1 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
3 * recog.c: Move pass_split_before_sched2 code in front of
4 pass_split_before_regstack.
5 (pass_data_split_before_sched2): Rename pass to split3 from split4.
6 (pass_data_split_before_regstack): Rename pass to split4 from split3.
7 (rest_of_handle_split_before_sched2): Remove.
8 (pass_split_before_sched2::execute): Unconditionally call
10 (enable_split_before_sched2): New function.
11 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
12 (pass_split_before_regstack::gate): Ditto.
13 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
14 Update name check for renamed split4 pass.
15 * config/sh/sh.c (register_sh_passes): Update pass insertion
16 point for renamed split4 pass.
18 2020-02-09 Jakub Jelinek <jakub@redhat.com>
20 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
21 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
22 copying them around between host and target.
24 2020-02-08 Andrew Pinski <apinski@marvell.com>
27 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
28 STRICT_ALIGNMENT also.
30 2020-02-08 Jim Wilson <jimw@sifive.com>
33 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
35 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
36 Jakub Jelinek <jakub@redhat.com>
39 * config/i386/i386.h (CALL_USED_REGISTERS): Make
40 xmm16-xmm31 call-used even in 64-bit ms-abi.
42 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
44 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
45 (simd_ummla, simd_usmmla): Likewise.
46 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
47 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
50 2020-02-07 Richard Biener <rguenther@suse.de>
53 * tree-inline.c (fold_marked_statements): Do a PRE walk,
54 skipping unreachable regions.
55 (optimize_inline_calls): Skip folding stmts when we didn't
58 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
61 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
62 Don't return aggregates with only SFmode and DFmode in SSE
64 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
66 2020-02-07 Jakub Jelinek <jakub@redhat.com>
69 * config/rs6000/rs6000-logue.c
70 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
71 if it fails, move rs into end_addr and retry. Add
72 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
73 the insn pattern doesn't describe well what exactly happens to
77 * config/i386/predicates.md (avx_identity_operand): Remove.
78 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
79 (avx_<castmode><avxsizesuffix>_<castmode>,
80 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
81 a VEC_CONCAT of the operand and UNSPEC_CAST.
82 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
83 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
87 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
88 recog_data.insn if distance_non_agu_define changed it.
90 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
93 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
94 we only had X-FORM (reg+reg) addressing for vectors. Also before
95 ISA 3.0, we only had X-FORM addressing for scalars in the
96 traditional Altivec registers.
98 2020-02-06 <zhongyunde@huawei.com>
99 Vladimir Makarov <vmakarov@redhat.com>
101 PR rtl-optimization/93561
102 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
105 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
107 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
110 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
112 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
113 where the low and the high 32 bits are equal to each other specially,
114 with an rldimi instruction.
116 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
118 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
120 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
122 * config/arm/arm-tables.opt: Regenerate.
124 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
127 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
128 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
129 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
131 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
133 PR rtl-optimization/87763
134 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
136 2020-02-06 Delia Burduv <delia.burduv@arm.com>
138 * config/aarch64/aarch64-simd-builtins.def
139 (bfmlaq): New built-in function.
140 (bfmlalb): New built-in function.
141 (bfmlalt): New built-in function.
142 (bfmlalb_lane): New built-in function.
143 (bfmlalt_lane): New built-in function.
144 * config/aarch64/aarch64-simd.md
145 (aarch64_bfmmlaqv4sf): New pattern.
146 (aarch64_bfmlal<bt>v4sf): New pattern.
147 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
148 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
149 (vbfmlalbq_f32): New intrinsic.
150 (vbfmlaltq_f32): New intrinsic.
151 (vbfmlalbq_lane_f32): New intrinsic.
152 (vbfmlaltq_lane_f32): New intrinsic.
153 (vbfmlalbq_laneq_f32): New intrinsic.
154 (vbfmlaltq_laneq_f32): New intrinsic.
155 * config/aarch64/iterators.md (BF_MLA): New int iterator.
156 (bt): New int attribute.
158 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
160 * config/i386/i386.md (*pushtf): Emit "#" instead of
161 calling gcc_unreachable in insn output.
164 (*pushsf_rex64): Ditto for alternatives other than 1.
165 (*pushsf): Ditto for alternatives other than 1.
167 2020-02-06 Martin Liska <mliska@suse.cz>
169 PR gcov-profile/91971
170 PR gcov-profile/93466
171 * coverage.c (coverage_init): Revert mangling of
172 path into filename. It can lead to huge filename length.
173 Creation of subfolders seem more natural.
175 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
178 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
179 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
180 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
182 2020-02-06 Jakub Jelinek <jakub@redhat.com>
185 * config/i386/predicates.md (avx_identity_operand): New predicate.
186 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
187 define_insn_and_split.
190 * omp-low.c (use_pointer_for_field): For nested constructs, also
191 look for map clauses on target construct.
192 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
193 taskreg_nesting_level.
196 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
197 shared clause, call omp_notice_variable on outer context if any.
199 2020-02-05 Jason Merrill <jason@redhat.com>
202 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
203 non-zero address even if weak and not yet defined.
205 2020-02-05 Martin Sebor <msebor@redhat.com>
207 PR tree-optimization/92765
208 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
209 * tree-ssa-strlen.c (compute_string_length): Remove.
210 (determine_min_objsize): Remove.
211 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
212 Avoid using type size as the upper bound on string length.
213 (handle_builtin_string_cmp): Add an argument. Adjust.
214 (strlen_check_and_optimize_call): Pass additional argument to
215 handle_builtin_string_cmp.
217 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
219 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
220 (*pushdi2_rex64 peephole2): Unconditionally split after
222 (*ashl<mode>3_doubleword): Ditto.
223 (*<shift_insn><mode>3_doubleword): Ditto.
225 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
228 * config/rs6000/rs6000.c (get_vector_offset): Fix
230 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
232 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
234 2020-02-05 David Malcolm <dmalcolm@redhat.com>
237 (Special Functions for Debugging the Analyzer): Update description
238 of __analyzer_dump_exploded_nodes.
240 2020-02-05 Jakub Jelinek <jakub@redhat.com>
243 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
244 include sets and not clobbers in the vzeroupper pattern.
245 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
246 the parallel has 17 (64-bit) or 9 (32-bit) elts.
247 (*avx_vzeroupper_1): New define_insn_and_split.
250 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
251 don't run when !optimize.
252 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
255 2020-02-05 Richard Biener <rguenther@suse.de>
258 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
259 checks before matching calls.
261 2020-02-05 Jakub Jelinek <jakub@redhat.com>
263 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
264 function comment typo.
267 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
268 simd_clone_create failed when i == 0, adjust clone->nargs by
271 2020-02-05 Martin Liska <mliska@suse.cz>
274 * doc/invoke.texi: Document that one should
275 not combine ASLR and -fpch.
277 2020-02-04 Richard Biener <rguenther@suse.de>
279 PR tree-optimization/93538
280 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
282 2020-02-04 Richard Biener <rguenther@suse.de>
284 PR tree-optimization/91123
285 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
286 (vn_walk_cb_data::last_vuse): New member.
287 (vn_walk_cb_data::saved_operands): Likewsie.
288 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
289 (vn_walk_cb_data::push_partial_def): Use finish.
290 (vn_reference_lookup_2): Update last_vuse and use finish if
291 we've saved operands.
292 (vn_reference_lookup_3): Use finish and update calls to
293 push_partial_defs everywhere. When translating through
294 memcpy or aggregate copies save off operands and alias-set.
295 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
296 operation for redundant store removal.
298 2020-02-04 Richard Biener <rguenther@suse.de>
300 PR tree-optimization/92819
301 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
302 generating more stmts than before.
304 2020-02-04 Martin Liska <mliska@suse.cz>
306 * config/arm/arm.c (arm_gen_far_branch): Move the function
307 outside of selftests.
309 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
311 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
312 function to adjust PC-relative vector addresses.
313 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
314 handle vectors with PC-relative addresses.
316 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
318 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
320 (hard_reg_and_mode_to_addr_mask): Delete.
321 (rs6000_adjust_vec_address): If the original vector address
322 was REG+REG or REG+OFFSET and the element is not zero, do the add
323 of the elements in the original address before adding the offset
324 for the vector element. Use address_to_insn_form to validate the
325 address using the register being loaded, rather than guessing
326 whether the address is a DS-FORM or DQ-FORM address.
328 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
330 * config/rs6000/rs6000.c (get_vector_offset): New helper function
331 to calculate the offset in memory from the start of a vector of a
332 particular element. Add code to keep the element number in
333 bounds if the element number is variable.
334 (rs6000_adjust_vec_address): Move calculation of offset of the
335 vector element to get_vector_offset.
336 (rs6000_split_vec_extract_var): Do not do the initial AND of
337 element here, move the code to get_vector_offset.
339 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
341 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
344 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
346 * config/rs6000/constraints.md: Improve documentation.
348 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
351 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
352 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
354 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
356 * config.gcc: Remove "carrizo" support.
357 * config/gcn/gcn-opts.h (processor_type): Likewise.
358 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
359 * config/gcn/gcn.opt (gpu_type): Likewise.
360 * config/gcn/t-omp-device: Likewise.
362 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
365 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
366 * config/arm/arm.c (arm_gen_far_branch): New function
368 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
370 2020-02-03 Julian Brown <julian@codesourcery.com>
371 Tobias Burnus <tobias@codesourcery.com>
373 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
375 2020-02-03 Jakub Jelinek <jakub@redhat.com>
378 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
379 valid RTL to sum up the lowest and second lowest bytes of the popcnt
382 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
384 PR rtl-optimization/91333
385 * ira-color.c (struct allocno_color_data): Add member
387 (init_allocno_threads): Set the member up.
388 (bucket_allocno_compare_func): Add compare hard reg
391 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
393 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
395 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
396 * config.in: Regenerated.
397 * configure: Regenerated.
398 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
399 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
400 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
402 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
404 * configure: Regenerate.
406 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
408 PR rtl-optimization/91333
409 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
410 reg preferences comparison up.
412 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
414 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
415 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
416 aarch64-sve-builtins-base.h.
417 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
418 aarch64-sve-builtins-base.cc.
419 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
420 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
422 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
423 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
424 (svcvtnt): New functions.
425 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
426 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
427 (svcvtnt): New functions.
428 (svcvt): Add a form that converts f32 to bf16.
429 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
430 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
432 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
433 Treat B as bfloat16_t.
434 (ternary_bfloat_lane_base): New class.
435 (ternary_bfloat_def): Likewise.
436 (ternary_bfloat): New shape.
437 (ternary_bfloat_lane_def): New class.
438 (ternary_bfloat_lane): New shape.
439 (ternary_bfloat_lanex2_def): New class.
440 (ternary_bfloat_lanex2): New shape.
441 (ternary_bfloat_opt_n_def): New class.
442 (ternary_bfloat_opt_n): New shape.
443 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
444 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
445 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
446 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
447 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
448 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
449 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
450 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
451 the pattern off the narrow mode instead of the wider one.
452 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
453 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
454 (sve_fp_op): Handle them.
455 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
456 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
458 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
460 * config/aarch64/arm_sve.h: Include arm_bf16.h.
461 * config/aarch64/aarch64-modes.def (BF): Move definition before
462 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
463 (SVE_MODES): Handle BF modes.
464 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
466 (aarch64_full_sve_mode): Likewise.
467 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
469 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
470 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
471 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
472 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
474 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
476 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
477 (TYPES_all_data): Add bf16.
478 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
479 (register_tuple_type): Increase buffer size.
480 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
481 (bf16): New type suffix.
482 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
483 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
484 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
485 Change type from all_data to all_arith.
486 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
489 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
490 Matthew Malcomson <matthew.malcomson@arm.com>
491 Richard Sandiford <richard.sandiford@arm.com>
493 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
494 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
495 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
496 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
497 __ARM_FEATURE_MATMUL_FP64.
498 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
499 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
500 be disabled at the same time.
501 (f32mm): New extension.
502 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
503 (AARCH64_FL_F64MM): Bump to the next bit up.
504 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
505 (TARGET_SVE_F64MM): New macros.
506 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
507 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
508 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
509 (UNSPEC_ZIP2Q): New unspeccs.
510 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
511 (optab, sur, perm_insn): Handle the new unspecs.
512 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
513 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
514 TARGET_SVE_F64MM instead of separate tests.
515 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
516 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
517 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
518 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
519 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
520 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
521 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
522 (TYPES_s_signed): New macro.
523 (TYPES_s_integer): Use it.
524 (TYPES_d_float): New macro.
525 (TYPES_d_data): Use it.
526 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
527 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
528 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
529 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
531 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
533 (ternary_resize2_lane_base): Likewise.
534 (ternary_resize2_base): New class.
535 (ternary_qq_lane_base): Likewise.
536 (ternary_intq_uintq_lane_def): Likewise.
537 (ternary_intq_uintq_lane): New shape.
538 (ternary_intq_uintq_opt_n_def): New class
539 (ternary_intq_uintq_opt_n): New shape.
540 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
541 (ternary_uintq_intq_def): New class.
542 (ternary_uintq_intq): New shape.
543 (ternary_uintq_intq_lane_def): New class.
544 (ternary_uintq_intq_lane): New shape.
545 (ternary_uintq_intq_opt_n_def): New class.
546 (ternary_uintq_intq_opt_n): New shape.
547 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
548 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
549 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
550 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
552 (svdotprod_lane_impl): ...this new class.
553 (svmmla_impl, svusdot_impl): New classes.
554 (svdot_lane): Update to use svdotprod_lane_impl.
555 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
556 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
558 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
559 function, with no types defined.
560 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
561 AARCH64_FL_I8MM functions.
562 (svmmla): New AARCH64_FL_F32MM function.
563 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
564 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
565 AARCH64_FL_F64MM function.
566 (REQUIRED_EXTENSIONS):
568 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
570 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
573 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
575 * config/i386/i386.md (*movoi_internal_avx): Do not check for
576 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
577 (*movti_internal): Do not check for
578 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
579 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
580 just after check for TARGET_AVX.
581 (*movdf_internal): Ditto.
582 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
583 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
584 * config/i386/sse.md (mov<mode>_internal): Only check
585 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
586 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
587 (<sse>_andnot<mode>3<mask_name>): Move check for
588 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
589 (<code><mode>3<mask_name>): Ditto.
590 (*andnot<mode>3): Ditto.
592 (*<code><mode>3): Ditto.
594 (*andnot<VI:mode>3): Remove
595 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
596 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
597 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
598 (sse4_1_blendv<ssemodesuffix>): Ditto.
599 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
600 Explain that tune applies to 128bit instructions only.
602 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
604 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
605 to definition of hsa_kernel_description. Parse assembly to find SGPR
606 and VGPR count of kernel and store in hsa_kernel_description.
608 2020-01-31 Tamar Christina <tamar.christina@arm.com>
610 PR rtl-optimization/91838
611 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
612 to truncate if allowed or reject combination.
614 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
616 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
617 (find_inv_vars_cb): Likewise.
619 2020-01-31 David Malcolm <dmalcolm@redhat.com>
621 * calls.c (special_function_p): Split out the check for DECL_NAME
622 being non-NULL and fndecl being extern at file scope into a
623 new maybe_special_function_p and call it. Drop check for fndecl
624 being non-NULL that was after a usage of DECL_NAME (fndecl).
625 * tree.h (maybe_special_function_p): New inline function.
627 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
629 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
630 (mask_gather_load<mode>): ... here, and zero-initialize the
632 (maskload<mode>di): Zero-initialize the destination.
635 2020-01-30 David Malcolm <dmalcolm@redhat.com>
638 * doc/analyzer.texi (Limitations): Note that constraints on
639 floating-point values are currently ignored.
641 2020-01-30 Jakub Jelinek <jakub@redhat.com>
644 * symtab.c (symtab_node::noninterposable_alias): If localalias
645 already exists, but is not usable, append numbers after it until
646 a unique name is found. Formatting fix.
649 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
652 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
654 * config/gcn/gcn.c (print_operand): Handle LTGT.
655 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
657 2020-01-30 Richard Biener <rguenther@suse.de>
659 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
660 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
662 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
664 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
665 without a DECL in .data.rel.ro.local.
667 2020-01-30 Jakub Jelinek <jakub@redhat.com>
670 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
675 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
676 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
677 any_extend code iterator instead of always zero_extend.
678 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
679 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
680 Use any_extend code iterator instead of always zero_extend.
681 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
682 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
683 Use any_extend code iterator instead of always zero_extend.
684 (*sse2_pmovmskb_ext): New define_insn.
685 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
688 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
689 (*popcountsi2_zext_falsedep): New define_insn.
691 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
693 * config.in: Regenerated.
694 * configure: Regenerated.
696 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
699 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
700 LLVM's assembler changed the default in version 9.
702 2020-01-24 Jeff Law <law@redhat.com>
704 PR tree-optimization/89689
705 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
707 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
711 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
713 PR rtl-optimization/87763
714 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
715 simplification to handle subregs as well as bare regs.
716 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
718 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
721 * ira.c (ira): Revert use of simplified LRA algorithm.
723 2020-01-29 Martin Jambor <mjambor@suse.cz>
725 PR tree-optimization/92706
726 * tree-sra.c (struct access): Fields first_link, last_link,
727 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
728 next_rhs_queued and grp_rhs_queued respectively, new fields
729 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
730 (struct assign_link): Field next renamed to next_rhs, new field
731 next_lhs. Updated comment.
732 (work_queue_head): Renamed to rhs_work_queue_head.
733 (lhs_work_queue_head): New variable.
734 (add_link_to_lhs): New function.
735 (relink_to_new_repr): Also relink LHS lists.
736 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
737 (add_access_to_lhs_work_queue): New function.
738 (pop_access_from_work_queue): Renamed to
739 pop_access_from_rhs_work_queue.
740 (pop_access_from_lhs_work_queue): New function.
741 (build_accesses_from_assign): Also add links to LHS lists and to LHS
743 (child_would_conflict_in_lacc): Renamed to
744 child_would_conflict_in_acc. Adjusted parameter names.
745 (create_artificial_child_access): New parameter set_grp_read, use it.
746 (subtree_mark_written_and_enqueue): Renamed to
747 subtree_mark_written_and_rhs_enqueue.
748 (propagate_subaccesses_across_link): Renamed to
749 propagate_subaccesses_from_rhs.
750 (propagate_subaccesses_from_lhs): New function.
751 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
754 2020-01-29 Martin Jambor <mjambor@suse.cz>
756 PR tree-optimization/92706
757 * tree-sra.c (struct access): Adjust comment of
758 grp_total_scalarization.
759 (find_access_in_subtree): Look for single children spanning an entire
761 (scalarizable_type_p): Allow register accesses, adjust callers.
762 (completely_scalarize): Remove function.
763 (scalarize_elem): Likewise.
764 (create_total_scalarization_access): Likewise.
765 (sort_and_splice_var_accesses): Do not track total scalarization
767 (analyze_access_subtree): New parameter totally, adjust to new meaning
768 of grp_total_scalarization.
769 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
770 (can_totally_scalarize_forest_p): New function.
771 (create_total_scalarization_access): Likewise.
772 (create_total_access_and_reshape): Likewise.
773 (total_should_skip_creating_access): Likewise.
774 (totally_scalarize_subtree): Likewise.
775 (analyze_all_variable_accesses): Perform total scalarization after
776 subaccess propagation using the new functions above.
777 (initialize_constant_pool_replacements): Output initializers by
778 traversing the access tree.
780 2020-01-29 Martin Jambor <mjambor@suse.cz>
782 * tree-sra.c (verify_sra_access_forest): New function.
783 (verify_all_sra_access_forests): Likewise.
784 (create_artificial_child_access): Set parent.
785 (analyze_all_variable_accesses): Call the verifier.
787 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
789 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
790 if called on indirect edge.
791 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
792 speculative call if needed.
794 2020-01-29 Richard Biener <rguenther@suse.de>
796 PR tree-optimization/93428
797 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
798 permutation when the load node is created.
799 (vect_analyze_slp_instance): Re-use it here.
801 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
803 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
805 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
807 PR rtl-optimization/93272
808 * ira-lives.c (process_out_of_region_eh_regs): New function.
809 (process_bb_node_lives): Call it.
811 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
813 * coverage.c (read_counts_file): Make error message lowercase.
815 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
817 * profile-count.c (profile_quality_display_names): Fix ordering.
819 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
822 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
823 hash only when edge is first within the sequence.
824 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
825 (symbol_table::create_edge): Do not set target_prob.
826 (cgraph_edge::remove_caller): Watch for speculative calls when updating
828 (cgraph_edge::make_speculative): Drop target_prob parameter.
829 (cgraph_edge::speculative_call_info): Remove.
830 (cgraph_edge::first_speculative_call_target): New member function.
831 (update_call_stmt_hash_for_removing_direct_edge): New function.
832 (cgraph_edge::resolve_speculation): Rewrite to new API.
833 (cgraph_edge::speculative_call_for_target): New member function.
834 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
835 multiple speculation targets.
836 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
838 (verify_speculative_call): Verify that targets form an interval.
839 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
840 (cgraph_edge::first_speculative_call_target): New member function.
841 (cgraph_edge::next_speculative_call_target): New member function.
842 (cgraph_edge::speculative_call_target_ref): New member function.
843 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
844 (cgraph_edge): Remove target_prob.
845 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
846 Fix handling of speculative calls.
847 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
848 * ipa-fnsummary.c (analyze_function_body): Likewise.
849 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
850 * ipa-profile.c (dump_histogram): Fix formating.
851 (ipa_profile_generate_summary): Watch for overflows.
852 (ipa_profile): Do not require probablity to be 1/2; update to new API.
853 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
854 (update_indirect_edges_after_inlining): Update to new API.
855 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
857 * profile-count.h: (profile_probability::adjusted): New.
858 * tree-inline.c (copy_bb): Update to new speculative call API; fix
860 * value-prof.c (gimple_ic_transform): Rename to ...
861 (dump_ic_profile): ... this one; update dumping.
862 (stream_in_histogram_value): Fix formating.
863 (gimple_value_profile_transformations): Update.
865 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
868 * config/i386/i386.md (*movoi_internal_avx): Remove
869 TARGET_SSE_TYPELESS_STORES check.
870 (*movti_internal): Prefer TARGET_AVX over
871 TARGET_SSE_TYPELESS_STORES.
872 (*movtf_internal): Likewise.
873 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
874 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
875 from TARGET_SSE_TYPELESS_STORES.
877 2020-01-28 David Malcolm <dmalcolm@redhat.com>
879 * diagnostic-core.h (warning_at): Rename overload to...
880 (warning_meta): ...this.
881 (emit_diagnostic_valist): Delete decl of overload taking
883 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
884 (warning_at): Rename overload taking diagnostic_metadata to...
885 (warning_meta): ...this.
887 2020-01-28 Richard Biener <rguenther@suse.de>
889 PR tree-optimization/93439
890 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
891 * tree-cfg.c (move_sese_region_to_fn): ... here.
892 (verify_types_in_gimple_reference): Verify used cliques are
895 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
898 * config/i386/i386-options.c (set_ix86_tune_features): Add an
899 argument of a pointer to struct gcc_options and pass it to
900 parse_mtune_ctrl_str.
901 (ix86_function_specific_restore): Pass opts to
902 set_ix86_tune_features.
903 (ix86_option_override_internal): Likewise.
904 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
905 gcc_options and use it for x_ix86_tune_ctrl_string.
907 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
909 PR rtl-optimization/87763
910 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
911 simplification to handle subregs as well as bare regs.
912 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
914 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
916 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
917 for reduction chains that (now) include a call.
919 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
921 PR tree-optimization/92822
922 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
923 out the don't-care elements of a vector whose significant elements
924 are duplicates, make the don't-care elements duplicates too.
926 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
928 PR tree-optimization/93434
929 * tree-predcom.c (split_data_refs_to_components): Record which
930 components have had aliasing loads removed. Prevent store-store
931 commoning for all such components.
933 2020-01-28 Jakub Jelinek <jakub@redhat.com>
936 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
937 -1 or is_vshift is true, use new_vector with number of elts npatterns
938 rather than new_unary_operation.
940 PR tree-optimization/93454
941 * gimple-fold.c (fold_array_ctor_reference): Perform
942 elt_size.to_uhwi () just once, instead of calling it in every
943 iteration. Punt if that value is above size of the temporary
944 buffer. Decrease third native_encode_expr argument when
945 bufoff + elt_sz is above size of buf.
947 2020-01-27 Joseph Myers <joseph@codesourcery.com>
949 * config/mips/mips.c (mips_declare_object_name)
950 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
952 2020-01-27 Martin Liska <mliska@suse.cz>
954 PR gcov-profile/93403
955 * tree-profile.c (gimple_init_gcov_profiler): Generate
956 both __gcov_indirect_call_profiler_v4 and
957 __gcov_indirect_call_profiler_v4_atomic.
959 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
962 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
964 (@aarch64_split_simd_mov<mode>): Use it.
965 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
966 Leave the vec_extract patterns to handle 2-element vectors.
967 (aarch64_simd_mov_from_<mode>high): Likewise.
968 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
969 (vec_extractv2dfv1df): Likewise.
971 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
973 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
974 jump conditions for *compare_condjump<GPI:mode>.
976 2020-01-27 David Malcolm <dmalcolm@redhat.com>
979 * digraph.cc (test_edge::test_edge): Specify template for base
982 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
984 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
986 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
988 * config/arc/arc-protos.h (gen_mlo): Remove.
990 * config/arc/arc.c (AUX_MULHI): Define.
991 (arc_must_save_reister): Special handling for r58/59.
992 (arc_compute_frame_size): Consider mlo/mhi registers.
993 (arc_save_callee_saves): Emit fp/sp move only when emit_move
995 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
996 mlo/mhi name selection.
997 (arc_restore_callee_saves): Don't early restore blink when ISR.
998 (arc_expand_prologue): Add mlo/mhi saving.
999 (arc_expand_epilogue): Add mlo/mhi restoring.
1002 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1003 numbering when MUL64 option is used.
1004 (DWARF2_FRAME_REG_OUT): Define.
1005 * config/arc/arc.md (arc600_stall): New pattern.
1006 (VUNSPEC_ARC_ARC600_STALL): Define.
1007 (mulsi64): Use correct mlo/mhi registers.
1008 (mulsi_600): Clean it up.
1009 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1011 (mhi_operand): Likewise.
1013 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1014 Petro Karashchenko <petro.karashchenko@ring.com>
1016 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1017 attributes if needed.
1018 (prepare_move_operands): Generate special unspec instruction for
1020 (arc_isuncached_mem_p): Propagate uncached attribute to each
1022 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1023 (VUNSPEC_ARC_STDI): Likewise.
1024 (ALLI): New mode iterator.
1025 (mALLI): New mode attribute.
1026 (lddi): New instruction pattern.
1028 (stdidi_split): Split instruction for architectures which are not
1029 supporting ll64 option.
1030 (lddidi_split): Likewise.
1032 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1034 PR rtl-optimization/92989
1035 * lra-lives.c (process_bb_lives): Update the live-in set before
1036 processing additional clobbers.
1038 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1040 PR rtl-optimization/93170
1041 * cselib.c (cselib_invalidate_regno_val): New function, split out
1043 (cselib_invalidate_regno): ...here.
1044 (cselib_invalidated_by_call_p): New function.
1045 (cselib_process_insn): Iterate over all the hard-register entries in
1046 REG_VALUES and invalidate any that cross call-clobbered registers.
1048 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1050 * dojump.c (split_comparison): Use HONOR_NANS rather than
1051 HONOR_SNANS when splitting LTGT.
1053 2020-01-27 Martin Liska <mliska@suse.cz>
1056 * opts.c (print_filtered_help): Exclude language-specific
1057 options from --help=common unless enabled in all FEs.
1059 2020-01-27 Martin Liska <mliska@suse.cz>
1061 * opts.c (print_help): Exclude params from
1062 all except --help=param.
1064 2020-01-27 Martin Liska <mliska@suse.cz>
1067 * config/i386/i386-features.c (make_resolver_func):
1068 Align the code with ppc64 target implementation.
1069 Do not generate a unique name for resolver function.
1071 2020-01-27 Richard Biener <rguenther@suse.de>
1073 PR tree-optimization/93397
1074 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1075 converted reduction chain SLP graph adjustment.
1077 2020-01-26 Marek Polacek <polacek@redhat.com>
1080 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1083 2020-01-26 Jason Merrill <jason@redhat.com>
1086 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1089 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1091 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1092 (rx_setmem): Likewise.
1094 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1097 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1098 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1099 drop <di> from constraint of last operand.
1102 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1103 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1104 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1106 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1109 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1112 2020-01-24 Jeff Law <law@redhat.com>
1114 PR tree-optimization/92788
1115 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1118 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1121 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1122 *avx_vperm_broadcast_<mode>,
1123 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1124 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1125 Move before avx2_perm<mode>/avx512f_perm<mode>.
1128 * simplify-rtx.c (simplify_const_unary_operation,
1129 simplify_const_binary_operation): Punt for mode precision above
1130 MAX_BITSIZE_MODE_ANY_INT.
1132 2020-01-24 Andrew Pinski <apinski@marvell.com>
1134 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1137 2020-01-24 Jeff Law <law@redhat.com>
1140 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1141 for REGs. Call output_operand_lossage to get more reasonable
1144 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1146 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1147 gcn_fp_compare_operator.
1148 (vec_cmpu<mode>di): Use gcn_compare_operator.
1149 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1150 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1151 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1152 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1153 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1154 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1155 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1156 gcn_fp_compare_operator.
1157 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1158 gcn_fp_compare_operator.
1159 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1160 gcn_fp_compare_operator.
1161 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1162 gcn_fp_compare_operator.
1164 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1166 * doc/install.texi (Cross-Compiler-Specific Options): Document
1167 `--with-toolexeclibdir' option.
1169 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1171 * target.def (flags_regnum): Also mention effect on delay slot filling.
1172 * doc/tm.texi: Regenerate.
1174 2020-01-23 Jeff Law <law@redhat.com>
1176 PR translation/90162
1177 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1179 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1182 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1185 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1187 PR rtl-optimization/93402
1188 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1191 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1193 * config.in: Regenerated.
1194 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1195 for TARGET_LIBC_GNUSTACK.
1196 * configure: Regenerated.
1197 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1198 found to be 2.31 or greater.
1200 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1202 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1204 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1205 (mips_asm_file_end): New function. Delegate to
1206 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1207 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1209 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1212 * config/i386/i386-modes.def (POImode): New mode.
1213 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1214 * config/i386/i386.md (DPWI): New mode attribute.
1215 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1217 (QPWI): ... this. Use POI instead of OI for TImode.
1218 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1219 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1222 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1225 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1227 (speculation_tracker_rev): New pattern.
1228 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1229 Use speculation_tracker_rev to track the inverse condition.
1231 2020-01-23 Richard Biener <rguenther@suse.de>
1233 PR tree-optimization/93381
1234 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1235 alias-set of the def as argument and record the first one.
1236 (vn_walk_cb_data::first_set): New member.
1237 (vn_reference_lookup_3): Pass the alias-set of the current def
1238 to push_partial_def. Fix alias-set used in the aggregate copy
1240 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1241 * real.c (clear_significand_below): Fix out-of-bound access.
1243 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1246 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1247 New define_insn patterns.
1249 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1251 * doc/sourcebuild.texi (check-function-bodies): Add an
1252 optional target/xfail selector.
1254 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1256 PR rtl-optimization/93124
1257 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1258 bare USE and CLOBBER insns.
1260 2020-01-22 Andrew Pinski <apinski@marvell.com>
1262 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1264 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1267 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1268 diagnostic_manager into "ana" namespace.
1269 * selftest-run-tests.c (selftest::run_tests): Update for move of
1270 selftest::run_analyzer_selftests to
1271 ana::selftest::run_analyzer_selftests.
1273 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1275 * cfgexpand.c (union_stack_vars): Update the size.
1277 2020-01-22 Richard Biener <rguenther@suse.de>
1279 PR tree-optimization/93381
1280 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1281 throughout, handle all conversions the same.
1283 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1286 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1287 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1288 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1289 Call force_reg on high_in2 unconditionally.
1291 2020-01-22 Martin Liska <mliska@suse.cz>
1293 PR tree-optimization/92924
1294 * profile.c (compute_value_histograms): Divide
1297 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1300 * output.h (assemble_name_resolve): Declare.
1301 * varasm.c (assemble_name_resolve): New function.
1302 (assemble_name): Use it.
1303 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1305 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1307 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1308 update_web_docs_git instead of update_web_docs_svn.
1310 2020-01-21 Andrew Pinski <apinski@marvell.com>
1313 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1314 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1315 (*tlsgd_small_<mode>): Likewise.
1316 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1317 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1318 register. Convert that register back to dest using convert_mode.
1320 2020-01-21 Jim Wilson <jimw@sifive.com>
1322 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1325 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1326 Uros Bizjak <ubizjak@gmail.com>
1329 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1331 (legitimize_tls_address): Do GNU2 TLS address computation in
1332 ptr_mode and zero-extend result to Pmode.
1333 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1334 :P with :PTR and Pmode with ptr_mode.
1335 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1336 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1337 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1339 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1342 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1343 the last two operands are CONST_INT_P before using them as such.
1345 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1347 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1348 to get the integer element types.
1350 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1352 * config/aarch64/aarch64-sve-builtins.h
1353 (function_expander::convert_to_pmode): Declare.
1354 * config/aarch64/aarch64-sve-builtins.cc
1355 (function_expander::convert_to_pmode): New function.
1356 (function_expander::get_contiguous_base): Use it.
1357 (function_expander::prepare_gather_address_operands): Likewise.
1358 * config/aarch64/aarch64-sve-builtins-sve2.cc
1359 (svwhilerw_svwhilewr_impl::expand): Likewise.
1361 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1364 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1365 cfun->machine->label_is_assembled.
1366 (aarch64_print_patchable_function_entry): New.
1367 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1368 * config/aarch64/aarch64.h (struct machine_function): New field,
1371 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1374 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1377 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1380 * cgraph.c (cgraph_edge::resolve_speculation,
1381 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1382 call_stmt_site_hash.
1384 2020-01-21 Martin Liska <mliska@suse.cz>
1386 * config/rs6000/rs6000.c (common_mode_defined): Remove
1389 2020-01-21 Richard Biener <rguenther@suse.de>
1391 PR tree-optimization/92328
1392 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1393 type when value-numbering same-sized store by inserting a
1395 (eliminate_dom_walker::eliminate_stmt): When eliminating
1396 a redundant store handle bit-reinterpretation of the same value.
1398 2020-01-21 Andrew Pinski <apinski@marvel.com>
1401 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1403 (prepare_block_for_update): This. Use a worklist instead of
1406 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1408 * gcc/config/arm/arm.c (clear_operation_p):
1409 Initialise last_regno, skip first iteration
1410 based on the first_set value and use ints instead
1411 of the unnecessary HOST_WIDE_INTs.
1413 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1416 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1417 compare_mode other than SFmode or DFmode.
1419 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1422 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1423 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1424 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1426 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1428 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1430 2020-01-20 Andrew Pinski <apinski@marvell.com>
1433 * targhooks.c (default_print_patchable_function_entry): Use
1434 output_asm_insn to emit the nop instruction.
1436 2020-01-20 Fangrui Song <maskray@google.com>
1439 * targhooks.c (default_print_patchable_function_entry): Align to
1442 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1445 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1446 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1447 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1448 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1449 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1450 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1451 Remove the {q} suffix from lea.
1452 (*tls_dynamic_gnu2_call_64): Renamed to ...
1453 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1454 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1455 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1456 Pass Pmode to gen_tls_dynamic_gnu2_64.
1458 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1460 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1462 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1464 * config/aarch64/aarch64-sve-builtins-base.cc
1465 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1467 2020-01-20 Richard Biener <rguenther@suse.de>
1470 * dwarf2out.c (prune_unused_types): Unconditionally mark
1471 called function DIEs.
1473 2020-01-20 Martin Liska <mliska@suse.cz>
1475 PR tree-optimization/93199
1476 * tree-eh.c (struct leh_state): Add
1477 new field outer_non_cleanup.
1478 (cleanup_is_dead_in): Pass leh_state instead
1479 of eh_region. Add a checking that state->outer_non_cleanup
1480 points to outer non-clean up region.
1481 (lower_try_finally): Record outer_non_cleanup
1483 (lower_catch): Likewise.
1484 (lower_eh_filter): Likewise.
1485 (lower_eh_must_not_throw): Likewise.
1486 (lower_cleanup): Likewise.
1488 2020-01-20 Richard Biener <rguenther@suse.de>
1490 PR tree-optimization/93094
1491 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1492 (vect_transform_loop): Likewise.
1493 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1494 loop_vectorized_call to vect_transform_loop.
1495 * tree-vect-loop.c (vect_transform_loop): Pass down
1496 loop_vectorized_call to vect_loop_versioning.
1497 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1498 the earlier discovered loop_vectorized_call.
1500 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1502 * doc/contribute.texi: Update for SVN -> Git transition.
1503 * doc/install.texi: Likewise.
1505 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1508 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1509 speculative targets.
1510 (verify_speculative_call): New function
1511 (cgraph_node::verify_node): Use it.
1512 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1515 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1518 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1519 (cgraph_edge::make_direct): Remove all indirect targets.
1520 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1521 (cgraph_node::verify_node): Verify that only one call_stmt or
1522 lto_stmt_uid is set.
1523 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1525 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1526 (lto_output_ref): Simplify streaming of stmt.
1527 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1529 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1531 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1532 Mark parameter unused.
1534 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1536 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1538 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1540 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1542 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1544 * Makefile.in: Add coroutine-passes.o.
1545 * builtin-types.def (BT_CONST_SIZE): New.
1546 (BT_FN_BOOL_PTR): New.
1547 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1548 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1549 * coroutine-builtins.def: New file.
1550 * coroutine-passes.cc: New file.
1551 * function.h (struct GTY function): Add a bit to indicate that the
1552 function is a coroutine component.
1553 * internal-fn.c (expand_CO_FRAME): New.
1554 (expand_CO_YIELD): New.
1555 (expand_CO_SUSPN): New.
1556 (expand_CO_ACTOR): New.
1557 * internal-fn.def (CO_ACTOR): New.
1561 * passes.def: Add pass_coroutine_lower_builtins,
1562 pass_coroutine_early_expand_ifns.
1563 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1564 (make_pass_coroutine_early_expand_ifns): New.
1565 * doc/invoke.texi: Document the fcoroutines command line
1568 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1570 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1573 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1574 after checking the argument is a REG. Don't use REGNO (reg)
1575 again to set last_regno, reuse regno variable instead.
1577 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1579 * doc/analyzer.texi (Limitations): Add note about NaN.
1581 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1582 Sudakshina Das <sudi.das@arm.com>
1584 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1585 and valid immediate.
1586 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1587 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1588 * config/arm/constraints.md (Pg): New.
1589 * config/arm/predicates.md (long_shift_imm): New.
1590 (arm_reg_or_long_shift_imm): Likewise.
1591 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1592 (thumb2_lsll): Likewise.
1595 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1596 Sudakshina Das <sudi.das@arm.com>
1598 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1599 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1600 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1601 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1602 * config/arm/thumb2.md (thumb2_asrl): New.
1603 (thumb2_lsll): Likewise.
1605 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1607 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1610 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1612 * gdbinit.in (help-gcc-hooks): New command.
1613 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1614 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1617 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1619 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1620 correct target macro.
1622 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1624 * config/aarch64/aarch64-protos.h
1625 (aarch64_sve_ld1ro_operand_p): New.
1626 * config/aarch64/aarch64-sve-builtins-base.cc
1627 (class load_replicate): New.
1628 (class svld1ro_impl): New.
1629 (class svld1rq_impl): Change to inherit from load_replicate.
1630 (svld1ro): New sve intrinsic function base.
1631 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1632 New DEF_SVE_FUNCTION.
1633 * config/aarch64/aarch64-sve-builtins-base.h
1634 (svld1ro): New decl.
1635 * config/aarch64/aarch64-sve-builtins.cc
1636 (function_expander::add_mem_operand): Modify assert to allow
1638 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1640 * config/aarch64/aarch64.c
1641 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1642 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1643 (aarch64_sve_ld1ro_operand_p): New.
1644 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1645 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1646 * config/aarch64/predicates.md
1647 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1649 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1651 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1652 Introduce this ACLE specified predefined macro.
1653 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1654 (fp): Disabling this disables f64mm.
1655 (simd): Disabling this disables f64mm.
1656 (fp16): Disabling this disables f64mm.
1657 (sve): Disabling this disables f64mm.
1658 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1659 (AARCH64_ISA_F64MM): New.
1660 (TARGET_F64MM): New.
1661 * doc/invoke.texi (f64mm): Document new option.
1663 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1665 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1666 (neoversen1_tunings): Likewise.
1668 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1671 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1672 Add assert to ensure prolog has been emitted.
1673 (aarch64_split_atomic_op): Likewise.
1674 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1675 Use epilogue_completed rather than reload_completed.
1676 (aarch64_atomic_exchange<mode>): Likewise.
1677 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1678 (atomic_nand<mode>): Likewise.
1679 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1680 (atomic_fetch_nand<mode>): Likewise.
1681 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1682 (atomic_nand_fetch<mode>): Likewise.
1684 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1687 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1689 (REVERSE_CONDITION): Delete.
1690 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1691 (CCFP_CCFPE): Likewise.
1692 (e): New mode attribute.
1693 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1694 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1695 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1696 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1697 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1698 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1699 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1700 name of generator from gen_ccmpdi to gen_ccmpccdi.
1701 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1702 the previous comparison but aren't able to, use the new ccmp_rev
1705 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1707 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1708 than testing directly for INTEGER_CST.
1709 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1711 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1713 PR tree-optimization/93292
1714 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1715 get_vectype_for_scalar_type returns NULL.
1717 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1719 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1720 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1722 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1724 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1726 * params.opt: (max-predicted-iterations): Set bounds.
1727 * predict.c (real_almost_one, real_br_prob_base,
1728 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1729 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1730 probabilities; do not truncate to reg_br_prob_bases.
1731 (estimate_loops_at_level): Pass max_cyclic_prob.
1732 (estimate_loops): Compute max_cyclic_prob.
1733 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1735 * profile-count.c (profile_probability::to_sreal): New.
1736 * profile-count.h (class sreal): Move up in file.
1737 (profile_probability::to_sreal): Declare.
1739 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1742 (arm_invalid_conversion): New function for target hook.
1743 (arm_invalid_unary_op): New function for target hook.
1744 (arm_invalid_binary_op): New function for target hook.
1746 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1748 * config.gcc: Add arm_bf16.h.
1749 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1750 (arm_simd_builtin_std_type): Add BFmode.
1751 (arm_init_simd_builtin_types): Define element types for vector types.
1752 (arm_init_bf16_types): New function.
1753 (arm_init_builtins): Add arm_init_bf16_types function call.
1754 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1755 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1756 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1757 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1758 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1759 (arm_mangle_type): Add __bf16.
1760 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1761 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1762 arm_bf16_ptr_type_node.
1763 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1764 define_split between ARM registers.
1765 * config/arm/arm_bf16.h: New file.
1766 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1767 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1769 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1770 * config/arm/vfp.md: Add BFmode to movhf patterns.
1772 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1773 Andre Vieira <andre.simoesdiasvieira@arm.com>
1775 * config/arm/arm-cpus.in (mve, mve_float): New features.
1776 (dsp, mve, mve.fp): New options.
1777 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1778 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1779 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1781 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1782 Thomas Preud'homme <thomas.preudhomme@arm.com>
1784 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1786 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1787 error for using -mcmse when targeting Armv8.1-M Mainline.
1789 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1790 Thomas Preud'homme <thomas.preudhomme@arm.com>
1792 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1793 address in r4 when targeting Armv8.1-M Mainline.
1794 (nonsecure_call_value_internal): Likewise.
1795 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1796 a register match_operand again. Emit BLXNS when targeting
1798 (nonsecure_call_value_reg_thumb2): Likewise.
1800 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1801 Thomas Preud'homme <thomas.preudhomme@arm.com>
1803 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1804 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1805 variable as true when floating-point ABI is not hard. Replace
1806 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1807 Generate VLSTM and VLLDM instruction respectively before and
1808 after a function call to cmse_nonsecure_call function.
1809 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1810 (VUNSPEC_VLLDM): Likewise.
1811 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1812 (lazy_load_multiple_insn): Likewise.
1814 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1815 Thomas Preud'homme <thomas.preudhomme@arm.com>
1817 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1818 (arm_emit_vfp_multi_reg_pop): Likewise.
1819 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1820 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1821 restore callee-saved VFP registers.
1823 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1824 Thomas Preud'homme <thomas.preudhomme@arm.com>
1826 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1827 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1828 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1829 callee-saved GPRs as well as clear ip register before doing a nonsecure
1830 call then restore callee-saved GPRs after it when targeting
1832 (arm_reorg): Adapt to function rename.
1834 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1835 Thomas Preud'homme <thomas.preudhomme@arm.com>
1837 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1838 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1839 clear_vfp_multiple pattern based on a new vfp parameter.
1840 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1841 targeting Armv8.1-M Mainline.
1842 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1843 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1844 whether VFP registers are available before looking call_used_regs for a
1846 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1847 of prototype of clear_operation_p.
1848 (clear_vfp_multiple_operation): New predicate.
1849 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1850 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1852 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1853 Thomas Preud'homme <thomas.preudhomme@arm.com>
1855 * config/arm/arm-protos.h (clear_operation_p): Declare.
1856 * config/arm/arm.c (clear_operation_p): New function.
1857 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1858 targeting Armv8.1-M Mainline or successor.
1859 (output_return_instruction): Only output APSR register clearing if
1860 Armv8.1-M Mainline instructions not available.
1861 (thumb_exit): Likewise.
1862 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1863 * config/arm/thumb2.md (clear_apsr): New define_insn.
1864 (clear_multiple): Likewise.
1865 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1867 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1868 Thomas Preud'homme <thomas.preudhomme@arm.com>
1870 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1871 (use_return_insn): Also return false for Armv8.1-M Mainline.
1872 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1873 Mainline instructions are available.
1874 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1875 when targeting Armv8.1-M Mainline Security Extensions.
1876 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1877 Mainline entry function.
1878 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1879 targeting Armv8.1-M Mainline or successor.
1880 (arm_expand_epilogue): Fix indentation of caller-saved register
1881 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1883 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1884 (FP_SYSREGS): Likewise.
1885 (enum vfp_sysregs_encoding): Define enum.
1886 (fp_sysreg_names): Declare.
1887 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1888 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1889 (pop_fpsysreg_insn): Likewise.
1891 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1892 Thomas Preud'homme <thomas.preudhomme@arm.com>
1894 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1895 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1896 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1897 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1898 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1899 (ARMv8_1m_main): New feature group.
1900 (armv8.1-m.main): New architecture.
1901 * config/arm/arm-tables.opt: Regenerate.
1902 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1903 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1904 (arm_options_perform_arch_sanity_checks): Error out when targeting
1905 Armv8.1-M Mainline Security Extensions.
1906 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1908 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1910 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1911 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1912 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1913 aarch64_bfdot_laneq): New.
1914 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1915 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1916 vbfdotq_laneq_f32): New.
1917 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1918 VBFMLA_W, VBF): New.
1919 (isquadop): Add V4BF, V8BF.
1921 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1923 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1924 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1925 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1926 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1927 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1928 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1929 usdot_laneq, sudot_lane,sudot_laneq): New.
1930 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1931 (aarch64_<sur>dot_lane): New.
1932 * config/aarch64/arm_neon.h (vusdot_s32): New.
1934 (vusdot_lane_s32): New.
1935 (vsudot_lane_s32): New.
1936 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1937 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1939 2020-01-16 Martin Liska <mliska@suse.cz>
1941 * value-prof.c (dump_histogram_value): Fix
1942 obvious spacing issue.
1944 2020-01-16 Andrew Pinski <apinski@marvell.com>
1946 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1947 !storage_order_barrier_p.
1949 2020-01-16 Andrew Pinski <apinski@marvell.com>
1951 * sched-int.h (_dep): Add unused bit-field field for the padding.
1952 * sched-deps.c (init_dep_1): Init unused field.
1954 2020-01-16 Andrew Pinski <apinski@marvell.com>
1956 * optabs.h (create_expand_operand): Initialize target field also.
1958 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
1960 PR tree-optimization/92429
1961 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1962 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1964 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1967 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
1969 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1970 aarch64_sve_int_mode to each mode.
1972 2020-01-15 David Malcolm <dmalcolm@redhat.com>
1974 * doc/analyzer.texi (Overview): Add note about
1975 -fdump-ipa-analyzer.
1977 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
1979 PR tree-optimization/93231
1980 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
1981 input_type is unsigned. Use tree_to_shwi for shift constant.
1982 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
1983 (simplify_count_trailing_zeroes): Add test to handle known non-zero
1984 inputs more efficiently.
1986 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
1988 * config/i386/i386.md (*movsf_internal): Do not require
1989 SSE2 ISA for alternatives 14 and 15.
1991 2020-01-15 Richard Biener <rguenther@suse.de>
1994 * tree-eh.c (sink_clobbers): If we already visited the destination
1995 block do not defer insertion.
1996 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
1997 the purpose of defered insertion.
1999 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2001 * BASE-VER: Bump to 10.0.1.
2003 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2005 PR tree-optimization/93247
2006 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2007 type of the stmt that we're going to vectorize.
2009 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2011 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2012 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2015 2020-01-15 Martin Liska <mliska@suse.cz>
2017 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2018 2 calls of streamer_read_hwi in a function call.
2020 2020-01-15 Richard Biener <rguenther@suse.de>
2022 * alias.c (record_alias_subset): Avoid redundant work when
2023 subset is already recorded.
2025 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2027 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2028 the analyzer options provide CWE identifiers.
2030 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2032 * tree-diagnostic-path.cc (path_summary::event_range::print):
2033 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2034 using get_pure_location.
2036 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2038 PR tree-optimization/93262
2039 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2040 perform head trimming only if the last argument is constant,
2041 either all ones, or larger or equal to head trim, in the latter
2042 case decrease the last argument by head_trim.
2044 PR tree-optimization/93249
2045 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2046 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2047 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2048 perform head trim unless we can prove there are no '\0' chars
2049 from the source among the first head_trim chars.
2051 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2053 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2055 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2058 * config/i386/sse.md
2059 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2060 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2061 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2062 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2063 just a single alternative instead of two, make operands 1 and 2
2066 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2069 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2072 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2074 * Makefile.in (lang_opt_files): Add analyzer.opt.
2075 (ANALYZER_OBJS): New.
2076 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2077 tristate.o and ANALYZER_OBJS.
2078 (TEXI_GCCINT_FILES): Add analyzer.texi.
2079 * common.opt (-fanalyzer): New driver option.
2080 * config.in: Regenerate.
2081 * configure: Regenerate.
2082 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2083 (gccdepdir): Also create depdir for "analyzer" subdir.
2084 * digraph.cc: New file.
2085 * digraph.h: New file.
2086 * doc/analyzer.texi: New file.
2087 * doc/gccint.texi ("Static Analyzer") New menu item.
2088 (analyzer.texi): Include it.
2089 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2090 ("Warning Options"): Add static analysis warnings to the list.
2091 (-Wno-analyzer-double-fclose): New option.
2092 (-Wno-analyzer-double-free): New option.
2093 (-Wno-analyzer-exposure-through-output-file): New option.
2094 (-Wno-analyzer-file-leak): New option.
2095 (-Wno-analyzer-free-of-non-heap): New option.
2096 (-Wno-analyzer-malloc-leak): New option.
2097 (-Wno-analyzer-possible-null-argument): New option.
2098 (-Wno-analyzer-possible-null-dereference): New option.
2099 (-Wno-analyzer-null-argument): New option.
2100 (-Wno-analyzer-null-dereference): New option.
2101 (-Wno-analyzer-stale-setjmp-buffer): New option.
2102 (-Wno-analyzer-tainted-array-index): New option.
2103 (-Wno-analyzer-use-after-free): New option.
2104 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2105 (-Wno-analyzer-use-of-uninitialized-value): New option.
2106 (-Wanalyzer-too-complex): New option.
2107 (-fanalyzer-call-summaries): New warning.
2108 (-fanalyzer-checker=): New warning.
2109 (-fanalyzer-fine-grained): New warning.
2110 (-fno-analyzer-state-merge): New warning.
2111 (-fno-analyzer-state-purge): New warning.
2112 (-fanalyzer-transitivity): New warning.
2113 (-fanalyzer-verbose-edges): New warning.
2114 (-fanalyzer-verbose-state-changes): New warning.
2115 (-fanalyzer-verbosity=): New warning.
2116 (-fdump-analyzer): New warning.
2117 (-fdump-analyzer-callgraph): New warning.
2118 (-fdump-analyzer-exploded-graph): New warning.
2119 (-fdump-analyzer-exploded-nodes): New warning.
2120 (-fdump-analyzer-exploded-nodes-2): New warning.
2121 (-fdump-analyzer-exploded-nodes-3): New warning.
2122 (-fdump-analyzer-supergraph): New warning.
2123 * doc/sourcebuild.texi (dg-require-dot): New.
2124 (dg-check-dot): New.
2125 * gdbinit.in (break-on-saved-diagnostic): New command.
2126 * graphviz.cc: New file.
2127 * graphviz.h: New file.
2128 * ordered-hash-map-tests.cc: New file.
2129 * ordered-hash-map.h: New file.
2130 * passes.def (pass_analyzer): Add before
2131 pass_ipa_whole_program_visibility.
2132 * selftest-run-tests.c (selftest::run_tests): Call
2133 selftest::ordered_hash_map_tests_cc_tests.
2134 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2136 * shortest-paths.h: New file.
2137 * timevar.def (TV_ANALYZER): New timevar.
2138 (TV_ANALYZER_SUPERGRAPH): Likewise.
2139 (TV_ANALYZER_STATE_PURGE): Likewise.
2140 (TV_ANALYZER_PLAN): Likewise.
2141 (TV_ANALYZER_SCC): Likewise.
2142 (TV_ANALYZER_WORKLIST): Likewise.
2143 (TV_ANALYZER_DUMP): Likewise.
2144 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2145 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2146 * tree-pass.h (make_pass_analyzer): New decl.
2147 * tristate.cc: New file.
2148 * tristate.h: New file.
2150 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2153 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2154 alternatives 9 and 10.
2156 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2158 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2159 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2160 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2161 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2162 (selftest::hash_map_tests_c_tests): Call it.
2163 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2164 New static constant, using the value of = H::empty_zero_p.
2165 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2166 from default_hash_traits <Value>.
2167 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2169 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2170 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2171 calls to mark_empty with !Descriptor::empty_zero_p.
2172 (hash_table::empty_slow): Conditionalize the memset call with a
2173 check that Descriptor::empty_zero_p; otherwise, loop through the
2174 entries calling mark_empty on them.
2175 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2176 (pointer_hash::empty_zero_p): Likewise.
2177 (pair_hash::empty_zero_p): Likewise.
2178 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2180 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2181 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2182 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2183 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2184 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2185 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2186 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2187 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2189 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2192 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2194 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2195 fix typo on return value.
2197 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2200 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2202 (cgraph_edge::make_speculative): Add param for setting speculative_id
2204 (cgraph_edge::speculative_call_info): Update comments and find reference
2205 by speculative_id for multiple indirect targets.
2206 (cgraph_edge::resolve_speculation): Decrease the speculations
2207 for indirect edge, drop it's speculative if not direct target
2208 left. Update comments.
2209 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2210 (cgraph_node::dump): Print num_speculative_call_targets.
2211 (cgraph_node::verify_node): Don't report error if speculative
2212 edge not include statement.
2213 (cgraph_edge::num_speculative_call_targets_p): New function.
2214 * cgraph.h (int common_target_id): Remove.
2215 (int common_target_probability): Remove.
2216 (num_speculative_call_targets): New variable.
2217 (make_speculative): Add param for setting speculative_id.
2218 (cgraph_edge::num_speculative_call_targets_p): New declare.
2219 (target_prob): New variable.
2220 (speculative_id): New variable.
2221 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2222 call summaries for multiple speculative call targets.
2223 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2224 * ipa-profile.c (struct speculative_call_target): New struct.
2225 (class speculative_call_summary): New class.
2226 (class speculative_call_summaries): New class.
2227 (call_sums): New variable.
2228 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2229 (ipa_profile_write_edge_summary): New function.
2230 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2231 (ipa_profile_dump_all_summaries): New function.
2232 (ipa_profile_read_edge_summary): New function.
2233 (ipa_profile_read_summary_section): New function.
2234 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2235 (ipa_profile): Generate num_speculative_call_targets from
2237 * ipa-ref.h (speculative_id): New variable.
2238 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2239 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2240 common_target_probability. Stream out speculative_id and
2241 num_speculative_call_targets.
2242 (input_edge): Likewise.
2243 * predict.c (dump_prediction): Remove edges count assert to be
2245 * symtab.c (symtab_node::create_reference): Init speculative_id.
2246 (symtab_node::clone_references): Clone speculative_id.
2247 (symtab_node::clone_referring): Clone speculative_id.
2248 (symtab_node::clone_reference): Clone speculative_id.
2249 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2250 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2251 if indirect call contains multiple speculative targets.
2252 * value-prof.h (check_ic_target): Remove.
2253 * value-prof.c (gimple_value_profile_transformations):
2254 Use void function gimple_ic_transform.
2255 * value-prof.c (gimple_ic_transform): Handle topn case.
2256 Fix comment typos. Change it to a void function.
2258 2020-01-13 Andrew Pinski <apinski@marvell.com>
2260 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2261 (octeontx2t98): New define.
2262 (octeontx2t96): New define.
2263 (octeontx2t93): New define.
2264 (octeontx2f95): New define.
2265 (octeontx2f95n): New define.
2266 (octeontx2f95mm): New define.
2267 * config/aarch64/aarch64-tune.md: Regenerate.
2268 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2270 2020-01-13 Jason Merrill <jason@redhat.com>
2272 PR c++/33799 - destroy return value if local cleanup throws.
2273 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2275 2020-01-13 Martin Liska <mliska@suse.cz>
2277 * ipa-cp.c (get_max_overall_size): Use newly
2278 renamed param param_ipa_cp_unit_growth.
2279 * params.opt: Remove legacy param name.
2281 2020-01-13 Martin Sebor <msebor@redhat.com>
2283 PR tree-optimization/93213
2284 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2285 stores to be eliminated.
2287 2020-01-13 Martin Liska <mliska@suse.cz>
2289 * opts.c (print_help): Do not print CL_PARAM
2290 and CL_WARNING for CL_OPTIMIZATION.
2292 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2295 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2296 depending on optimization settings.
2298 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2300 PR tree-optimization/90838
2301 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2302 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2303 argument rather than to initialize temporary for targets that
2304 don't use the mode argument at all. Initialize ctzval to avoid
2307 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2309 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2310 * tree-core.h: Document it.
2311 * gimplify.c (gimplify_omp_workshare): Set it.
2312 * omp-low.c (lower_omp_target): Use it.
2313 * tree-pretty-print.c (dump_omp_clause): Print it.
2315 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2316 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2318 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2320 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2321 * common.opt (fdiagnostics-path-format=): New option.
2322 (diagnostic_path_format): New enum.
2323 (fdiagnostics-show-path-depths): New option.
2324 * coretypes.h (diagnostic_event_id_t): New forward decl.
2325 * diagnostic-color.c (color_dict): Add "path".
2326 * diagnostic-event-id.h: New file.
2327 * diagnostic-format-json.cc (json_from_expanded_location): Make
2329 (json_end_diagnostic): Call context->make_json_for_path if it
2330 exists and the diagnostic has a path.
2331 (diagnostic_output_format_init): Clear context->print_path.
2332 * diagnostic-path.h: New file.
2333 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2334 when printing a run of events in a diagnostic_path so that they
2335 all get the same color.
2336 (layout::m_diagnostic_path_p): New field.
2337 (layout::layout): Initialize it.
2338 (layout::print_any_labels): Don't colorize the label text for an
2339 event in a diagnostic_path.
2340 (gcc_rich_location::add_location_if_nearby): Add
2341 "restrict_to_current_line_spans" and "label" params. Pass the
2342 former to layout.maybe_add_location_range; pass the latter
2343 when calling add_range.
2344 * diagnostic.c: Include "diagnostic-path.h".
2345 (diagnostic_initialize): Initialize context->path_format and
2346 context->show_path_depths.
2347 (diagnostic_show_any_path): New function.
2348 (diagnostic_path::interprocedural_p): New function.
2349 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2350 (simple_diagnostic_path::num_events): New function.
2351 (simple_diagnostic_path::get_event): New function.
2352 (simple_diagnostic_path::add_event): New function.
2353 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2354 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2355 (debug): New overload taking a diagnostic_path *.
2356 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2357 * diagnostic.h (enum diagnostic_path_format): New enum.
2358 (json::value): New forward decl.
2359 (diagnostic_context::path_format): New field.
2360 (diagnostic_context::show_path_depths): New field.
2361 (diagnostic_context::print_path): New callback field.
2362 (diagnostic_context::make_json_for_path): New callback field.
2363 (diagnostic_show_any_path): New decl.
2364 (json_from_expanded_location): New decl.
2365 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2366 (-fdiagnostics-show-path-depths): New option.
2367 (-fdiagnostics-color): Add "path" to description of default
2368 GCC_COLORS; describe it.
2369 (-fdiagnostics-format=json): Document how diagnostic paths are
2370 represented in the JSON output format.
2371 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2372 Add optional params "restrict_to_current_line_spans" and "label".
2373 * opts.c (common_handle_option): Handle
2374 OPT_fdiagnostics_path_format_ and
2375 OPT_fdiagnostics_show_path_depths.
2376 * pretty-print.c: Include "diagnostic-event-id.h".
2377 (pp_format): Implement "%@" format code for printing
2378 diagnostic_event_id_t *.
2379 (selftest::test_pp_format): Add tests for "%@".
2380 * selftest-run-tests.c (selftest::run_tests): Call
2381 selftest::tree_diagnostic_path_cc_tests.
2382 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2383 * toplev.c (general_init): Initialize global_dc->path_format and
2384 global_dc->show_path_depths.
2385 * tree-diagnostic-path.cc: New file.
2386 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2387 non-static. Drop "diagnostic" param in favor of storing the
2388 original value of "where" and re-using it.
2389 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2390 maybe_unwind_expanded_macro_loc.
2391 (tree_diagnostics_defaults): Initialize context->print_path and
2392 context->make_json_for_path.
2393 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2395 (default_tree_make_json_for_path): New decl.
2396 (maybe_unwind_expanded_macro_loc): New decl.
2398 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2400 PR tree-optimization/93210
2401 * fold-const.h (native_encode_initializer,
2402 can_native_interpret_type_p): Declare.
2403 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2405 (native_encode_initializer): New function, moved from dwarf2out.c.
2406 Adjust to native_encode_expr compatible arguments, including dry-run
2407 and partial extraction modes. Don't handle STRING_CST.
2408 (can_native_interpret_type_p): No longer static.
2409 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2410 offset / BITS_PER_UNIT fits into int and don't call it if
2411 can_native_interpret_type_p fails. If suboff is NULL and for
2412 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2413 native_encode_initializer.
2414 (fold_const_aggregate_ref_1): Formatting fix.
2415 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2416 (tree_add_const_value_attribute): Adjust caller.
2418 PR tree-optimization/90838
2419 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2420 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2421 CTZ_DEFINED_VALUE_AT_ZERO.
2423 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2426 * lra-constraints.c (match_reload): Permit input operands have the
2427 same mode as output while other input operands have a different
2430 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2432 PR tree-optimization/90838
2433 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2434 (check_ctz_string): Likewise.
2435 (optimize_count_trailing_zeroes): Likewise.
2436 (simplify_count_trailing_zeroes): Likewise.
2437 (pass_forwprop::execute): Try ctz simplification.
2438 * match.pd: Add matching for ctz idioms.
2440 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2442 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2444 (aarch64_invalid_unary_op): New function for target hook.
2445 (aarch64_invalid_binary_op): New function for target hook.
2447 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2449 * config.gcc: Add arm_bf16.h.
2450 * config/aarch64/aarch64-builtins.c
2451 (aarch64_simd_builtin_std_type): Add BFmode.
2452 (aarch64_init_simd_builtin_types): Define element types for vector
2454 (aarch64_init_bf16_types): New function.
2455 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2456 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2458 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2459 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2461 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2462 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2463 * config/aarch64/aarch64.c
2464 (aarch64_classify_vector_mode): Add support for BF types.
2465 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2466 (aarch64_vq_mode): Add support for BF types.
2467 (aarch64_simd_container_mode): Add support for BF types.
2468 (aarch64_mangle_type): Add support for BF scalar type.
2469 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2470 * config/aarch64/arm_bf16.h: New file.
2471 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2472 * config/aarch64/iterators.md: Add BF types to mode attributes.
2473 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2475 2020-01-10 Jason Merrill <jason@redhat.com>
2477 PR c++/93173 - incorrect tree sharing.
2478 * gimplify.c (copy_if_shared): No longer static.
2479 * gimplify.h: Declare it.
2481 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2483 * doc/invoke.texi (-msve-vector-bits=): Document that
2484 -msve-vector-bits=128 now generates VL-specific code for
2485 little-endian targets.
2486 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2487 build_vector_type_for_mode to construct the data vector types.
2488 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2489 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2490 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2491 for 128-bit vectors.
2493 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2495 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2498 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2500 * config/aarch64/aarch64-builtins.c
2501 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2502 rather than checking the number of elements and the element mode.
2504 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2506 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2507 get_related_vectype_for_scalar_type rather than build_vector_type
2508 to create the index type for a conditional reduction.
2510 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2512 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2513 for any type of gather or scatter, including strided accesses.
2515 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2517 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2520 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2522 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2524 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2525 parameter and its use to reset DR_OFFSET's.
2526 (vect_transform_loop): Remove orig_drs_init argument.
2527 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2528 member of dr_vec_info rather than the offset of the associated
2529 data_reference's innermost_loop_behavior.
2530 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2531 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2532 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2533 get_dr_vinfo_offset.
2534 (vectorizable_store): Likewise.
2535 (vectorizable_load): Likewise.
2537 2020-01-10 Richard Biener <rguenther@suse.de>
2539 * gimple-ssa-store-merging
2540 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2542 2020-01-10 Martin Liska <mliska@suse.cz>
2545 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2546 encapsulation that was there before r280040.
2548 2020-01-10 Richard Biener <rguenther@suse.de>
2551 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2552 sequences to avoid walking them again for secondary opportunities.
2553 (pass_lower_eh_dispatch::execute): Instead actually insert
2556 2020-01-10 Richard Biener <rguenther@suse.de>
2559 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2560 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2561 avoid quadraticness.
2563 2020-01-10 Martin Jambor <mjambor@suse.cz>
2565 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2566 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2567 to get param_ipa_sra_max_replacements.
2568 (param_splitting_across_edge): Pass the caller to
2569 pull_accesses_from_callee.
2571 2020-01-10 Martin Jambor <mjambor@suse.cz>
2573 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2574 * ipa-cp.c (max_new_size): Removed.
2575 (orig_overall_size): New variable.
2576 (get_max_overall_size): New function.
2577 (estimate_local_effects): Use it. Adjust dump.
2578 (decide_about_value): Likewise.
2579 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2580 orig_overall_size. Adjust dump.
2581 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2583 2020-01-10 Martin Jambor <mjambor@suse.cz>
2585 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2586 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2587 instead of param_ipa_max_agg_items.
2588 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2589 optimization info for the callee.
2591 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2593 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2594 markers if debug_inline_points is false.
2596 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2598 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2600 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2601 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2602 aarch64-sve-builtins-sve2.h.
2603 (aarch64-sve-builtins-sve2.o): New rule.
2604 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2605 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2606 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2607 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2608 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2609 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2611 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2612 instructions that are handled here.
2613 (@cond_asrd<mode>): Generalize to...
2614 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2615 (*cond_asrd<mode>_2): Generalize to...
2616 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2617 (*cond_asrd<mode>_z): Generalize to...
2618 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2619 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2620 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2621 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2622 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2624 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2625 (@aarch64_scatter_stnt<mode>): Likewise.
2626 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2627 (@aarch64_mul_lane_<mode>): Likewise.
2628 (@aarch64_sve_suqadd<mode>_const): Likewise.
2629 (*<sur>h<addsub><mode>): Generalize to...
2630 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2632 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2633 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2634 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2635 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2636 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2637 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2638 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2639 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2640 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2641 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2642 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2643 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2644 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2645 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2646 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2647 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2648 (@aarch64_sve2_xar<mode>): Likewise.
2649 (@aarch64_sve2_bcax<mode>): Likewise.
2650 (*aarch64_sve2_eor3<mode>): Rename to...
2651 (@aarch64_sve2_eor3<mode>): ...this.
2652 (@aarch64_sve2_bsl<mode>): New expander.
2653 (@aarch64_sve2_nbsl<mode>): Likewise.
2654 (@aarch64_sve2_bsl1n<mode>): Likewise.
2655 (@aarch64_sve2_bsl2n<mode>): Likewise.
2656 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2657 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2658 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2659 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2660 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2661 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2662 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2663 (<su>mull<bt><Vwide>): Generalize to...
2664 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2666 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2667 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2668 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2669 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2670 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2671 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2672 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2673 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2674 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2675 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2676 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2677 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2678 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2679 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2680 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2681 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2682 (<SHRNB:r>shrnb<mode>): Generalize to...
2683 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2685 (<SHRNT:r>shrnt<mode>): Generalize to...
2686 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2688 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2689 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2690 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2691 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2692 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2693 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2694 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2695 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2696 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2697 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2698 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2699 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2700 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2701 (@aarch64_sve2_cvtnt<mode>): Likewise.
2702 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2703 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2704 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2705 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2706 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2707 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2708 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2709 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2710 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2711 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2712 (@aarch64_sve2_pmul<mode>): Likewise.
2713 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2714 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2715 (@aarch64_sve2_tbl2<mode>): Likewise.
2716 (@aarch64_sve2_tbx<mode>): Likewise.
2717 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2718 (@aarch64_sve2_histcnt<mode>): Likewise.
2719 (@aarch64_sve2_histseg<mode>): Likewise.
2720 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2721 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2722 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2723 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2724 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2725 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2726 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2727 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2728 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2729 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2730 (SVE2_PMULL_PAIR_I): New mode iterators.
2731 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2732 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2733 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2734 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2735 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2736 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2737 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2738 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2739 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2740 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2741 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2742 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2743 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2744 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2745 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2746 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2747 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2748 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2749 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2750 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2751 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2752 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2753 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2754 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2755 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2756 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2757 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2758 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2759 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2760 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2761 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2763 (VNARROW, Ventype): New mode attributes.
2764 (Vewtype): Handle VNx2DI. Fix typo in comment.
2765 (VDOUBLE): New mode attribute.
2766 (sve_lane_con): Handle VNx8HI.
2767 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2768 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2769 (sve_int_op, sve_int_op_rev): Handle the above codes.
2770 (sve_pred_int_rhs2_operand): Likewise.
2771 (MULLBT, SHRNB, SHRNT): Delete.
2772 (SVE_INT_SHIFT_IMM): New int iterator.
2773 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2774 and UNSPEC_WHILEHS for TARGET_SVE2.
2775 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2776 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2777 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2778 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2779 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2780 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2781 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2782 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2783 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2784 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2785 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2786 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2787 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2788 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2789 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2790 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2791 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2792 (optab): Handle the new unspecs.
2793 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2795 (lr): Handle the new unspecs.
2797 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2798 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2799 (sve_int_qsub_op): New int attributes.
2800 (sve_fp_op, rot): Handle the new unspecs.
2801 * config/aarch64/aarch64-sve-builtins.h
2802 (function_resolver::require_matching_pointer_type): Declare.
2803 (function_resolver::resolve_unary): Add an optional boolean argument.
2804 (function_resolver::finish_opt_n_resolution): Add an optional
2805 type_suffix_index argument.
2806 (gimple_folder::redirect_call): Declare.
2807 (gimple_expander::prepare_gather_address_operands): Add an optional
2809 * config/aarch64/aarch64-sve-builtins.cc: Include
2810 aarch64-sve-builtins-sve2.h.
2811 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2812 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2813 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2814 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2815 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2816 (TYPES_s_unsigned): Likewise.
2817 (TYPES_s_integer): Use TYPES_s_unsigned.
2818 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2819 (TYPES_sd_integer): Use them.
2820 (TYPES_d_unsigned): New macro.
2821 (TYPES_d_integer): Use it.
2822 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2823 (TYPES_cvt_narrow): Likewise.
2824 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2825 (preds_mx): New variable.
2826 (function_builder::add_overloaded_function): Allow the new feature
2827 set to be more restrictive than the original one.
2828 (function_resolver::infer_pointer_type): Remove qualifiers from
2829 the pointer type before printing it.
2830 (function_resolver::require_matching_pointer_type): New function.
2831 (function_resolver::resolve_sv_displacement): Handle functions
2832 that don't support 32-bit vector indices or svint32_t vector offsets.
2833 (function_resolver::finish_opt_n_resolution): Take the inferred type
2834 as a separate argument.
2835 (function_resolver::resolve_unary): Optionally treat all forms in
2836 the same way as normal merging functions.
2837 (gimple_folder::redirect_call): New function.
2838 (function_expander::prepare_gather_address_operands): Add an argument
2839 that says whether scaled forms are available. If they aren't,
2840 handle scaling of vector indices and don't add the extension and
2842 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2843 fall back to using cond_* instead.
2844 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2845 Split out the member variables into...
2846 (rtx_code_function_base): ...this new base class.
2847 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2848 (unspec_based_function): Split out the member variables into...
2849 (unspec_based_function_base): ...this new base class.
2850 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2851 (unspec_based_function_exact_insn): New class.
2852 (unspec_based_add_function, unspec_based_add_lane_function)
2853 (unspec_based_lane_function, unspec_based_pred_function)
2854 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2855 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2856 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2858 (unspec_based_fused_function): New class.
2859 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2860 (unspec_based_fused_lane_function): New class.
2861 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2863 (CODE_FOR_MODE1): New macro.
2864 (fixed_insn_function): New class.
2865 (while_comparison): Likewise.
2866 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2867 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2868 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2869 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2870 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2871 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2872 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2873 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2874 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2875 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2876 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2877 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2878 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2879 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2880 Also add an initial argument for unary_convert_narrowt, regardless
2881 of the predication type.
2882 (build_32_64): Allow loads and stores to specify MODE_none.
2883 (build_sv_index64, build_sv_uint_offset): New functions.
2884 (long_type_suffix): New function.
2885 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2886 (binary_imm_long_base, load_gather_sv_base): Likewise.
2887 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2888 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2889 (unary_narrowb_base, unary_narrowt_base): Likewise.
2890 (binary_long_lane_def, binary_long_lane): New shape.
2891 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2892 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2893 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2894 (binary_to_uint_def, binary_to_uint): Likewise.
2895 (binary_wide_def, binary_wide): Likewise.
2896 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2897 (compare_def, compare): Likewise.
2898 (compare_ptr_def, compare_ptr): Likewise.
2899 (load_ext_gather_index_restricted_def,
2900 load_ext_gather_index_restricted): Likewise.
2901 (load_ext_gather_offset_restricted_def,
2902 load_ext_gather_offset_restricted): Likewise.
2903 (load_gather_sv_def): Inherit from load_gather_sv_base.
2904 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2905 (shift_left_imm_def, shift_left_imm): Likewise.
2906 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2907 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2908 (store_scatter_index_restricted_def,
2909 store_scatter_index_restricted): Likewise.
2910 (store_scatter_offset_restricted_def,
2911 store_scatter_offset_restricted): Likewise.
2912 (tbl_tuple_def, tbl_tuple): Likewise.
2913 (ternary_long_lane_def, ternary_long_lane): Likewise.
2914 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2915 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2916 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2917 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2918 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2919 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2920 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2921 (ternary_uint_def, ternary_uint): Likewise.
2922 (unary_convert): Fix typo in comment.
2923 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2924 (unary_long_def, unary_long): Likewise.
2925 (unary_narrowb_def, unary_narrowb): Likewise.
2926 (unary_narrowt_def, unary_narrowt): Likewise.
2927 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2928 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2929 (unary_to_int_def, unary_to_int): Likewise.
2930 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2931 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2932 (svasrd_impl): Delete.
2933 (svcadd_impl::expand): Handle integer operations too.
2934 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2935 new functions to derive the unspec numbers.
2936 (svmla_svmls_lane_impl): Replace with...
2937 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2938 integer operations too.
2939 (svwhile_impl): Rename to...
2940 (svwhilelx_impl): ...this and inherit from while_comparison.
2941 (svasrd): Use unspec_based_function.
2942 (svmla_lane): Use svmla_lane_impl.
2943 (svmls_lane): Use svmls_lane_impl.
2944 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2945 (svwhilele, svwhilelt): Use svwhilelx_impl.
2946 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2947 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2948 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2949 * config/aarch64/aarch64-sve-builtins.def: Include
2950 aarch64-sve-builtins-sve2.def.
2952 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2954 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2955 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2956 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2957 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
2958 immediates as well as vector ones.
2959 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2960 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2961 (aarch64_sve_qsub_immediate): Update calls accordingly.
2963 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2965 * config/aarch64/aarch64-sve2.md: Add banner comments.
2966 (<su>mulh<r>s<mode>3): Move further up file.
2967 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2968 (*aarch64_sve2_sra<mode>): Move further down file.
2969 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2971 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2973 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2975 (while_optab_cmp): Handle them.
2976 * config/aarch64/aarch64-sve.md
2977 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
2978 and add a "@" marker.
2979 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
2980 instead of gen_aarch64_sve2_while_ptest.
2981 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
2983 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2985 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
2986 (UNSPEC_WHILELE): ...this.
2987 (UNSPEC_WHILE_LO): Rename to...
2988 (UNSPEC_WHILELO): ...this.
2989 (UNSPEC_WHILE_LS): Rename to...
2990 (UNSPEC_WHILELS): ...this.
2991 (UNSPEC_WHILE_LT): Rename to...
2992 (UNSPEC_WHILELT): ...this.
2993 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
2994 (cmp_op, while_optab_cmp): Likewise.
2995 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
2996 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
2997 (svwhilelt): Likewise.
2999 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3001 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3002 (unary_to_uint): Define.
3003 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3004 (unary_count): Rename to...
3005 (unary_to_uint_def, unary_to_uint): ...this.
3006 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3008 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3010 * config/aarch64/aarch64-sve-builtins-functions.h
3011 (code_for_mode_function): New class.
3012 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3013 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3014 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3015 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3016 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3018 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3020 * config/aarch64/iterators.md (addsub): New code attribute.
3021 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3023 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3024 in the asm string and attributes. Fix indentation.
3025 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3027 (@aarch64_sve_<optab><mode>): ...this.
3028 * config/aarch64/aarch64-sve-builtins.h
3029 (function_expander::expand_signed_unpred_op): Delete.
3030 * config/aarch64/aarch64-sve-builtins.cc
3031 (function_expander::expand_signed_unpred_op): Likewise.
3032 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3033 try using code_for_aarch64_sve instead.
3034 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3035 (svqsub_impl): Likewise.
3036 (svqadd, svqsub): Use rtx_code_function instead.
3038 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3040 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3041 (HADDSUB, sur, addsub): Remove them.
3043 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3045 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3046 functions the same way, rather than singling out those that
3047 aren't mapped directly to optabs.
3049 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3051 * target.def (compatible_vector_types_p): New target hook.
3052 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3053 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3054 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3055 * doc/tm.texi: Regenerate.
3056 * gimple-expr.c: Include target.h.
3057 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3058 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3060 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3061 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3062 Use the original predicate if it already has a suitable type.
3064 2020-01-09 Martin Jambor <mjambor@suse.cz>
3066 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3067 resolve_speculation and redirect_call_stmt_to_callee static. Change
3068 return type of set_call_stmt to cgraph_edge *.
3069 * auto-profile.c (afdo_indirect_call): Adjust call to
3070 redirect_call_stmt_to_callee.
3071 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3072 make the this pointer explicit, adjust self-recursive calls and the
3073 call top make_direct. Return the resulting edge.
3074 (cgraph_edge::remove): Make this pointer explicit.
3075 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3076 (cgraph_edge::make_direct): Likewise, adjust call to
3077 resolve_speculation.
3078 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3079 call to set_call_stmt.
3080 (cgraph_update_edges_for_call_stmt_node): Update call to
3081 set_call_stmt and remove.
3082 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3083 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3084 (cgraph_node::create_edge_including_clones): Moved "first" definition
3085 of edge to the block where it was used. Adjusted calls to
3087 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3088 cgraph_edge::remove.
3089 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3090 make_direct and redirect_call_stmt_to_callee.
3091 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3092 resolve_speculation and make_direct.
3093 * ipa-inline-transform.c (inline_transform): Adjust call to
3094 redirect_call_stmt_to_callee.
3095 (check_speculations_1):: Adjust call to resolve_speculation.
3096 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3097 resolve-speculation.
3098 (inline_small_functions): Adjust call to resolve_speculation.
3099 (ipa_inline): Likewise.
3100 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3102 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3103 safe with regards to edge removal, adjust calls to
3104 redirect_call_stmt_to_callee.
3105 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3106 and redirect_call_stmt_to_callee.
3107 * multiple_target.c (create_dispatcher_calls): Adjust call to
3108 redirect_call_stmt_to_callee
3109 (redirect_to_specific_clone): Likewise.
3110 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3111 Adjust calls to cgraph_edge::remove.
3112 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3113 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3114 (expand_call_inline): Adjust call to cgraph_edge::remove.
3116 2020-01-09 Martin Liska <mliska@suse.cz>
3118 * params.opt: Set Optimization for
3119 param_max_speculative_devirt_maydefs.
3121 2020-01-09 Martin Sebor <msebor@redhat.com>
3125 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3127 2020-01-09 Martin Liska <mliska@suse.cz>
3129 * auto-profile.c (auto_profile): Use opt_for_fn
3131 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3132 (propagate_vals_across_arith_jfunc): Likewise.
3133 (hint_time_bonus): Likewise.
3134 (incorporate_penalties): Likewise.
3135 (good_cloning_opportunity_p): Likewise.
3136 (perform_estimation_of_a_value): Likewise.
3137 (estimate_local_effects): Likewise.
3138 (ipcp_propagate_stage): Likewise.
3139 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3140 (set_switch_stmt_execution_predicate): Likewise.
3141 (analyze_function_body): Likewise.
3142 * ipa-inline-analysis.c (offline_size): Likewise.
3143 * ipa-inline.c (early_inliner): Likewise.
3144 * ipa-prop.c (ipa_analyze_node): Likewise.
3145 (ipcp_transform_function): Likewise.
3146 * ipa-sra.c (process_scan_results): Likewise.
3147 (ipa_sra_summarize_function): Likewise.
3148 * params.opt: Rename ipcp-unit-growth to
3149 ipa-cp-unit-growth. Add Optimization for various
3150 IPA-related parameters.
3152 2020-01-09 Richard Biener <rguenther@suse.de>
3155 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3157 2020-01-09 Richard Biener <rguenther@suse.de>
3159 PR tree-optimization/93040
3160 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3162 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3164 * common/config/avr/avr-common.c (avr_option_optimization_table)
3165 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3167 2020-01-09 Martin Liska <mliska@suse.cz>
3169 * cgraphclones.c (symbol_table::materialize_all_clones):
3170 Use cgraph_node::dump_name.
3172 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3175 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3176 output_operand_lossage instead of gcc_unreachable.
3177 * doc/md.texi (riscv f constraint): Fix typo.
3180 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3181 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3182 CONST_SCALAR_INT_P instead of CONST_INT_P.
3183 (*subv<mode>4_1): Rename to ...
3184 (subv<mode>4_1): ... this.
3185 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3186 define_insn_and_split patterns.
3187 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3190 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3192 * vec.c (class selftest::count_dtor): New class.
3193 (selftest::test_auto_delete_vec): New test.
3194 (selftest::vec_c_tests): Call it.
3195 * vec.h (class auto_delete_vec): New class template.
3196 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3198 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3200 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3202 2020-01-08 Jim Wilson <jimw@sifive.com>
3204 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3205 use of TLS_MODEL_LOCAL_EXEC when not pic.
3207 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3209 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3212 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3215 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3216 *stack_protect_set_3 peephole2): Also check that the second
3217 insns source is general_operand.
3220 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3221 predicate for output operand instead of register_operand.
3222 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3223 memory destination and non-memory operands[2].
3225 2020-01-08 Martin Liska <mliska@suse.cz>
3227 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3228 ::dump_asm_name instead of (::name or ::asm_name).
3229 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3230 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3231 (analyze_functions): Likewise.
3232 (expand_all_functions): Likewise.
3233 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3234 (propagate_bits_across_jump_function): Likewise.
3235 (dump_profile_updates): Likewise.
3236 (ipcp_store_bits_results): Likewise.
3237 (ipcp_store_vr_results): Likewise.
3238 * ipa-devirt.c (dump_targets): Likewise.
3239 * ipa-fnsummary.c (analyze_function_body): Likewise.
3240 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3241 (process_hsa_functions): Likewise.
3242 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3243 (set_alias_uids): Likewise.
3244 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3245 * ipa-inline.c (recursive_inlining): Likewise.
3246 (inline_to_all_callers_1): Likewise.
3247 (ipa_inline): Likewise.
3248 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3249 (ipa_propagate_frequency): Likewise.
3250 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3251 (remove_described_reference): Likewise.
3252 * ipa-pure-const.c (worse_state): Likewise.
3253 (check_retval_uses): Likewise.
3254 (analyze_function): Likewise.
3255 (propagate_pure_const): Likewise.
3256 (propagate_nothrow): Likewise.
3257 (dump_malloc_lattice): Likewise.
3258 (propagate_malloc): Likewise.
3259 (pass_local_pure_const::execute): Likewise.
3260 * ipa-visibility.c (optimize_weakref): Likewise.
3261 (function_and_variable_visibility): Likewise.
3262 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3263 (ipa_discover_variable_flags): Likewise.
3264 * lto-streamer-out.c (output_function): Likewise.
3265 (output_constructor): Likewise.
3266 * tree-inline.c (copy_bb): Likewise.
3267 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3268 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3270 2020-01-08 Richard Biener <rguenther@suse.de>
3273 * tree-eh.c (sink_clobbers): Update virtual operands for
3274 the first and last stmt only. Add a dry-run capability.
3275 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3276 after CFG manipulations and in RPO order to catch all
3277 secondary opportunities reliably.
3279 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3282 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3284 2019-01-08 Richard Biener <rguenther@suse.de>
3287 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3288 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3289 virtual operand, also updating SSA use.
3290 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3291 Update stmt after resetting virtual operand.
3292 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3293 * gimple-iterator.c (gsi_remove): When not removing the stmt
3294 permanently do not delink immediate uses or mark the stmt modified.
3296 2020-01-08 Martin Liska <mliska@suse.cz>
3298 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3299 (ipa_call_context::estimate_size_and_time): Likewise.
3300 (inline_analyze_function): Likewise.
3302 2020-01-08 Martin Liska <mliska@suse.cz>
3304 * cgraph.c (cgraph_node::dump): Use systematically
3307 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3309 Add -nodevicespecs option for avr.
3312 * config/avr/avr.opt (-nodevicespecs): New driver option.
3313 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3314 "-specs=device-specs/..." if that option is not set.
3315 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3317 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3319 Implement 64-bit double functions for avr.
3322 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3323 --with-double-comparison.
3324 * doc/install.texi: Document them.
3325 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3326 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3327 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3328 * doc/invoke.texi (AVR Built-in Macros): Document them.
3329 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3330 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3331 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3333 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3336 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3337 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3338 when only building rm-profile multilibs.
3340 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3343 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3344 lattice for a value to check.
3345 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3346 finite propagation in self-recursive scc.
3348 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3350 * ipa-inline.c (caller_growth_limits): Restore the AND.
3352 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3354 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3355 (VEC_ALLREG_ALT): New iterator.
3356 (VEC_ALLREG_INT_MODE): New iterator.
3357 (VCMP_MODE): New iterator.
3358 (VCMP_MODE_INT): New iterator.
3359 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3360 (vec_cmp<u>v64qidi): New define_expand.
3361 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3362 (vec_cmpu<mode>di_exec): New define_expand.
3363 (vec_cmp<u>v64qidi_exec): New define_expand.
3364 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3365 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3366 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3367 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3368 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3369 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3370 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3371 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3372 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3373 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3375 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3376 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3378 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3380 * config/gcn/constraints.md (DA): Update description and match.
3382 (Db): New constraint.
3383 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3385 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3386 Implement 'Db' mixed immediate type.
3387 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3388 (addcv64si3_dup<exec_vcc>): Delete.
3389 (subcv64si3<exec_vcc>): Rework constraints.
3390 (addv64di3): Rework constraints.
3391 (addv64di3_exec): Rework constraints.
3392 (subv64di3): Rework constraints.
3393 (addv64di3_dup): Delete.
3394 (addv64di3_dup_exec): Delete.
3395 (addv64di3_zext): Rework constraints.
3396 (addv64di3_zext_exec): Rework constraints.
3397 (addv64di3_zext_dup): Rework constraints.
3398 (addv64di3_zext_dup_exec): Rework constraints.
3399 (addv64di3_zext_dup2): Rework constraints.
3400 (addv64di3_zext_dup2_exec): Rework constraints.
3401 (addv64di3_sext_dup2): Rework constraints.
3402 (addv64di3_sext_dup2_exec): Rework constraints.
3404 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3406 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3407 existing target checks.
3409 2020-01-07 Richard Biener <rguenther@suse.de>
3411 * doc/install.texi: Bump minimal supported MPC version.
3413 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3415 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3416 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3417 * langhooks.c: Include stor-layout.h.
3418 (lhd_simulate_enum_decl): New function.
3419 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3420 handle_arm_sve_h for the LTO frontend.
3421 (register_vector_type): Cope with null returns from pushdecl.
3423 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3425 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3426 (aarch64_sve::nvectors_if_data_type): Replace with...
3427 (aarch64_sve::builtin_type_p): ...this.
3428 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3429 (find_vector_type): Delete.
3430 (add_sve_type_attribute): New function.
3431 (lookup_sve_type_attribute): Likewise.
3432 (register_builtin_types): Add an "SVE type" attribute to each type.
3433 (register_tuple_type): Likewise.
3434 (svbool_type_p, nvectors_if_data_type): Delete.
3435 (mangle_builtin_type): Use lookup_sve_type_attribute.
3436 (builtin_type_p): Likewise. Add an overload that returns the
3437 number of constituent vector and predicate registers.
3438 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3439 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3440 instead of aarch64_sve_argument_p.
3441 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3442 (aarch64_pass_by_reference): Likewise.
3443 (aarch64_function_value_1): Likewise.
3444 (aarch64_return_in_memory): Likewise.
3445 (aarch64_layout_arg): Likewise.
3447 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3449 PR tree-optimization/93156
3450 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3451 least significant bit is always clear.
3453 PR tree-optimization/93118
3454 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3455 simplifier with two intermediate conversions.
3457 2020-01-07 Martin Liska <mliska@suse.cz>
3459 * params.opt: Add Optimization for various parameters.
3461 2020-01-07 Martin Liska <mliska@suse.cz>
3464 * doc/extend.texi: Explain cloning for target_clone
3467 2020-01-07 Martin Liska <mliska@suse.cz>
3469 PR tree-optimization/92860
3470 * common.opt: Make in Optimization option
3471 as it is affected by -O0, which is an Optimization
3473 * tree-inline.c (tree_inlinable_function_p):
3474 Use opt_for_fn for warn_inline.
3475 (expand_call_inline): Likewise.
3477 2020-01-07 Martin Liska <mliska@suse.cz>
3479 PR tree-optimization/92860
3480 * common.opt: Make flag_ree as optimization
3483 2020-01-07 Martin Liska <mliska@suse.cz>
3485 PR optimization/92860
3486 * params.opt: Mark param_min_crossjump_insns with Optimization
3489 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3491 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3492 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3494 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3496 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3497 helper function to return the valid addressing formats for a given
3498 hard register and mode.
3499 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3501 * config/rs6000/constraints.md (Q constraint): Update
3503 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3506 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3507 Use 'Q' for doing vector extract from memory.
3508 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3510 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3511 doing vector extract from memory.
3512 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3513 extract from memory.
3515 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3516 for the offset being 34-bits when -mcpu=future is used.
3518 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3520 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3521 instead of cmpib_comparison_operator in cmpib patterns.
3522 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3523 of cmpib_comparison_operator. Revise comment.
3525 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3527 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3528 in an IFN_DIV_POW2 node to be equal.
3530 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3532 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3533 (vect_check_scalar_mask): ...this.
3534 (vectorizable_store, vectorizable_load): Update call accordingly.
3535 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3536 argument in calls to conditional internal functions.
3538 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3540 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3541 '0' matching inputs.
3542 (subv64di3_exec): Likewise.
3544 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3546 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3547 * doc/md.texi (movstr): Likewise.
3549 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3551 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3554 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3556 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3558 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3559 to a temporary file and use move-if-change to update the real
3560 file where necessary.
3562 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3564 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3565 rather than Upa for CPY /M.
3567 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3569 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3572 2020-01-06 Martin Liska <mliska@suse.cz>
3574 PR tree-optimization/92860
3575 * params.opt: Mark param_max_combine_insns with Optimization
3578 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3581 * config/i386/i386.md (SWIDWI): New mode iterator.
3582 (DWI, dwi): Add TImode variants.
3583 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3584 <general_hilo_operand> instead of <general_operand>. Use
3585 CONST_SCALAR_INT_P instead of CONST_INT_P.
3586 (*addv<mode>4_1): Rename to ...
3587 (addv<mode>4_1): ... this.
3588 (QWI): New mode attribute.
3589 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3590 define_insn_and_split patterns.
3591 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3593 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3594 <general_hilo_operand> instead of <general_operand>.
3595 (*addcarry<mode>_1): New define_insn.
3596 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3598 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3600 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3601 Use "call" instead of "set".
3603 2020-01-03 Martin Jambor <mjambor@suse.cz>
3606 * ipa-cp.c (print_all_lattices): Skip functions without info.
3608 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3611 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3612 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3613 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3614 for 'e' simd clones.
3617 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3619 (mprefer-vector-width=): Add Save.
3620 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3621 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3622 (ix86_debug_options, ix86_function_specific_print): Adjust
3623 ix86_target_string callers.
3624 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3625 (ix86_valid_target_attribute_tree): Likewise.
3626 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3627 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3628 ix86_target_string caller.
3631 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3632 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3633 instead of gen_int_shift_amount + convert_modes.
3635 PR rtl-optimization/93088
3636 * loop-iv.c (find_single_def_src): Punt after looking through
3637 128 reg copies for regs with single definitions. Move definitions
3640 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3642 * config/arm/arm-c.c (arm_cpu_builtins): Define
3643 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3644 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3645 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3646 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3647 * config/arm/arm-tables.opt: Regenerated.
3648 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3649 arm_arch_i8mm and arm_arch_bf16 when enabled.
3650 * config/arm/arm.h (TARGET_I8MM): New macro.
3651 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3652 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3653 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3654 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3655 (v8_6_a_simd_variants): New.
3656 (v8_*_a_simd_variants): Add i8mm and bf16.
3657 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3659 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3662 * predict.c (compute_function_frequency): Don't call
3663 warn_function_cold on functions that already have cold attribute.
3665 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3668 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3669 COMDAT group function labels in .data.rel.ro.local section.
3670 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3673 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3674 comparison_operator in B and S integer comparisons. Likewise, use
3675 ordered_comparison_operator instead of cmpib_comparison_operator in
3677 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3679 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3681 Update copyright years.
3683 * gcc.c (process_command): Update copyright notice dates.
3684 * gcov-dump.c (print_version): Ditto.
3685 * gcov.c (print_version): Ditto.
3686 * gcov-tool.c (print_version): Ditto.
3687 * gengtype.c (create_file): Ditto.
3688 * doc/cpp.texi: Bump @copying's copyright year.
3689 * doc/cppinternals.texi: Ditto.
3690 * doc/gcc.texi: Ditto.
3691 * doc/gccint.texi: Ditto.
3692 * doc/gcov.texi: Ditto.
3693 * doc/install.texi: Ditto.
3694 * doc/invoke.texi: Ditto.
3696 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3698 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3701 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3703 PR tree-optimization/93098
3704 * match.pd (popcount): For shift amounts, use integer_onep
3705 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3706 tests. Make sure that precision is power of two larger than or equal
3707 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3708 instead of ULL suffixed constants. Formatting fixes.
3710 Copyright (C) 2020 Free Software Foundation, Inc.
3712 Copying and distribution of this file, with or without modification,
3713 are permitted in any medium without royalty provided the copyright
3714 notice and this notice are preserved.