saving.
2941 (arc_expand_epilogue): Add mlo/mhi restoring.
2944 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2945 numbering when MUL64 option is used.
2946 (DWARF2_FRAME_REG_OUT): Define.
2947 * config/arc/arc.md (arc600_stall): New pattern.
2948 (VUNSPEC_ARC_ARC600_STALL): Define.
2949 (mulsi64): Use correct mlo/mhi registers.
2950 (mulsi_600): Clean it up.
2951 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2953 (mhi_operand): Likewise.
2955 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2956 Petro Karashchenko <petro.karashchenko@ring.com>
2958 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2959 attributes if needed.
2960 (prepare_move_operands): Generate special unspec instruction for
2962 (arc_isuncached_mem_p): Propagate uncached attribute to each
2964 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2965 (VUNSPEC_ARC_STDI): Likewise.
2966 (ALLI): New mode iterator.
2967 (mALLI): New mode attribute.
2968 (lddi): New instruction pattern.
2970 (stdidi_split): Split instruction for architectures which are not
2971 supporting ll64 option.
2972 (lddidi_split): Likewise.
2974 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2976 PR rtl-optimization/92989
2977 * lra-lives.c (process_bb_lives): Update the live-in set before
2978 processing additional clobbers.
2980 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2982 PR rtl-optimization/93170
2983 * cselib.c (cselib_invalidate_regno_val): New function, split out
2985 (cselib_invalidate_regno): ...here.
2986 (cselib_invalidated_by_call_p): New function.
2987 (cselib_process_insn): Iterate over all the hard-register entries in
2988 REG_VALUES and invalidate any that cross call-clobbered registers.
2990 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2992 * dojump.c (split_comparison): Use HONOR_NANS rather than
2993 HONOR_SNANS when splitting LTGT.
2995 2020-01-27 Martin Liska <mliska@suse.cz>
2998 * opts.c (print_filtered_help): Exclude language-specific
2999 options from --help=common unless enabled in all FEs.
3001 2020-01-27 Martin Liska <mliska@suse.cz>
3003 * opts.c (print_help): Exclude params from
3004 all except --help=param.
3006 2020-01-27 Martin Liska <mliska@suse.cz>
3009 * config/i386/i386-features.c (make_resolver_func):
3010 Align the code with ppc64 target implementation.
3011 Do not generate a unique name for resolver function.
3013 2020-01-27 Richard Biener <rguenther@suse.de>
3015 PR tree-optimization/93397
3016 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
3017 converted reduction chain SLP graph adjustment.
3019 2020-01-26 Marek Polacek <polacek@redhat.com>
3022 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
3025 2020-01-26 Jason Merrill <jason@redhat.com>
3028 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
3031 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
3033 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
3034 (rx_setmem): Likewise.
3036 2020-01-26 Jakub Jelinek <jakub@redhat.com>
3039 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
3040 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
3041 drop <di> from constraint of last operand.
3044 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
3045 TARGET_AVX2 and V4DFmode not in the split condition, but in the
3046 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
3048 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
3051 * ipa-cp.c (get_info_about_necessary_edges): Remove value
3054 2020-01-24 Jeff Law <law@redhat.com>
3056 PR tree-optimization/92788
3057 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
3060 2020-01-24 Jakub Jelinek <jakub@redhat.com>
3063 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
3064 *avx_vperm_broadcast_<mode>,
3065 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
3066 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
3067 Move before avx2_perm<mode>/avx512f_perm<mode>.
3070 * simplify-rtx.c (simplify_const_unary_operation,
3071 simplify_const_binary_operation): Punt for mode precision above
3072 MAX_BITSIZE_MODE_ANY_INT.
3074 2020-01-24 Andrew Pinski <apinski@marvell.com>
3076 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
3079 2020-01-24 Jeff Law <law@redhat.com>
3082 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
3083 for REGs. Call output_operand_lossage to get more reasonable
3086 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
3088 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
3089 gcn_fp_compare_operator.
3090 (vec_cmpu<mode>di): Use gcn_compare_operator.
3091 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
3092 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
3093 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
3094 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
3095 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
3096 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
3097 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
3098 gcn_fp_compare_operator.
3099 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
3100 gcn_fp_compare_operator.
3101 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
3102 gcn_fp_compare_operator.
3103 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
3104 gcn_fp_compare_operator.
3106 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
3108 * doc/install.texi (Cross-Compiler-Specific Options): Document
3109 `--with-toolexeclibdir' option.
3111 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
3113 * target.def (flags_regnum): Also mention effect on delay slot filling.
3114 * doc/tm.texi: Regenerate.
3116 2020-01-23 Jeff Law <law@redhat.com>
3118 PR translation/90162
3119 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
3121 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
3124 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
3127 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3129 PR rtl-optimization/93402
3130 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
3133 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3135 * config.in: Regenerated.
3136 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
3137 for TARGET_LIBC_GNUSTACK.
3138 * configure: Regenerated.
3139 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
3140 found to be 2.31 or greater.
3142 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
3144 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
3146 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
3147 (mips_asm_file_end): New function. Delegate to
3148 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
3149 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3151 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3154 * config/i386/i386-modes.def (POImode): New mode.
3155 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
3156 * config/i386/i386.md (DPWI): New mode attribute.
3157 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
3159 (QPWI): ... this. Use POI instead of OI for TImode.
3160 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
3161 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
3164 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3167 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
3169 (speculation_tracker_rev): New pattern.
3170 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
3171 Use speculation_tracker_rev to track the inverse condition.
3173 2020-01-23 Richard Biener <rguenther@suse.de>
3175 PR tree-optimization/93381
3176 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
3177 alias-set of the def as argument and record the first one.
3178 (vn_walk_cb_data::first_set): New member.
3179 (vn_reference_lookup_3): Pass the alias-set of the current def
3180 to push_partial_def. Fix alias-set used in the aggregate copy
3182 (vn_reference_lookup): Consistently set *last_vuse_ptr.
3183 * real.c (clear_significand_below): Fix out-of-bound access.
3185 2020-01-23 Jakub Jelinek <jakub@redhat.com>
3188 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
3189 New define_insn patterns.
3191 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3193 * doc/sourcebuild.texi (check-function-bodies): Add an
3194 optional target/xfail selector.
3196 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
3198 PR rtl-optimization/93124
3199 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
3200 bare USE and CLOBBER insns.
3202 2020-01-22 Andrew Pinski <apinski@marvell.com>
3204 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
3206 2020-01-22 David Malcolm <dmalcolm@redhat.com>
3209 * gdbinit.in (break-on-saved-diagnostic): Update for move of
3210 diagnostic_manager into "ana" namespace.
3211 * selftest-run-tests.c (selftest::run_tests): Update for move of
3212 selftest::run_analyzer_selftests to
3213 ana::selftest::run_analyzer_selftests.
3215 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
3217 * cfgexpand.c (union_stack_vars): Update the size.
3219 2020-01-22 Richard Biener <rguenther@suse.de>
3221 PR tree-optimization/93381
3222 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
3223 throughout, handle all conversions the same.
3225 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3228 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
3229 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
3230 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
3231 Call force_reg on high_in2 unconditionally.
3233 2020-01-22 Martin Liska <mliska@suse.cz>
3235 PR tree-optimization/92924
3236 * profile.c (compute_value_histograms): Divide
3239 2020-01-22 Jakub Jelinek <jakub@redhat.com>
3242 * output.h (assemble_name_resolve): Declare.
3243 * varasm.c (assemble_name_resolve): New function.
3244 (assemble_name): Use it.
3245 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
3247 2020-01-22 Joseph Myers <joseph@codesourcery.com>
3249 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
3250 update_web_docs_git instead of update_web_docs_svn.
3252 2020-01-21 Andrew Pinski <apinski@marvell.com>
3255 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
3256 as PTR mode. Have operand 1 as being modeless, it can be P mode.
3257 (*tlsgd_small_<mode>): Likewise.
3258 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
3259 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
3260 register. Convert that register back to dest using convert_mode.
3262 2020-01-21 Jim Wilson <jimw@sifive.com>
3264 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
3267 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
3268 Uros Bizjak <ubizjak@gmail.com>
3271 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
3273 (legitimize_tls_address): Do GNU2 TLS address computation in
3274 ptr_mode and zero-extend result to Pmode.
3275 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
3276 :P with :PTR and Pmode with ptr_mode.
3277 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
3278 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
3279 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
3281 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3284 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
3285 the last two operands are CONST_INT_P before using them as such.
3287 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3289 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
3290 to get the integer element types.
3292 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
3294 * config/aarch64/aarch64-sve-builtins.h
3295 (function_expander::convert_to_pmode): Declare.
3296 * config/aarch64/aarch64-sve-builtins.cc
3297 (function_expander::convert_to_pmode): New function.
3298 (function_expander::get_contiguous_base): Use it.
3299 (function_expander::prepare_gather_address_operands): Likewise.
3300 * config/aarch64/aarch64-sve-builtins-sve2.cc
3301 (svwhilerw_svwhilewr_impl::expand): Likewise.
3303 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
3306 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
3307 cfun->machine->label_is_assembled.
3308 (aarch64_print_patchable_function_entry): New.
3309 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
3310 * config/aarch64/aarch64.h (struct machine_function): New field,
3313 2020-01-21 David Malcolm <dmalcolm@redhat.com>
3316 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
3319 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3322 * cgraph.c (cgraph_edge::resolve_speculation,
3323 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
3324 call_stmt_site_hash.
3326 2020-01-21 Martin Liska <mliska@suse.cz>
3328 * config/rs6000/rs6000.c (common_mode_defined): Remove
3331 2020-01-21 Richard Biener <rguenther@suse.de>
3333 PR tree-optimization/92328
3334 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
3335 type when value-numbering same-sized store by inserting a
3337 (eliminate_dom_walker::eliminate_stmt): When eliminating
3338 a redundant store handle bit-reinterpretation of the same value.
3340 2020-01-21 Andrew Pinski <apinski@marvel.com>
3343 * tree-into-ssa.c (prepare_block_for_update_1): Split out
3345 (prepare_block_for_update): This. Use a worklist instead of
3348 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3350 * gcc/config/arm/arm.c (clear_operation_p):
3351 Initialise last_regno, skip first iteration
3352 based on the first_set value and use ints instead
3353 of the unnecessary HOST_WIDE_INTs.
3355 2020-01-21 Jakub Jelinek <jakub@redhat.com>
3358 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
3359 compare_mode other than SFmode or DFmode.
3361 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
3364 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
3365 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
3366 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
3368 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3370 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
3372 2020-01-20 Andrew Pinski <apinski@marvell.com>
3375 * targhooks.c (default_print_patchable_function_entry): Use
3376 output_asm_insn to emit the nop instruction.
3378 2020-01-20 Fangrui Song <maskray@google.com>
3381 * targhooks.c (default_print_patchable_function_entry): Align to
3384 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
3387 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
3388 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
3389 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
3390 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
3391 (*tls_dynamic_gnu2_lea_64): Renamed to ...
3392 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
3393 Remove the {q} suffix from lea.
3394 (*tls_dynamic_gnu2_call_64): Renamed to ...
3395 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
3396 (*tls_dynamic_gnu2_combine_64): Renamed to ...
3397 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
3398 Pass Pmode to gen_tls_dynamic_gnu2_64.
3400 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
3402 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
3404 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
3406 * config/aarch64/aarch64-sve-builtins-base.cc
3407 (svld1ro_impl::memory_vector_mode): Remove parameter name.
3409 2020-01-20 Richard Biener <rguenther@suse.de>
3412 * dwarf2out.c (prune_unused_types): Unconditionally mark
3413 called function DIEs.
3415 2020-01-20 Martin Liska <mliska@suse.cz>
3417 PR tree-optimization/93199
3418 * tree-eh.c (struct leh_state): Add
3419 new field outer_non_cleanup.
3420 (cleanup_is_dead_in): Pass leh_state instead
3421 of eh_region. Add a checking that state->outer_non_cleanup
3422 points to outer non-clean up region.
3423 (lower_try_finally): Record outer_non_cleanup
3425 (lower_catch): Likewise.
3426 (lower_eh_filter): Likewise.
3427 (lower_eh_must_not_throw): Likewise.
3428 (lower_cleanup): Likewise.
3430 2020-01-20 Richard Biener <rguenther@suse.de>
3432 PR tree-optimization/93094
3433 * tree-vectorizer.h (vect_loop_versioning): Adjust.
3434 (vect_transform_loop): Likewise.
3435 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
3436 loop_vectorized_call to vect_transform_loop.
3437 * tree-vect-loop.c (vect_transform_loop): Pass down
3438 loop_vectorized_call to vect_loop_versioning.
3439 * tree-vect-loop-manip.c (vect_loop_versioning): Use
3440 the earlier discovered loop_vectorized_call.
3442 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
3444 * doc/contribute.texi: Update for SVN -> Git transition.
3445 * doc/install.texi: Likewise.
3447 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3450 * cgraph.c (cgraph_edge::make_speculative): Increase number of
3451 speculative targets.
3452 (verify_speculative_call): New function
3453 (cgraph_node::verify_node): Use it.
3454 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
3457 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
3460 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
3461 (cgraph_edge::make_direct): Remove all indirect targets.
3462 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
3463 (cgraph_node::verify_node): Verify that only one call_stmt or
3464 lto_stmt_uid is set.
3465 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
3467 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
3468 (lto_output_ref): Simplify streaming of stmt.
3469 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
3471 2020-01-18 Tamar Christina <tamar.christina@arm.com>
3473 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
3474 Mark parameter unused.
3476 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
3478 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
3480 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
3482 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
3484 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
3486 * Makefile.in: Add coroutine-passes.o.
3487 * builtin-types.def (BT_CONST_SIZE): New.
3488 (BT_FN_BOOL_PTR): New.
3489 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
3490 * builtins.def (DEF_COROUTINE_BUILTIN): New.
3491 * coroutine-builtins.def: New file.
3492 * coroutine-passes.cc: New file.
3493 * function.h (struct GTY function): Add a bit to indicate that the
3494 function is a coroutine component.
3495 * internal-fn.c (expand_CO_FRAME): New.
3496 (expand_CO_YIELD): New.
3497 (expand_CO_SUSPN): New.
3498 (expand_CO_ACTOR): New.
3499 * internal-fn.def (CO_ACTOR): New.
3503 * passes.def: Add pass_coroutine_lower_builtins,
3504 pass_coroutine_early_expand_ifns.
3505 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
3506 (make_pass_coroutine_early_expand_ifns): New.
3507 * doc/invoke.texi: Document the fcoroutines command line
3510 2020-01-18 Jakub Jelinek <jakub@redhat.com>
3512 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
3515 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
3516 after checking the argument is a REG. Don't use REGNO (reg)
3517 again to set last_regno, reuse regno variable instead.
3519 2020-01-17 David Malcolm <dmalcolm@redhat.com>
3521 * doc/analyzer.texi (Limitations): Add note about NaN.
3523 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3524 Sudakshina Das <sudi.das@arm.com>
3526 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
3527 and valid immediate.
3528 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
3529 (lshrdi3): Generate thumb2_lsrl for valid immediates.
3530 * config/arm/constraints.md (Pg): New.
3531 * config/arm/predicates.md (long_shift_imm): New.
3532 (arm_reg_or_long_shift_imm): Likewise.
3533 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
3534 (thumb2_lsll): Likewise.
3537 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3538 Sudakshina Das <sudi.das@arm.com>
3540 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
3541 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
3542 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
3543 register pairs for doubleword quantities for ARMv8.1M-Mainline.
3544 * config/arm/thumb2.md (thumb2_asrl): New.
3545 (thumb2_lsll): Likewise.
3547 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3549 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
3552 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
3554 * gdbinit.in (help-gcc-hooks): New command.
3555 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
3556 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
3559 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3561 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
3562 correct target macro.
3564 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3566 * config/aarch64/aarch64-protos.h
3567 (aarch64_sve_ld1ro_operand_p): New.
3568 * config/aarch64/aarch64-sve-builtins-base.cc
3569 (class load_replicate): New.
3570 (class svld1ro_impl): New.
3571 (class svld1rq_impl): Change to inherit from load_replicate.
3572 (svld1ro): New sve intrinsic function base.
3573 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
3574 New DEF_SVE_FUNCTION.
3575 * config/aarch64/aarch64-sve-builtins-base.h
3576 (svld1ro): New decl.
3577 * config/aarch64/aarch64-sve-builtins.cc
3578 (function_expander::add_mem_operand): Modify assert to allow
3580 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
3582 * config/aarch64/aarch64.c
3583 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
3584 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
3585 (aarch64_sve_ld1ro_operand_p): New.
3586 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
3587 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
3588 * config/aarch64/predicates.md
3589 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
3591 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
3593 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
3594 Introduce this ACLE specified predefined macro.
3595 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
3596 (fp): Disabling this disables f64mm.
3597 (simd): Disabling this disables f64mm.
3598 (fp16): Disabling this disables f64mm.
3599 (sve): Disabling this disables f64mm.
3600 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
3601 (AARCH64_ISA_F64MM): New.
3602 (TARGET_F64MM): New.
3603 * doc/invoke.texi (f64mm): Document new option.
3605 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3607 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
3608 (neoversen1_tunings): Likewise.
3610 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3613 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
3614 Add assert to ensure prolog has been emitted.
3615 (aarch64_split_atomic_op): Likewise.
3616 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
3617 Use epilogue_completed rather than reload_completed.
3618 (aarch64_atomic_exchange<mode>): Likewise.
3619 (aarch64_atomic_<atomic_optab><mode>): Likewise.
3620 (atomic_nand<mode>): Likewise.
3621 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
3622 (atomic_fetch_nand<mode>): Likewise.
3623 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
3624 (atomic_nand_fetch<mode>): Likewise.
3626 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3629 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
3631 (REVERSE_CONDITION): Delete.
3632 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
3633 (CCFP_CCFPE): Likewise.
3634 (e): New mode attribute.
3635 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3636 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3637 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3638 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3639 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3640 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3641 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3642 name of generator from gen_ccmpdi to gen_ccmpccdi.
3643 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3644 the previous comparison but aren't able to, use the new ccmp_rev
3647 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3649 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3650 than testing directly for INTEGER_CST.
3651 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3653 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3655 PR tree-optimization/93292
3656 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3657 get_vectype_for_scalar_type returns NULL.
3659 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3661 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3662 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3664 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3666 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3668 * params.opt: (max-predicted-iterations): Set bounds.
3669 * predict.c (real_almost_one, real_br_prob_base,
3670 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3671 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3672 probabilities; do not truncate to reg_br_prob_bases.
3673 (estimate_loops_at_level): Pass max_cyclic_prob.
3674 (estimate_loops): Compute max_cyclic_prob.
3675 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3677 * profile-count.c (profile_probability::to_sreal): New.
3678 * profile-count.h (class sreal): Move up in file.
3679 (profile_probability::to_sreal): Declare.
3681 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3684 (arm_invalid_conversion): New function for target hook.
3685 (arm_invalid_unary_op): New function for target hook.
3686 (arm_invalid_binary_op): New function for target hook.
3688 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3690 * config.gcc: Add arm_bf16.h.
3691 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3692 (arm_simd_builtin_std_type): Add BFmode.
3693 (arm_init_simd_builtin_types): Define element types for vector types.
3694 (arm_init_bf16_types): New function.
3695 (arm_init_builtins): Add arm_init_bf16_types function call.
3696 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3697 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3698 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3699 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3700 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3701 (arm_mangle_type): Add __bf16.
3702 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3703 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3704 arm_bf16_ptr_type_node.
3705 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3706 define_split between ARM registers.
3707 * config/arm/arm_bf16.h: New file.
3708 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3709 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3711 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3712 * config/arm/vfp.md: Add BFmode to movhf patterns.
3714 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3715 Andre Vieira <andre.simoesdiasvieira@arm.com>
3717 * config/arm/arm-cpus.in (mve, mve_float): New features.
3718 (dsp, mve, mve.fp): New options.
3719 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3720 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3721 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3723 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3724 Thomas Preud'homme <thomas.preudhomme@arm.com>
3726 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3728 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3729 error for using -mcmse when targeting Armv8.1-M Mainline.
3731 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3732 Thomas Preud'homme <thomas.preudhomme@arm.com>
3734 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3735 address in r4 when targeting Armv8.1-M Mainline.
3736 (nonsecure_call_value_internal): Likewise.
3737 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3738 a register match_operand again. Emit BLXNS when targeting
3740 (nonsecure_call_value_reg_thumb2): Likewise.
3742 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3743 Thomas Preud'homme <thomas.preudhomme@arm.com>
3745 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3746 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3747 variable as true when floating-point ABI is not hard. Replace
3748 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3749 Generate VLSTM and VLLDM instruction respectively before and
3750 after a function call to cmse_nonsecure_call function.
3751 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3752 (VUNSPEC_VLLDM): Likewise.
3753 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3754 (lazy_load_multiple_insn): Likewise.
3756 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3757 Thomas Preud'homme <thomas.preudhomme@arm.com>
3759 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3760 (arm_emit_vfp_multi_reg_pop): Likewise.
3761 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3762 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3763 restore callee-saved VFP registers.
3765 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3766 Thomas Preud'homme <thomas.preudhomme@arm.com>
3768 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3769 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3770 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3771 callee-saved GPRs as well as clear ip register before doing a nonsecure
3772 call then restore callee-saved GPRs after it when targeting
3774 (arm_reorg): Adapt to function rename.
3776 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3777 Thomas Preud'homme <thomas.preudhomme@arm.com>
3779 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3780 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3781 clear_vfp_multiple pattern based on a new vfp parameter.
3782 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3783 targeting Armv8.1-M Mainline.
3784 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3785 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3786 whether VFP registers are available before looking call_used_regs for a
3788 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3789 of prototype of clear_operation_p.
3790 (clear_vfp_multiple_operation): New predicate.
3791 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3792 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3794 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3795 Thomas Preud'homme <thomas.preudhomme@arm.com>
3797 * config/arm/arm-protos.h (clear_operation_p): Declare.
3798 * config/arm/arm.c (clear_operation_p): New function.
3799 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3800 targeting Armv8.1-M Mainline or successor.
3801 (output_return_instruction): Only output APSR register clearing if
3802 Armv8.1-M Mainline instructions not available.
3803 (thumb_exit): Likewise.
3804 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3805 * config/arm/thumb2.md (clear_apsr): New define_insn.
3806 (clear_multiple): Likewise.
3807 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3809 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3810 Thomas Preud'homme <thomas.preudhomme@arm.com>
3812 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3813 (use_return_insn): Also return false for Armv8.1-M Mainline.
3814 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3815 Mainline instructions are available.
3816 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3817 when targeting Armv8.1-M Mainline Security Extensions.
3818 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3819 Mainline entry function.
3820 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3821 targeting Armv8.1-M Mainline or successor.
3822 (arm_expand_epilogue): Fix indentation of caller-saved register
3823 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3825 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3826 (FP_SYSREGS): Likewise.
3827 (enum vfp_sysregs_encoding): Define enum.
3828 (fp_sysreg_names): Declare.
3829 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3830 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3831 (pop_fpsysreg_insn): Likewise.
3833 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3834 Thomas Preud'homme <thomas.preudhomme@arm.com>
3836 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3837 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3838 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3839 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3840 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3841 (ARMv8_1m_main): New feature group.
3842 (armv8.1-m.main): New architecture.
3843 * config/arm/arm-tables.opt: Regenerate.
3844 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3845 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3846 (arm_options_perform_arch_sanity_checks): Error out when targeting
3847 Armv8.1-M Mainline Security Extensions.
3848 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3850 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3852 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3853 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3854 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3855 aarch64_bfdot_laneq): New.
3856 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3857 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3858 vbfdotq_laneq_f32): New.
3859 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3860 VBFMLA_W, VBF): New.
3861 (isquadop): Add V4BF, V8BF.
3863 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3865 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3866 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3867 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3868 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3869 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3870 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3871 usdot_laneq, sudot_lane,sudot_laneq): New.
3872 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3873 (aarch64_<sur>dot_lane): New.
3874 * config/aarch64/arm_neon.h (vusdot_s32): New.
3876 (vusdot_lane_s32): New.
3877 (vsudot_lane_s32): New.
3878 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3879 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3881 2020-01-16 Martin Liska <mliska@suse.cz>
3883 * value-prof.c (dump_histogram_value): Fix
3884 obvious spacing issue.
3886 2020-01-16 Andrew Pinski <apinski@marvell.com>
3888 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3889 !storage_order_barrier_p.
3891 2020-01-16 Andrew Pinski <apinski@marvell.com>
3893 * sched-int.h (_dep): Add unused bit-field field for the padding.
3894 * sched-deps.c (init_dep_1): Init unused field.
3896 2020-01-16 Andrew Pinski <apinski@marvell.com>
3898 * optabs.h (create_expand_operand): Initialize target field also.
3900 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3902 PR tree-optimization/92429
3903 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3904 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3906 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3909 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3911 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3912 aarch64_sve_int_mode to each mode.
3914 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3916 * doc/analyzer.texi (Overview): Add note about
3917 -fdump-ipa-analyzer.
3919 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3921 PR tree-optimization/93231
3922 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3923 input_type is unsigned. Use tree_to_shwi for shift constant.
3924 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3925 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3926 inputs more efficiently.
3928 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3930 * config/i386/i386.md (*movsf_internal): Do not require
3931 SSE2 ISA for alternatives 14 and 15.
3933 2020-01-15 Richard Biener <rguenther@suse.de>
3936 * tree-eh.c (sink_clobbers): If we already visited the destination
3937 block do not defer insertion.
3938 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3939 the purpose of defered insertion.
3941 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3943 * BASE-VER: Bump to 10.0.1.
3945 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3947 PR tree-optimization/93247
3948 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3949 type of the stmt that we're going to vectorize.
3951 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3953 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3954 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3957 2020-01-15 Martin Liska <mliska@suse.cz>
3959 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3960 2 calls of streamer_read_hwi in a function call.
3962 2020-01-15 Richard Biener <rguenther@suse.de>
3964 * alias.c (record_alias_subset): Avoid redundant work when
3965 subset is already recorded.
3967 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3969 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3970 the analyzer options provide CWE identifiers.
3972 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3974 * tree-diagnostic-path.cc (path_summary::event_range::print):
3975 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3976 using get_pure_location.
3978 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3980 PR tree-optimization/93262
3981 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3982 perform head trimming only if the last argument is constant,
3983 either all ones, or larger or equal to head trim, in the latter
3984 case decrease the last argument by head_trim.
3986 PR tree-optimization/93249
3987 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3988 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3989 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3990 perform head trim unless we can prove there are no '\0' chars
3991 from the source among the first head_trim chars.
3993 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3995 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3997 2020-01-15 Jakub Jelinek <jakub@redhat.com>
4000 * config/i386/sse.md
4001 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
4002 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
4003 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
4004 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
4005 just a single alternative instead of two, make operands 1 and 2
4008 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
4011 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
4014 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4016 * Makefile.in (lang_opt_files): Add analyzer.opt.
4017 (ANALYZER_OBJS): New.
4018 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
4019 tristate.o and ANALYZER_OBJS.
4020 (TEXI_GCCINT_FILES): Add analyzer.texi.
4021 * common.opt (-fanalyzer): New driver option.
4022 * config.in: Regenerate.
4023 * configure: Regenerate.
4024 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
4025 (gccdepdir): Also create depdir for "analyzer" subdir.
4026 * digraph.cc: New file.
4027 * digraph.h: New file.
4028 * doc/analyzer.texi: New file.
4029 * doc/gccint.texi ("Static Analyzer") New menu item.
4030 (analyzer.texi): Include it.
4031 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
4032 ("Warning Options"): Add static analysis warnings to the list.
4033 (-Wno-analyzer-double-fclose): New option.
4034 (-Wno-analyzer-double-free): New option.
4035 (-Wno-analyzer-exposure-through-output-file): New option.
4036 (-Wno-analyzer-file-leak): New option.
4037 (-Wno-analyzer-free-of-non-heap): New option.
4038 (-Wno-analyzer-malloc-leak): New option.
4039 (-Wno-analyzer-possible-null-argument): New option.
4040 (-Wno-analyzer-possible-null-dereference): New option.
4041 (-Wno-analyzer-null-argument): New option.
4042 (-Wno-analyzer-null-dereference): New option.
4043 (-Wno-analyzer-stale-setjmp-buffer): New option.
4044 (-Wno-analyzer-tainted-array-index): New option.
4045 (-Wno-analyzer-use-after-free): New option.
4046 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
4047 (-Wno-analyzer-use-of-uninitialized-value): New option.
4048 (-Wanalyzer-too-complex): New option.
4049 (-fanalyzer-call-summaries): New warning.
4050 (-fanalyzer-checker=): New warning.
4051 (-fanalyzer-fine-grained): New warning.
4052 (-fno-analyzer-state-merge): New warning.
4053 (-fno-analyzer-state-purge): New warning.
4054 (-fanalyzer-transitivity): New warning.
4055 (-fanalyzer-verbose-edges): New warning.
4056 (-fanalyzer-verbose-state-changes): New warning.
4057 (-fanalyzer-verbosity=): New warning.
4058 (-fdump-analyzer): New warning.
4059 (-fdump-analyzer-callgraph): New warning.
4060 (-fdump-analyzer-exploded-graph): New warning.
4061 (-fdump-analyzer-exploded-nodes): New warning.
4062 (-fdump-analyzer-exploded-nodes-2): New warning.
4063 (-fdump-analyzer-exploded-nodes-3): New warning.
4064 (-fdump-analyzer-supergraph): New warning.
4065 * doc/sourcebuild.texi (dg-require-dot): New.
4066 (dg-check-dot): New.
4067 * gdbinit.in (break-on-saved-diagnostic): New command.
4068 * graphviz.cc: New file.
4069 * graphviz.h: New file.
4070 * ordered-hash-map-tests.cc: New file.
4071 * ordered-hash-map.h: New file.
4072 * passes.def (pass_analyzer): Add before
4073 pass_ipa_whole_program_visibility.
4074 * selftest-run-tests.c (selftest::run_tests): Call
4075 selftest::ordered_hash_map_tests_cc_tests.
4076 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
4078 * shortest-paths.h: New file.
4079 * timevar.def (TV_ANALYZER): New timevar.
4080 (TV_ANALYZER_SUPERGRAPH): Likewise.
4081 (TV_ANALYZER_STATE_PURGE): Likewise.
4082 (TV_ANALYZER_PLAN): Likewise.
4083 (TV_ANALYZER_SCC): Likewise.
4084 (TV_ANALYZER_WORKLIST): Likewise.
4085 (TV_ANALYZER_DUMP): Likewise.
4086 (TV_ANALYZER_DIAGNOSTICS): Likewise.
4087 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
4088 * tree-pass.h (make_pass_analyzer): New decl.
4089 * tristate.cc: New file.
4090 * tristate.h: New file.
4092 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
4095 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
4096 alternatives 9 and 10.
4098 2020-01-14 David Malcolm <dmalcolm@redhat.com>
4100 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
4101 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
4102 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
4103 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
4104 (selftest::hash_map_tests_c_tests): Call it.
4105 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
4106 New static constant, using the value of = H::empty_zero_p.
4107 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
4108 from default_hash_traits <Value>.
4109 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
4111 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
4112 * hash-table.h (hash_table::alloc_entries): Guard the loop of
4113 calls to mark_empty with !Descriptor::empty_zero_p.
4114 (hash_table::empty_slow): Conditionalize the memset call with a
4115 check that Descriptor::empty_zero_p; otherwise, loop through the
4116 entries calling mark_empty on them.
4117 * hash-traits.h (int_hash::empty_zero_p): New static constant.
4118 (pointer_hash::empty_zero_p): Likewise.
4119 (pair_hash::empty_zero_p): Likewise.
4120 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
4122 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
4123 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
4124 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
4125 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
4126 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
4127 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
4128 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
4129 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
4131 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
4134 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
4136 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
4137 fix typo on return value.
4139 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
4142 * cgraph.c (symbol_table::create_edge): Init speculative_id and
4144 (cgraph_edge::make_speculative): Add param for setting speculative_id
4146 (cgraph_edge::speculative_call_info): Update comments and find reference
4147 by speculative_id for multiple indirect targets.
4148 (cgraph_edge::resolve_speculation): Decrease the speculations
4149 for indirect edge, drop it's speculative if not direct target
4150 left. Update comments.
4151 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4152 (cgraph_node::dump): Print num_speculative_call_targets.
4153 (cgraph_node::verify_node): Don't report error if speculative
4154 edge not include statement.
4155 (cgraph_edge::num_speculative_call_targets_p): New function.
4156 * cgraph.h (int common_target_id): Remove.
4157 (int common_target_probability): Remove.
4158 (num_speculative_call_targets): New variable.
4159 (make_speculative): Add param for setting speculative_id.
4160 (cgraph_edge::num_speculative_call_targets_p): New declare.
4161 (target_prob): New variable.
4162 (speculative_id): New variable.
4163 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
4164 call summaries for multiple speculative call targets.
4165 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
4166 * ipa-profile.c (struct speculative_call_target): New struct.
4167 (class speculative_call_summary): New class.
4168 (class speculative_call_summaries): New class.
4169 (call_sums): New variable.
4170 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
4171 (ipa_profile_write_edge_summary): New function.
4172 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
4173 (ipa_profile_dump_all_summaries): New function.
4174 (ipa_profile_read_edge_summary): New function.
4175 (ipa_profile_read_summary_section): New function.
4176 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
4177 (ipa_profile): Generate num_speculative_call_targets from
4179 * ipa-ref.h (speculative_id): New variable.
4180 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
4181 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
4182 common_target_probability. Stream out speculative_id and
4183 num_speculative_call_targets.
4184 (input_edge): Likewise.
4185 * predict.c (dump_prediction): Remove edges count assert to be
4187 * symtab.c (symtab_node::create_reference): Init speculative_id.
4188 (symtab_node::clone_references): Clone speculative_id.
4189 (symtab_node::clone_referring): Clone speculative_id.
4190 (symtab_node::clone_reference): Clone speculative_id.
4191 (symtab_node::clear_stmts_in_references): Clear speculative_id.
4192 * tree-inline.c (copy_bb): Duplicate all the speculative edges
4193 if indirect call contains multiple speculative targets.
4194 * value-prof.h (check_ic_target): Remove.
4195 * value-prof.c (gimple_value_profile_transformations):
4196 Use void function gimple_ic_transform.
4197 * value-prof.c (gimple_ic_transform): Handle topn case.
4198 Fix comment typos. Change it to a void function.
4200 2020-01-13 Andrew Pinski <apinski@marvell.com>
4202 * config/aarch64/aarch64-cores.def (octeontx2): New define.
4203 (octeontx2t98): New define.
4204 (octeontx2t96): New define.
4205 (octeontx2t93): New define.
4206 (octeontx2f95): New define.
4207 (octeontx2f95n): New define.
4208 (octeontx2f95mm): New define.
4209 * config/aarch64/aarch64-tune.md: Regenerate.
4210 * doc/invoke.texi (-mcpu=): Document the new cpu types.
4212 2020-01-13 Jason Merrill <jason@redhat.com>
4214 PR c++/33799 - destroy return value if local cleanup throws.
4215 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
4217 2020-01-13 Martin Liska <mliska@suse.cz>
4219 * ipa-cp.c (get_max_overall_size): Use newly
4220 renamed param param_ipa_cp_unit_growth.
4221 * params.opt: Remove legacy param name.
4223 2020-01-13 Martin Sebor <msebor@redhat.com>
4225 PR tree-optimization/93213
4226 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
4227 stores to be eliminated.
4229 2020-01-13 Martin Liska <mliska@suse.cz>
4231 * opts.c (print_help): Do not print CL_PARAM
4232 and CL_WARNING for CL_OPTIMIZATION.
4234 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
4237 * doc/invoke.texi (Warning Options): Add caveat about some warnings
4238 depending on optimization settings.
4240 2020-01-13 Jakub Jelinek <jakub@redhat.com>
4242 PR tree-optimization/90838
4243 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4244 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
4245 argument rather than to initialize temporary for targets that
4246 don't use the mode argument at all. Initialize ctzval to avoid
4249 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
4251 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
4252 * tree-core.h: Document it.
4253 * gimplify.c (gimplify_omp_workshare): Set it.
4254 * omp-low.c (lower_omp_target): Use it.
4255 * tree-pretty-print.c (dump_omp_clause): Print it.
4257 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
4258 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
4260 2020-01-10 David Malcolm <dmalcolm@redhat.com>
4262 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
4263 * common.opt (fdiagnostics-path-format=): New option.
4264 (diagnostic_path_format): New enum.
4265 (fdiagnostics-show-path-depths): New option.
4266 * coretypes.h (diagnostic_event_id_t): New forward decl.
4267 * diagnostic-color.c (color_dict): Add "path".
4268 * diagnostic-event-id.h: New file.
4269 * diagnostic-format-json.cc (json_from_expanded_location): Make
4271 (json_end_diagnostic): Call context->make_json_for_path if it
4272 exists and the diagnostic has a path.
4273 (diagnostic_output_format_init): Clear context->print_path.
4274 * diagnostic-path.h: New file.
4275 * diagnostic-show-locus.c (colorizer::set_range): Special-case
4276 when printing a run of events in a diagnostic_path so that they
4277 all get the same color.
4278 (layout::m_diagnostic_path_p): New field.
4279 (layout::layout): Initialize it.
4280 (layout::print_any_labels): Don't colorize the label text for an
4281 event in a diagnostic_path.
4282 (gcc_rich_location::add_location_if_nearby): Add
4283 "restrict_to_current_line_spans" and "label" params. Pass the
4284 former to layout.maybe_add_location_range; pass the latter
4285 when calling add_range.
4286 * diagnostic.c: Include "diagnostic-path.h".
4287 (diagnostic_initialize): Initialize context->path_format and
4288 context->show_path_depths.
4289 (diagnostic_show_any_path): New function.
4290 (diagnostic_path::interprocedural_p): New function.
4291 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
4292 (simple_diagnostic_path::num_events): New function.
4293 (simple_diagnostic_path::get_event): New function.
4294 (simple_diagnostic_path::add_event): New function.
4295 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
4296 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
4297 (debug): New overload taking a diagnostic_path *.
4298 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
4299 * diagnostic.h (enum diagnostic_path_format): New enum.
4300 (json::value): New forward decl.
4301 (diagnostic_context::path_format): New field.
4302 (diagnostic_context::show_path_depths): New field.
4303 (diagnostic_context::print_path): New callback field.
4304 (diagnostic_context::make_json_for_path): New callback field.
4305 (diagnostic_show_any_path): New decl.
4306 (json_from_expanded_location): New decl.
4307 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
4308 (-fdiagnostics-show-path-depths): New option.
4309 (-fdiagnostics-color): Add "path" to description of default
4310 GCC_COLORS; describe it.
4311 (-fdiagnostics-format=json): Document how diagnostic paths are
4312 represented in the JSON output format.
4313 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4314 Add optional params "restrict_to_current_line_spans" and "label".
4315 * opts.c (common_handle_option): Handle
4316 OPT_fdiagnostics_path_format_ and
4317 OPT_fdiagnostics_show_path_depths.
4318 * pretty-print.c: Include "diagnostic-event-id.h".
4319 (pp_format): Implement "%@" format code for printing
4320 diagnostic_event_id_t *.
4321 (selftest::test_pp_format): Add tests for "%@".
4322 * selftest-run-tests.c (selftest::run_tests): Call
4323 selftest::tree_diagnostic_path_cc_tests.
4324 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
4325 * toplev.c (general_init): Initialize global_dc->path_format and
4326 global_dc->show_path_depths.
4327 * tree-diagnostic-path.cc: New file.
4328 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
4329 non-static. Drop "diagnostic" param in favor of storing the
4330 original value of "where" and re-using it.
4331 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
4332 maybe_unwind_expanded_macro_loc.
4333 (tree_diagnostics_defaults): Initialize context->print_path and
4334 context->make_json_for_path.
4335 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
4337 (default_tree_make_json_for_path): New decl.
4338 (maybe_unwind_expanded_macro_loc): New decl.
4340 2020-01-10 Jakub Jelinek <jakub@redhat.com>
4342 PR tree-optimization/93210
4343 * fold-const.h (native_encode_initializer,
4344 can_native_interpret_type_p): Declare.
4345 * fold-const.c (native_encode_string): Fix up handling with off != -1,
4347 (native_encode_initializer): New function, moved from dwarf2out.c.
4348 Adjust to native_encode_expr compatible arguments, including dry-run
4349 and partial extraction modes. Don't handle STRING_CST.
4350 (can_native_interpret_type_p): No longer static.
4351 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
4352 offset / BITS_PER_UNIT fits into int and don't call it if
4353 can_native_interpret_type_p fails. If suboff is NULL and for
4354 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
4355 native_encode_initializer.
4356 (fold_const_aggregate_ref_1): Formatting fix.
4357 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
4358 (tree_add_const_value_attribute): Adjust caller.
4360 PR tree-optimization/90838
4361 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
4362 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
4363 CTZ_DEFINED_VALUE_AT_ZERO.
4365 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
4368 * lra-constraints.c (match_reload): Permit input operands have the
4369 same mode as output while other input operands have a different
4372 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
4374 PR tree-optimization/90838
4375 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
4376 (check_ctz_string): Likewise.
4377 (optimize_count_trailing_zeroes): Likewise.
4378 (simplify_count_trailing_zeroes): Likewise.
4379 (pass_forwprop::execute): Try ctz simplification.
4380 * match.pd: Add matching for ctz idioms.
4382 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4384 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
4386 (aarch64_invalid_unary_op): New function for target hook.
4387 (aarch64_invalid_binary_op): New function for target hook.
4389 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
4391 * config.gcc: Add arm_bf16.h.
4392 * config/aarch64/aarch64-builtins.c
4393 (aarch64_simd_builtin_std_type): Add BFmode.
4394 (aarch64_init_simd_builtin_types): Define element types for vector
4396 (aarch64_init_bf16_types): New function.
4397 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
4398 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
4400 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
4401 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
4403 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
4404 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
4405 * config/aarch64/aarch64.c
4406 (aarch64_classify_vector_mode): Add support for BF types.
4407 (aarch64_gimplify_va_arg_expr): Add support for BF types.
4408 (aarch64_vq_mode): Add support for BF types.
4409 (aarch64_simd_container_mode): Add support for BF types.
4410 (aarch64_mangle_type): Add support for BF scalar type.
4411 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
4412 * config/aarch64/arm_bf16.h: New file.
4413 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
4414 * config/aarch64/iterators.md: Add BF types to mode attributes.
4415 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
4417 2020-01-10 Jason Merrill <jason@redhat.com>
4419 PR c++/93173 - incorrect tree sharing.
4420 * gimplify.c (copy_if_shared): No longer static.
4421 * gimplify.h: Declare it.
4423 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4425 * doc/invoke.texi (-msve-vector-bits=): Document that
4426 -msve-vector-bits=128 now generates VL-specific code for
4427 little-endian targets.
4428 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
4429 build_vector_type_for_mode to construct the data vector types.
4430 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
4431 VL-specific code for -msve-vector-bits=128 on little-endian targets.
4432 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
4433 for 128-bit vectors.
4435 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4437 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
4440 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4442 * config/aarch64/aarch64-builtins.c
4443 (aarch64_builtin_vectorized_function): Check for specific vector modes,
4444 rather than checking the number of elements and the element mode.
4446 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4448 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
4449 get_related_vectype_for_scalar_type rather than build_vector_type
4450 to create the index type for a conditional reduction.
4452 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
4454 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
4455 for any type of gather or scatter, including strided accesses.
4457 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4459 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
4462 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
4464 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
4466 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
4467 parameter and its use to reset DR_OFFSET's.
4468 (vect_transform_loop): Remove orig_drs_init argument.
4469 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
4470 member of dr_vec_info rather than the offset of the associated
4471 data_reference's innermost_loop_behavior.
4472 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
4473 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
4474 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
4475 get_dr_vinfo_offset.
4476 (vectorizable_store): Likewise.
4477 (vectorizable_load): Likewise.
4479 2020-01-10 Richard Biener <rguenther@suse.de>
4481 * gimple-ssa-store-merging
4482 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
4484 2020-01-10 Martin Liska <mliska@suse.cz>
4487 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
4488 encapsulation that was there before r280040.
4490 2020-01-10 Richard Biener <rguenther@suse.de>
4493 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
4494 sequences to avoid walking them again for secondary opportunities.
4495 (pass_lower_eh_dispatch::execute): Instead actually insert
4498 2020-01-10 Richard Biener <rguenther@suse.de>
4501 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
4502 (cleanup_all_empty_eh): Walk landing pads in reverse order to
4503 avoid quadraticness.
4505 2020-01-10 Martin Jambor <mjambor@suse.cz>
4507 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
4508 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
4509 to get param_ipa_sra_max_replacements.
4510 (param_splitting_across_edge): Pass the caller to
4511 pull_accesses_from_callee.
4513 2020-01-10 Martin Jambor <mjambor@suse.cz>
4515 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
4516 * ipa-cp.c (max_new_size): Removed.
4517 (orig_overall_size): New variable.
4518 (get_max_overall_size): New function.
4519 (estimate_local_effects): Use it. Adjust dump.
4520 (decide_about_value): Likewise.
4521 (ipcp_propagate_stage): Do not calculate max_new_size, just store
4522 orig_overall_size. Adjust dump.
4523 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
4525 2020-01-10 Martin Jambor <mjambor@suse.cz>
4527 * params.opt (param_ipa_max_agg_items): Mark as Optimization
4528 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
4529 instead of param_ipa_max_agg_items.
4530 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
4531 optimization info for the callee.
4533 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
4535 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
4536 markers if debug_inline_points is false.
4538 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4540 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
4542 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
4543 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
4544 aarch64-sve-builtins-sve2.h.
4545 (aarch64-sve-builtins-sve2.o): New rule.
4546 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
4547 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
4548 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
4549 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
4550 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4551 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
4553 * config/aarch64/aarch64-sve.md: Update comments with SVE2
4554 instructions that are handled here.
4555 (@cond_asrd<mode>): Generalize to...
4556 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
4557 (*cond_asrd<mode>_2): Generalize to...
4558 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
4559 (*cond_asrd<mode>_z): Generalize to...
4560 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
4561 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
4562 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
4563 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
4564 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
4566 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4567 (@aarch64_scatter_stnt<mode>): Likewise.
4568 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
4569 (@aarch64_mul_lane_<mode>): Likewise.
4570 (@aarch64_sve_suqadd<mode>_const): Likewise.
4571 (*<sur>h<addsub><mode>): Generalize to...
4572 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
4574 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
4575 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
4576 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
4577 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
4578 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
4579 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
4580 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
4581 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
4582 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
4583 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
4584 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
4585 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
4586 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
4587 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
4588 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
4589 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
4590 (@aarch64_sve2_xar<mode>): Likewise.
4591 (@aarch64_sve2_bcax<mode>): Likewise.
4592 (*aarch64_sve2_eor3<mode>): Rename to...
4593 (@aarch64_sve2_eor3<mode>): ...this.
4594 (@aarch64_sve2_bsl<mode>): New expander.
4595 (@aarch64_sve2_nbsl<mode>): Likewise.
4596 (@aarch64_sve2_bsl1n<mode>): Likewise.
4597 (@aarch64_sve2_bsl2n<mode>): Likewise.
4598 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
4599 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
4600 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
4601 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
4602 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
4603 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
4604 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
4605 (<su>mull<bt><Vwide>): Generalize to...
4606 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
4608 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
4609 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
4610 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
4611 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4612 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
4613 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4614 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
4615 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4616 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
4617 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
4618 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
4619 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
4620 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
4621 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
4622 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
4623 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
4624 (<SHRNB:r>shrnb<mode>): Generalize to...
4625 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
4627 (<SHRNT:r>shrnt<mode>): Generalize to...
4628 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
4630 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
4631 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
4632 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
4633 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
4634 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4635 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4636 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4637 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4638 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4639 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4640 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4641 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4642 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4643 (@aarch64_sve2_cvtnt<mode>): Likewise.
4644 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4645 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4646 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4647 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4648 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4649 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4650 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4651 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4652 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4653 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4654 (@aarch64_sve2_pmul<mode>): Likewise.
4655 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4656 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4657 (@aarch64_sve2_tbl2<mode>): Likewise.
4658 (@aarch64_sve2_tbx<mode>): Likewise.
4659 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4660 (@aarch64_sve2_histcnt<mode>): Likewise.
4661 (@aarch64_sve2_histseg<mode>): Likewise.
4662 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4663 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4664 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4665 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4666 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4667 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4668 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4669 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4670 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4671 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4672 (SVE2_PMULL_PAIR_I): New mode iterators.
4673 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4674 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4675 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4676 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4677 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4678 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4679 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4680 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4681 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4682 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4683 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4684 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4685 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4686 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4687 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4688 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4689 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4690 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4691 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4692 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4693 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4694 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4695 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4696 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4697 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4698 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4699 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4700 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4701 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4702 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4703 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4705 (VNARROW, Ventype): New mode attributes.
4706 (Vewtype): Handle VNx2DI. Fix typo in comment.
4707 (VDOUBLE): New mode attribute.
4708 (sve_lane_con): Handle VNx8HI.
4709 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4710 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4711 (sve_int_op, sve_int_op_rev): Handle the above codes.
4712 (sve_pred_int_rhs2_operand): Likewise.
4713 (MULLBT, SHRNB, SHRNT): Delete.
4714 (SVE_INT_SHIFT_IMM): New int iterator.
4715 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4716 and UNSPEC_WHILEHS for TARGET_SVE2.
4717 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4718 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4719 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4720 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4721 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4722 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4723 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4724 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4725 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4726 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4727 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4728 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4729 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4730 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4731 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4732 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4733 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4734 (optab): Handle the new unspecs.
4735 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4737 (lr): Handle the new unspecs.
4739 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4740 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4741 (sve_int_qsub_op): New int attributes.
4742 (sve_fp_op, rot): Handle the new unspecs.
4743 * config/aarch64/aarch64-sve-builtins.h
4744 (function_resolver::require_matching_pointer_type): Declare.
4745 (function_resolver::resolve_unary): Add an optional boolean argument.
4746 (function_resolver::finish_opt_n_resolution): Add an optional
4747 type_suffix_index argument.
4748 (gimple_folder::redirect_call): Declare.
4749 (gimple_expander::prepare_gather_address_operands): Add an optional
4751 * config/aarch64/aarch64-sve-builtins.cc: Include
4752 aarch64-sve-builtins-sve2.h.
4753 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4754 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4755 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4756 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4757 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4758 (TYPES_s_unsigned): Likewise.
4759 (TYPES_s_integer): Use TYPES_s_unsigned.
4760 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4761 (TYPES_sd_integer): Use them.
4762 (TYPES_d_unsigned): New macro.
4763 (TYPES_d_integer): Use it.
4764 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4765 (TYPES_cvt_narrow): Likewise.
4766 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4767 (preds_mx): New variable.
4768 (function_builder::add_overloaded_function): Allow the new feature
4769 set to be more restrictive than the original one.
4770 (function_resolver::infer_pointer_type): Remove qualifiers from
4771 the pointer type before printing it.
4772 (function_resolver::require_matching_pointer_type): New function.
4773 (function_resolver::resolve_sv_displacement): Handle functions
4774 that don't support 32-bit vector indices or svint32_t vector offsets.
4775 (function_resolver::finish_opt_n_resolution): Take the inferred type
4776 as a separate argument.
4777 (function_resolver::resolve_unary): Optionally treat all forms in
4778 the same way as normal merging functions.
4779 (gimple_folder::redirect_call): New function.
4780 (function_expander::prepare_gather_address_operands): Add an argument
4781 that says whether scaled forms are available. If they aren't,
4782 handle scaling of vector indices and don't add the extension and
4784 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4785 fall back to using cond_* instead.
4786 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4787 Split out the member variables into...
4788 (rtx_code_function_base): ...this new base class.
4789 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4790 (unspec_based_function): Split out the member variables into...
4791 (unspec_based_function_base): ...this new base class.
4792 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4793 (unspec_based_function_exact_insn): New class.
4794 (unspec_based_add_function, unspec_based_add_lane_function)
4795 (unspec_based_lane_function, unspec_based_pred_function)
4796 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4797 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4798 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4800 (unspec_based_fused_function): New class.
4801 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4802 (unspec_based_fused_lane_function): New class.
4803 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4805 (CODE_FOR_MODE1): New macro.
4806 (fixed_insn_function): New class.
4807 (while_comparison): Likewise.
4808 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4809 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4810 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4811 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4812 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4813 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4814 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4815 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4816 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4817 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4818 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4819 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4820 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4821 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4822 Also add an initial argument for unary_convert_narrowt, regardless
4823 of the predication type.
4824 (build_32_64): Allow loads and stores to specify MODE_none.
4825 (build_sv_index64, build_sv_uint_offset): New functions.
4826 (long_type_suffix): New function.
4827 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4828 (binary_imm_long_base, load_gather_sv_base): Likewise.
4829 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4830 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4831 (unary_narrowb_base, unary_narrowt_base): Likewise.
4832 (binary_long_lane_def, binary_long_lane): New shape.
4833 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4834 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4835 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4836 (binary_to_uint_def, binary_to_uint): Likewise.
4837 (binary_wide_def, binary_wide): Likewise.
4838 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4839 (compare_def, compare): Likewise.
4840 (compare_ptr_def, compare_ptr): Likewise.
4841 (load_ext_gather_index_restricted_def,
4842 load_ext_gather_index_restricted): Likewise.
4843 (load_ext_gather_offset_restricted_def,
4844 load_ext_gather_offset_restricted): Likewise.
4845 (load_gather_sv_def): Inherit from load_gather_sv_base.
4846 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4847 (shift_left_imm_def, shift_left_imm): Likewise.
4848 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4849 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4850 (store_scatter_index_restricted_def,
4851 store_scatter_index_restricted): Likewise.
4852 (store_scatter_offset_restricted_def,
4853 store_scatter_offset_restricted): Likewise.
4854 (tbl_tuple_def, tbl_tuple): Likewise.
4855 (ternary_long_lane_def, ternary_long_lane): Likewise.
4856 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4857 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4858 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4859 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4860 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4861 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4862 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4863 (ternary_uint_def, ternary_uint): Likewise.
4864 (unary_convert): Fix typo in comment.
4865 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4866 (unary_long_def, unary_long): Likewise.
4867 (unary_narrowb_def, unary_narrowb): Likewise.
4868 (unary_narrowt_def, unary_narrowt): Likewise.
4869 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4870 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4871 (unary_to_int_def, unary_to_int): Likewise.
4872 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4873 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4874 (svasrd_impl): Delete.
4875 (svcadd_impl::expand): Handle integer operations too.
4876 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4877 new functions to derive the unspec numbers.
4878 (svmla_svmls_lane_impl): Replace with...
4879 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4880 integer operations too.
4881 (svwhile_impl): Rename to...
4882 (svwhilelx_impl): ...this and inherit from while_comparison.
4883 (svasrd): Use unspec_based_function.
4884 (svmla_lane): Use svmla_lane_impl.
4885 (svmls_lane): Use svmls_lane_impl.
4886 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4887 (svwhilele, svwhilelt): Use svwhilelx_impl.
4888 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4889 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4890 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4891 * config/aarch64/aarch64-sve-builtins.def: Include
4892 aarch64-sve-builtins-sve2.def.
4894 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4896 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4897 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4898 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4899 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4900 immediates as well as vector ones.
4901 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4902 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4903 (aarch64_sve_qsub_immediate): Update calls accordingly.
4905 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4907 * config/aarch64/aarch64-sve2.md: Add banner comments.
4908 (<su>mulh<r>s<mode>3): Move further up file.
4909 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4910 (*aarch64_sve2_sra<mode>): Move further down file.
4911 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4913 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4915 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4917 (while_optab_cmp): Handle them.
4918 * config/aarch64/aarch64-sve.md
4919 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4920 and add a "@" marker.
4921 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4922 instead of gen_aarch64_sve2_while_ptest.
4923 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4925 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4927 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4928 (UNSPEC_WHILELE): ...this.
4929 (UNSPEC_WHILE_LO): Rename to...
4930 (UNSPEC_WHILELO): ...this.
4931 (UNSPEC_WHILE_LS): Rename to...
4932 (UNSPEC_WHILELS): ...this.
4933 (UNSPEC_WHILE_LT): Rename to...
4934 (UNSPEC_WHILELT): ...this.
4935 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4936 (cmp_op, while_optab_cmp): Likewise.
4937 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4938 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4939 (svwhilelt): Likewise.
4941 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4943 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4944 (unary_to_uint): Define.
4945 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4946 (unary_count): Rename to...
4947 (unary_to_uint_def, unary_to_uint): ...this.
4948 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4950 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4952 * config/aarch64/aarch64-sve-builtins-functions.h
4953 (code_for_mode_function): New class.
4954 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4955 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4956 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4957 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4958 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4960 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4962 * config/aarch64/iterators.md (addsub): New code attribute.
4963 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4965 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4966 in the asm string and attributes. Fix indentation.
4967 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4969 (@aarch64_sve_<optab><mode>): ...this.
4970 * config/aarch64/aarch64-sve-builtins.h
4971 (function_expander::expand_signed_unpred_op): Delete.
4972 * config/aarch64/aarch64-sve-builtins.cc
4973 (function_expander::expand_signed_unpred_op): Likewise.
4974 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4975 try using code_for_aarch64_sve instead.
4976 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4977 (svqsub_impl): Likewise.
4978 (svqadd, svqsub): Use rtx_code_function instead.
4980 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4982 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4983 (HADDSUB, sur, addsub): Remove them.
4985 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4987 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4988 functions the same way, rather than singling out those that
4989 aren't mapped directly to optabs.
4991 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4993 * target.def (compatible_vector_types_p): New target hook.
4994 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4995 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4996 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4997 * doc/tm.texi: Regenerate.
4998 * gimple-expr.c: Include target.h.
4999 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
5000 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
5002 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
5003 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
5004 Use the original predicate if it already has a suitable type.
5006 2020-01-09 Martin Jambor <mjambor@suse.cz>
5008 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
5009 resolve_speculation and redirect_call_stmt_to_callee static. Change
5010 return type of set_call_stmt to cgraph_edge *.
5011 * auto-profile.c (afdo_indirect_call): Adjust call to
5012 redirect_call_stmt_to_callee.
5013 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
5014 make the this pointer explicit, adjust self-recursive calls and the
5015 call top make_direct. Return the resulting edge.
5016 (cgraph_edge::remove): Make this pointer explicit.
5017 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
5018 (cgraph_edge::make_direct): Likewise, adjust call to
5019 resolve_speculation.
5020 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
5021 call to set_call_stmt.
5022 (cgraph_update_edges_for_call_stmt_node): Update call to
5023 set_call_stmt and remove.
5024 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
5025 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
5026 (cgraph_node::create_edge_including_clones): Moved "first" definition
5027 of edge to the block where it was used. Adjusted calls to
5029 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
5030 cgraph_edge::remove.
5031 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
5032 make_direct and redirect_call_stmt_to_callee.
5033 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
5034 resolve_speculation and make_direct.
5035 * ipa-inline-transform.c (inline_transform): Adjust call to
5036 redirect_call_stmt_to_callee.
5037 (check_speculations_1):: Adjust call to resolve_speculation.
5038 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
5039 resolve-speculation.
5040 (inline_small_functions): Adjust call to resolve_speculation.
5041 (ipa_inline): Likewise.
5042 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
5044 * ipa-visibility.c (function_and_variable_visibility): Make iteration
5045 safe with regards to edge removal, adjust calls to
5046 redirect_call_stmt_to_callee.
5047 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
5048 and redirect_call_stmt_to_callee.
5049 * multiple_target.c (create_dispatcher_calls): Adjust call to
5050 redirect_call_stmt_to_callee
5051 (redirect_to_specific_clone): Likewise.
5052 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
5053 Adjust calls to cgraph_edge::remove.
5054 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
5055 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
5056 (expand_call_inline): Adjust call to cgraph_edge::remove.
5058 2020-01-09 Martin Liska <mliska@suse.cz>
5060 * params.opt: Set Optimization for
5061 param_max_speculative_devirt_maydefs.
5063 2020-01-09 Martin Sebor <msebor@redhat.com>
5067 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
5069 2020-01-09 Martin Liska <mliska@suse.cz>
5071 * auto-profile.c (auto_profile): Use opt_for_fn
5073 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
5074 (propagate_vals_across_arith_jfunc): Likewise.
5075 (hint_time_bonus): Likewise.
5076 (incorporate_penalties): Likewise.
5077 (good_cloning_opportunity_p): Likewise.
5078 (perform_estimation_of_a_value): Likewise.
5079 (estimate_local_effects): Likewise.
5080 (ipcp_propagate_stage): Likewise.
5081 * ipa-fnsummary.c (decompose_param_expr): Likewise.
5082 (set_switch_stmt_execution_predicate): Likewise.
5083 (analyze_function_body): Likewise.
5084 * ipa-inline-analysis.c (offline_size): Likewise.
5085 * ipa-inline.c (early_inliner): Likewise.
5086 * ipa-prop.c (ipa_analyze_node): Likewise.
5087 (ipcp_transform_function): Likewise.
5088 * ipa-sra.c (process_scan_results): Likewise.
5089 (ipa_sra_summarize_function): Likewise.
5090 * params.opt: Rename ipcp-unit-growth to
5091 ipa-cp-unit-growth. Add Optimization for various
5092 IPA-related parameters.
5094 2020-01-09 Richard Biener <rguenther@suse.de>
5097 * gimplify.c (gimplify_expr): Deal with NOP definitions.
5099 2020-01-09 Richard Biener <rguenther@suse.de>
5101 PR tree-optimization/93040
5102 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
5104 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
5106 * common/config/avr/avr-common.c (avr_option_optimization_table)
5107 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
5109 2020-01-09 Martin Liska <mliska@suse.cz>
5111 * cgraphclones.c (symbol_table::materialize_all_clones):
5112 Use cgraph_node::dump_name.
5114 2020-01-09 Jakub Jelinek <jakub@redhat.com>
5117 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
5118 output_operand_lossage instead of gcc_unreachable.
5119 * doc/md.texi (riscv f constraint): Fix typo.
5122 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
5123 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
5124 CONST_SCALAR_INT_P instead of CONST_INT_P.
5125 (*subv<mode>4_1): Rename to ...
5126 (subv<mode>4_1): ... this.
5127 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5128 define_insn_and_split patterns.
5129 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5132 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5134 * vec.c (class selftest::count_dtor): New class.
5135 (selftest::test_auto_delete_vec): New test.
5136 (selftest::vec_c_tests): Call it.
5137 * vec.h (class auto_delete_vec): New class template.
5138 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
5140 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5142 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
5144 2020-01-08 Jim Wilson <jimw@sifive.com>
5146 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
5147 use of TLS_MODEL_LOCAL_EXEC when not pic.
5149 2020-01-08 David Malcolm <dmalcolm@redhat.com>
5151 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
5154 2020-01-08 Jakub Jelinek <jakub@redhat.com>
5157 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
5158 *stack_protect_set_3 peephole2): Also check that the second
5159 insns source is general_operand.
5162 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
5163 predicate for output operand instead of register_operand.
5164 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
5165 memory destination and non-memory operands[2].
5167 2020-01-08 Martin Liska <mliska@suse.cz>
5169 * cgraph.c (cgraph_node::dump): Use ::dump_name or
5170 ::dump_asm_name instead of (::name or ::asm_name).
5171 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
5172 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
5173 (analyze_functions): Likewise.
5174 (expand_all_functions): Likewise.
5175 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5176 (propagate_bits_across_jump_function): Likewise.
5177 (dump_profile_updates): Likewise.
5178 (ipcp_store_bits_results): Likewise.
5179 (ipcp_store_vr_results): Likewise.
5180 * ipa-devirt.c (dump_targets): Likewise.
5181 * ipa-fnsummary.c (analyze_function_body): Likewise.
5182 * ipa-hsa.c (check_warn_node_versionable): Likewise.
5183 (process_hsa_functions): Likewise.
5184 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
5185 (set_alias_uids): Likewise.
5186 * ipa-inline-transform.c (save_inline_function_body): Likewise.
5187 * ipa-inline.c (recursive_inlining): Likewise.
5188 (inline_to_all_callers_1): Likewise.
5189 (ipa_inline): Likewise.
5190 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5191 (ipa_propagate_frequency): Likewise.
5192 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5193 (remove_described_reference): Likewise.
5194 * ipa-pure-const.c (worse_state): Likewise.
5195 (check_retval_uses): Likewise.
5196 (analyze_function): Likewise.
5197 (propagate_pure_const): Likewise.
5198 (propagate_nothrow): Likewise.
5199 (dump_malloc_lattice): Likewise.
5200 (propagate_malloc): Likewise.
5201 (pass_local_pure_const::execute): Likewise.
5202 * ipa-visibility.c (optimize_weakref): Likewise.
5203 (function_and_variable_visibility): Likewise.
5204 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
5205 (ipa_discover_variable_flags): Likewise.
5206 * lto-streamer-out.c (output_function): Likewise.
5207 (output_constructor): Likewise.
5208 * tree-inline.c (copy_bb): Likewise.
5209 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
5210 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
5212 2020-01-08 Richard Biener <rguenther@suse.de>
5215 * tree-eh.c (sink_clobbers): Update virtual operands for
5216 the first and last stmt only. Add a dry-run capability.
5217 (pass_lower_eh_dispatch::execute): Perform clobber sinking
5218 after CFG manipulations and in RPO order to catch all
5219 secondary opportunities reliably.
5221 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5224 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5226 2019-01-08 Richard Biener <rguenther@suse.de>
5229 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
5230 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
5231 virtual operand, also updating SSA use.
5232 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
5233 Update stmt after resetting virtual operand.
5234 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
5235 * gimple-iterator.c (gsi_remove): When not removing the stmt
5236 permanently do not delink immediate uses or mark the stmt modified.
5238 2020-01-08 Martin Liska <mliska@suse.cz>
5240 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
5241 (ipa_call_context::estimate_size_and_time): Likewise.
5242 (inline_analyze_function): Likewise.
5244 2020-01-08 Martin Liska <mliska@suse.cz>
5246 * cgraph.c (cgraph_node::dump): Use systematically
5249 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5251 Add -nodevicespecs option for avr.
5254 * config/avr/avr.opt (-nodevicespecs): New driver option.
5255 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
5256 "-specs=device-specs/..." if that option is not set.
5257 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
5259 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
5261 Implement 64-bit double functions for avr.
5264 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
5265 --with-double-comparison.
5266 * doc/install.texi: Document them.
5267 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
5268 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
5269 <WITH_DOUBLE_COMPARISON>: New built-in defines.
5270 * doc/invoke.texi (AVR Built-in Macros): Document them.
5271 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
5272 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
5273 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
5275 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
5278 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
5279 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
5280 when only building rm-profile multilibs.
5282 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
5285 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
5286 lattice for a value to check.
5287 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
5288 finite propagation in self-recursive scc.
5290 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
5292 * ipa-inline.c (caller_growth_limits): Restore the AND.
5294 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5296 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
5297 (VEC_ALLREG_ALT): New iterator.
5298 (VEC_ALLREG_INT_MODE): New iterator.
5299 (VCMP_MODE): New iterator.
5300 (VCMP_MODE_INT): New iterator.
5301 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
5302 (vec_cmp<u>v64qidi): New define_expand.
5303 (vec_cmp<mode>di_exec): Use VCMP_MODE.
5304 (vec_cmpu<mode>di_exec): New define_expand.
5305 (vec_cmp<u>v64qidi_exec): New define_expand.
5306 (vec_cmp<mode>di_dup): Use VCMP_MODE.
5307 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
5308 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
5309 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
5310 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
5311 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
5312 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
5313 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
5314 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
5315 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
5317 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
5318 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
5320 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
5322 * config/gcn/constraints.md (DA): Update description and match.
5324 (Db): New constraint.
5325 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
5327 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
5328 Implement 'Db' mixed immediate type.
5329 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
5330 (addcv64si3_dup<exec_vcc>): Delete.
5331 (subcv64si3<exec_vcc>): Rework constraints.
5332 (addv64di3): Rework constraints.
5333 (addv64di3_exec): Rework constraints.
5334 (subv64di3): Rework constraints.
5335 (addv64di3_dup): Delete.
5336 (addv64di3_dup_exec): Delete.
5337 (addv64di3_zext): Rework constraints.
5338 (addv64di3_zext_exec): Rework constraints.
5339 (addv64di3_zext_dup): Rework constraints.
5340 (addv64di3_zext_dup_exec): Rework constraints.
5341 (addv64di3_zext_dup2): Rework constraints.
5342 (addv64di3_zext_dup2_exec): Rework constraints.
5343 (addv64di3_sext_dup2): Rework constraints.
5344 (addv64di3_sext_dup2_exec): Rework constraints.
5346 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
5348 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
5349 existing target checks.
5351 2020-01-07 Richard Biener <rguenther@suse.de>
5353 * doc/install.texi: Bump minimal supported MPC version.
5355 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5357 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
5358 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
5359 * langhooks.c: Include stor-layout.h.
5360 (lhd_simulate_enum_decl): New function.
5361 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
5362 handle_arm_sve_h for the LTO frontend.
5363 (register_vector_type): Cope with null returns from pushdecl.
5365 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
5367 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
5368 (aarch64_sve::nvectors_if_data_type): Replace with...
5369 (aarch64_sve::builtin_type_p): ...this.
5370 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
5371 (find_vector_type): Delete.
5372 (add_sve_type_attribute): New function.
5373 (lookup_sve_type_attribute): Likewise.
5374 (register_builtin_types): Add an "SVE type" attribute to each type.
5375 (register_tuple_type): Likewise.
5376 (svbool_type_p, nvectors_if_data_type): Delete.
5377 (mangle_builtin_type): Use lookup_sve_type_attribute.
5378 (builtin_type_p): Likewise. Add an overload that returns the
5379 number of constituent vector and predicate registers.
5380 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
5381 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
5382 instead of aarch64_sve_argument_p.
5383 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
5384 (aarch64_pass_by_reference): Likewise.
5385 (aarch64_function_value_1): Likewise.
5386 (aarch64_return_in_memory): Likewise.
5387 (aarch64_layout_arg): Likewise.
5389 2020-01-07 Jakub Jelinek <jakub@redhat.com>
5391 PR tree-optimization/93156
5392 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
5393 least significant bit is always clear.
5395 PR tree-optimization/93118
5396 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
5397 simplifier with two intermediate conversions.
5399 2020-01-07 Martin Liska <mliska@suse.cz>
5401 * params.opt: Add Optimization for various parameters.
5403 2020-01-07 Martin Liska <mliska@suse.cz>
5406 * doc/extend.texi: Explain cloning for target_clone
5409 2020-01-07 Martin Liska <mliska@suse.cz>
5411 PR tree-optimization/92860
5412 * common.opt: Make in Optimization option
5413 as it is affected by -O0, which is an Optimization
5415 * tree-inline.c (tree_inlinable_function_p):
5416 Use opt_for_fn for warn_inline.
5417 (expand_call_inline): Likewise.
5419 2020-01-07 Martin Liska <mliska@suse.cz>
5421 PR tree-optimization/92860
5422 * common.opt: Make flag_ree as optimization
5425 2020-01-07 Martin Liska <mliska@suse.cz>
5427 PR optimization/92860
5428 * params.opt: Mark param_min_crossjump_insns with Optimization
5431 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
5433 * ipa-inline-analysis.c (estimate_growth): Fix typo.
5434 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
5436 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
5438 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
5439 helper function to return the valid addressing formats for a given
5440 hard register and mode.
5441 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
5443 * config/rs6000/constraints.md (Q constraint): Update
5445 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
5448 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
5449 Use 'Q' for doing vector extract from memory.
5450 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
5452 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
5453 doing vector extract from memory.
5454 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
5455 extract from memory.
5457 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
5458 for the offset being 34-bits when -mcpu=future is used.
5460 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
5462 * config/pa/pa.md: Revert change to use ordered_comparison_operator
5463 instead of cmpib_comparison_operator in cmpib patterns.
5464 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
5465 of cmpib_comparison_operator. Revise comment.
5467 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5469 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
5470 in an IFN_DIV_POW2 node to be equal.
5472 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5474 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
5475 (vect_check_scalar_mask): ...this.
5476 (vectorizable_store, vectorizable_load): Update call accordingly.
5477 (vectorizable_call): Use vect_check_scalar_mask to check the mask
5478 argument in calls to conditional internal functions.
5480 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5482 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
5483 '0' matching inputs.
5484 (subv64di3_exec): Likewise.
5486 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
5488 * config/mips/mips.c (vr4130_align_insns): Fix typo.
5489 * doc/md.texi (movstr): Likewise.
5491 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5493 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
5496 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5498 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
5500 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
5501 to a temporary file and use move-if-change to update the real
5502 file where necessary.
5504 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
5506 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
5507 rather than Upa for CPY /M.
5509 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
5511 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
5514 2020-01-06 Martin Liska <mliska@suse.cz>
5516 PR tree-optimization/92860
5517 * params.opt: Mark param_max_combine_insns with Optimization
5520 2020-01-05 Jakub Jelinek <jakub@redhat.com>
5523 * config/i386/i386.md (SWIDWI): New mode iterator.
5524 (DWI, dwi): Add TImode variants.
5525 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
5526 <general_hilo_operand> instead of <general_operand>. Use
5527 CONST_SCALAR_INT_P instead of CONST_INT_P.
5528 (*addv<mode>4_1): Rename to ...
5529 (addv<mode>4_1): ... this.
5530 (QWI): New mode attribute.
5531 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
5532 define_insn_and_split patterns.
5533 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
5535 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
5536 <general_hilo_operand> instead of <general_operand>.
5537 (*addcarry<mode>_1): New define_insn.
5538 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
5540 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
5542 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
5543 Use "call" instead of "set".
5545 2020-01-03 Martin Jambor <mjambor@suse.cz>
5548 * ipa-cp.c (print_all_lattices): Skip functions without info.
5550 2020-01-03 Jakub Jelinek <jakub@redhat.com>
5553 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
5554 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
5555 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
5556 for 'e' simd clones.
5559 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
5561 (mprefer-vector-width=): Add Save.
5562 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
5563 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
5564 (ix86_debug_options, ix86_function_specific_print): Adjust
5565 ix86_target_string callers.
5566 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
5567 (ix86_valid_target_attribute_tree): Likewise.
5568 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
5569 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
5570 ix86_target_string caller.
5573 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
5574 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
5575 instead of gen_int_shift_amount + convert_modes.
5577 PR rtl-optimization/93088
5578 * loop-iv.c (find_single_def_src): Punt after looking through
5579 128 reg copies for regs with single definitions. Move definitions
5582 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
5584 * config/arm/arm-c.c (arm_cpu_builtins): Define
5585 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
5586 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
5587 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
5588 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
5589 * config/arm/arm-tables.opt: Regenerated.
5590 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
5591 arm_arch_i8mm and arm_arch_bf16 when enabled.
5592 * config/arm/arm.h (TARGET_I8MM): New macro.
5593 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
5594 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
5595 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
5596 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
5597 (v8_6_a_simd_variants): New.
5598 (v8_*_a_simd_variants): Add i8mm and bf16.
5599 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
5601 2020-01-02 Jakub Jelinek <jakub@redhat.com>
5604 * predict.c (compute_function_frequency): Don't call
5605 warn_function_cold on functions that already have cold attribute.
5607 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
5610 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
5611 COMDAT group function labels in .data.rel.ro.local section.
5612 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
5615 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
5616 comparison_operator in B and S integer comparisons. Likewise, use
5617 ordered_comparison_operator instead of cmpib_comparison_operator in
5619 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
5621 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5623 Update copyright years.
5625 * gcc.c (process_command): Update copyright notice dates.
5626 * gcov-dump.c (print_version): Ditto.
5627 * gcov.c (print_version): Ditto.
5628 * gcov-tool.c (print_version): Ditto.
5629 * gengtype.c (create_file): Ditto.
5630 * doc/cpp.texi: Bump @copying's copyright year.
5631 * doc/cppinternals.texi: Ditto.
5632 * doc/gcc.texi: Ditto.
5633 * doc/gccint.texi: Ditto.
5634 * doc/gcov.texi: Ditto.
5635 * doc/install.texi: Ditto.
5636 * doc/invoke.texi: Ditto.
5638 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5640 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5643 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5645 PR tree-optimization/93098
5646 * match.pd (popcount): For shift amounts, use integer_onep
5647 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5648 tests. Make sure that precision is power of two larger than or equal
5649 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5650 instead of ULL suffixed constants. Formatting fixes.
5652 Copyright (C) 2020 Free Software Foundation, Inc.
5654 Copying and distribution of this file, with or without modification,
5655 are permitted in any medium without royalty provided the copyright
5656 notice and this notice are preserved.