1 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
4 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
7 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
9 Try to generate zero-based comparisons.
10 * config/cris/cris.c (cris_reduce_compare): New function.
11 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
12 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
13 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
15 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
18 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
19 in Thumb state and also as a destination in Arm state. Add T16
22 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
24 * md.texi (Define Subst): Match closing paren in example.
26 2020-02-10 Jakub Jelinek <jakub@redhat.com>
30 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
33 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
36 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
37 but different source value.
38 (adjust_callers_for_value_intersection): New function.
39 (gather_edges_for_value): Adjust order of callers to let a
40 non-self-recursive caller be the first element.
41 (self_recursive_pass_through_p): Add a new parameter "simple", and
42 check generalized self-recursive pass-through jump function.
43 (self_recursive_agg_pass_through_p): Likewise.
44 (find_more_scalar_values_for_callers_subset): Compute value from
45 pass-through jump function for self-recursive.
46 (intersect_with_plats): Cleanup previous implementation code for value
47 itersection with self-recursive call edge.
48 (intersect_with_agg_replacements): Likewise.
49 (intersect_aggregates_with_edge): Deduce value from pass-through jump
50 function for self-recursive call edge. Cleanup previous implementation
51 code for value intersection with self-recursive call edge.
52 (decide_whether_version_node): Remove dead callers and adjust order
53 to let a non-self-recursive caller be the first element.
55 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
57 * recog.c: Move pass_split_before_sched2 code in front of
58 pass_split_before_regstack.
59 (pass_data_split_before_sched2): Rename pass to split3 from split4.
60 (pass_data_split_before_regstack): Rename pass to split4 from split3.
61 (rest_of_handle_split_before_sched2): Remove.
62 (pass_split_before_sched2::execute): Unconditionally call
64 (enable_split_before_sched2): New function.
65 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
66 (pass_split_before_regstack::gate): Ditto.
67 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
68 Update name check for renamed split4 pass.
69 * config/sh/sh.c (register_sh_passes): Update pass insertion
70 point for renamed split4 pass.
72 2020-02-09 Jakub Jelinek <jakub@redhat.com>
74 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
75 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
76 copying them around between host and target.
78 2020-02-08 Andrew Pinski <apinski@marvell.com>
81 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
82 STRICT_ALIGNMENT also.
84 2020-02-08 Jim Wilson <jimw@sifive.com>
87 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
89 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
90 Jakub Jelinek <jakub@redhat.com>
93 * config/i386/i386.h (CALL_USED_REGISTERS): Make
94 xmm16-xmm31 call-used even in 64-bit ms-abi.
96 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
98 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
99 (simd_ummla, simd_usmmla): Likewise.
100 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
101 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
104 2020-02-07 Richard Biener <rguenther@suse.de>
107 * tree-inline.c (fold_marked_statements): Do a PRE walk,
108 skipping unreachable regions.
109 (optimize_inline_calls): Skip folding stmts when we didn't
112 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
115 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
116 Don't return aggregates with only SFmode and DFmode in SSE
118 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
120 2020-02-07 Jakub Jelinek <jakub@redhat.com>
123 * config/rs6000/rs6000-logue.c
124 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
125 if it fails, move rs into end_addr and retry. Add
126 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
127 the insn pattern doesn't describe well what exactly happens to
131 * config/i386/predicates.md (avx_identity_operand): Remove.
132 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
133 (avx_<castmode><avxsizesuffix>_<castmode>,
134 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
135 a VEC_CONCAT of the operand and UNSPEC_CAST.
136 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
137 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
141 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
142 recog_data.insn if distance_non_agu_define changed it.
144 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
147 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
148 we only had X-FORM (reg+reg) addressing for vectors. Also before
149 ISA 3.0, we only had X-FORM addressing for scalars in the
150 traditional Altivec registers.
152 2020-02-06 <zhongyunde@huawei.com>
153 Vladimir Makarov <vmakarov@redhat.com>
155 PR rtl-optimization/93561
156 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
159 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
161 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
164 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
166 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
167 where the low and the high 32 bits are equal to each other specially,
168 with an rldimi instruction.
170 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
172 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
174 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
176 * config/arm/arm-tables.opt: Regenerate.
178 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
181 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
182 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
183 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
185 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
187 PR rtl-optimization/87763
188 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
190 2020-02-06 Delia Burduv <delia.burduv@arm.com>
192 * config/aarch64/aarch64-simd-builtins.def
193 (bfmlaq): New built-in function.
194 (bfmlalb): New built-in function.
195 (bfmlalt): New built-in function.
196 (bfmlalb_lane): New built-in function.
197 (bfmlalt_lane): New built-in function.
198 * config/aarch64/aarch64-simd.md
199 (aarch64_bfmmlaqv4sf): New pattern.
200 (aarch64_bfmlal<bt>v4sf): New pattern.
201 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
202 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
203 (vbfmlalbq_f32): New intrinsic.
204 (vbfmlaltq_f32): New intrinsic.
205 (vbfmlalbq_lane_f32): New intrinsic.
206 (vbfmlaltq_lane_f32): New intrinsic.
207 (vbfmlalbq_laneq_f32): New intrinsic.
208 (vbfmlaltq_laneq_f32): New intrinsic.
209 * config/aarch64/iterators.md (BF_MLA): New int iterator.
210 (bt): New int attribute.
212 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
214 * config/i386/i386.md (*pushtf): Emit "#" instead of
215 calling gcc_unreachable in insn output.
218 (*pushsf_rex64): Ditto for alternatives other than 1.
219 (*pushsf): Ditto for alternatives other than 1.
221 2020-02-06 Martin Liska <mliska@suse.cz>
223 PR gcov-profile/91971
224 PR gcov-profile/93466
225 * coverage.c (coverage_init): Revert mangling of
226 path into filename. It can lead to huge filename length.
227 Creation of subfolders seem more natural.
229 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
232 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
233 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
234 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
236 2020-02-06 Jakub Jelinek <jakub@redhat.com>
239 * config/i386/predicates.md (avx_identity_operand): New predicate.
240 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
241 define_insn_and_split.
244 * omp-low.c (use_pointer_for_field): For nested constructs, also
245 look for map clauses on target construct.
246 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
247 taskreg_nesting_level.
250 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
251 shared clause, call omp_notice_variable on outer context if any.
253 2020-02-05 Jason Merrill <jason@redhat.com>
256 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
257 non-zero address even if weak and not yet defined.
259 2020-02-05 Martin Sebor <msebor@redhat.com>
261 PR tree-optimization/92765
262 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
263 * tree-ssa-strlen.c (compute_string_length): Remove.
264 (determine_min_objsize): Remove.
265 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
266 Avoid using type size as the upper bound on string length.
267 (handle_builtin_string_cmp): Add an argument. Adjust.
268 (strlen_check_and_optimize_call): Pass additional argument to
269 handle_builtin_string_cmp.
271 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
273 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
274 (*pushdi2_rex64 peephole2): Unconditionally split after
276 (*ashl<mode>3_doubleword): Ditto.
277 (*<shift_insn><mode>3_doubleword): Ditto.
279 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
282 * config/rs6000/rs6000.c (get_vector_offset): Fix
284 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
286 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
288 2020-02-05 David Malcolm <dmalcolm@redhat.com>
291 (Special Functions for Debugging the Analyzer): Update description
292 of __analyzer_dump_exploded_nodes.
294 2020-02-05 Jakub Jelinek <jakub@redhat.com>
297 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
298 include sets and not clobbers in the vzeroupper pattern.
299 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
300 the parallel has 17 (64-bit) or 9 (32-bit) elts.
301 (*avx_vzeroupper_1): New define_insn_and_split.
304 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
305 don't run when !optimize.
306 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
309 2020-02-05 Richard Biener <rguenther@suse.de>
312 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
313 checks before matching calls.
315 2020-02-05 Jakub Jelinek <jakub@redhat.com>
317 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
318 function comment typo.
321 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
322 simd_clone_create failed when i == 0, adjust clone->nargs by
325 2020-02-05 Martin Liska <mliska@suse.cz>
328 * doc/invoke.texi: Document that one should
329 not combine ASLR and -fpch.
331 2020-02-04 Richard Biener <rguenther@suse.de>
333 PR tree-optimization/93538
334 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
336 2020-02-04 Richard Biener <rguenther@suse.de>
338 PR tree-optimization/91123
339 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
340 (vn_walk_cb_data::last_vuse): New member.
341 (vn_walk_cb_data::saved_operands): Likewsie.
342 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
343 (vn_walk_cb_data::push_partial_def): Use finish.
344 (vn_reference_lookup_2): Update last_vuse and use finish if
345 we've saved operands.
346 (vn_reference_lookup_3): Use finish and update calls to
347 push_partial_defs everywhere. When translating through
348 memcpy or aggregate copies save off operands and alias-set.
349 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
350 operation for redundant store removal.
352 2020-02-04 Richard Biener <rguenther@suse.de>
354 PR tree-optimization/92819
355 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
356 generating more stmts than before.
358 2020-02-04 Martin Liska <mliska@suse.cz>
360 * config/arm/arm.c (arm_gen_far_branch): Move the function
361 outside of selftests.
363 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
365 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
366 function to adjust PC-relative vector addresses.
367 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
368 handle vectors with PC-relative addresses.
370 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
372 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
374 (hard_reg_and_mode_to_addr_mask): Delete.
375 (rs6000_adjust_vec_address): If the original vector address
376 was REG+REG or REG+OFFSET and the element is not zero, do the add
377 of the elements in the original address before adding the offset
378 for the vector element. Use address_to_insn_form to validate the
379 address using the register being loaded, rather than guessing
380 whether the address is a DS-FORM or DQ-FORM address.
382 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
384 * config/rs6000/rs6000.c (get_vector_offset): New helper function
385 to calculate the offset in memory from the start of a vector of a
386 particular element. Add code to keep the element number in
387 bounds if the element number is variable.
388 (rs6000_adjust_vec_address): Move calculation of offset of the
389 vector element to get_vector_offset.
390 (rs6000_split_vec_extract_var): Do not do the initial AND of
391 element here, move the code to get_vector_offset.
393 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
395 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
398 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
400 * config/rs6000/constraints.md: Improve documentation.
402 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
405 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
406 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
408 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
410 * config.gcc: Remove "carrizo" support.
411 * config/gcn/gcn-opts.h (processor_type): Likewise.
412 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
413 * config/gcn/gcn.opt (gpu_type): Likewise.
414 * config/gcn/t-omp-device: Likewise.
416 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
419 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
420 * config/arm/arm.c (arm_gen_far_branch): New function
422 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
424 2020-02-03 Julian Brown <julian@codesourcery.com>
425 Tobias Burnus <tobias@codesourcery.com>
427 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
429 2020-02-03 Jakub Jelinek <jakub@redhat.com>
432 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
433 valid RTL to sum up the lowest and second lowest bytes of the popcnt
436 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
438 PR rtl-optimization/91333
439 * ira-color.c (struct allocno_color_data): Add member
441 (init_allocno_threads): Set the member up.
442 (bucket_allocno_compare_func): Add compare hard reg
445 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
447 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
449 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
450 * config.in: Regenerated.
451 * configure: Regenerated.
452 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
453 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
454 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
456 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
458 * configure: Regenerate.
460 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
462 PR rtl-optimization/91333
463 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
464 reg preferences comparison up.
466 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
468 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
469 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
470 aarch64-sve-builtins-base.h.
471 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
472 aarch64-sve-builtins-base.cc.
473 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
474 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
476 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
477 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
478 (svcvtnt): New functions.
479 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
480 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
481 (svcvtnt): New functions.
482 (svcvt): Add a form that converts f32 to bf16.
483 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
484 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
486 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
487 Treat B as bfloat16_t.
488 (ternary_bfloat_lane_base): New class.
489 (ternary_bfloat_def): Likewise.
490 (ternary_bfloat): New shape.
491 (ternary_bfloat_lane_def): New class.
492 (ternary_bfloat_lane): New shape.
493 (ternary_bfloat_lanex2_def): New class.
494 (ternary_bfloat_lanex2): New shape.
495 (ternary_bfloat_opt_n_def): New class.
496 (ternary_bfloat_opt_n): New shape.
497 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
498 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
499 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
500 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
501 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
502 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
503 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
504 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
505 the pattern off the narrow mode instead of the wider one.
506 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
507 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
508 (sve_fp_op): Handle them.
509 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
510 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
512 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
514 * config/aarch64/arm_sve.h: Include arm_bf16.h.
515 * config/aarch64/aarch64-modes.def (BF): Move definition before
516 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
517 (SVE_MODES): Handle BF modes.
518 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
520 (aarch64_full_sve_mode): Likewise.
521 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
523 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
524 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
525 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
526 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
528 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
530 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
531 (TYPES_all_data): Add bf16.
532 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
533 (register_tuple_type): Increase buffer size.
534 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
535 (bf16): New type suffix.
536 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
537 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
538 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
539 Change type from all_data to all_arith.
540 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
543 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
544 Matthew Malcomson <matthew.malcomson@arm.com>
545 Richard Sandiford <richard.sandiford@arm.com>
547 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
548 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
549 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
550 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
551 __ARM_FEATURE_MATMUL_FP64.
552 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
553 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
554 be disabled at the same time.
555 (f32mm): New extension.
556 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
557 (AARCH64_FL_F64MM): Bump to the next bit up.
558 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
559 (TARGET_SVE_F64MM): New macros.
560 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
561 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
562 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
563 (UNSPEC_ZIP2Q): New unspeccs.
564 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
565 (optab, sur, perm_insn): Handle the new unspecs.
566 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
567 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
568 TARGET_SVE_F64MM instead of separate tests.
569 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
570 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
571 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
572 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
573 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
574 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
575 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
576 (TYPES_s_signed): New macro.
577 (TYPES_s_integer): Use it.
578 (TYPES_d_float): New macro.
579 (TYPES_d_data): Use it.
580 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
581 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
582 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
583 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
585 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
587 (ternary_resize2_lane_base): Likewise.
588 (ternary_resize2_base): New class.
589 (ternary_qq_lane_base): Likewise.
590 (ternary_intq_uintq_lane_def): Likewise.
591 (ternary_intq_uintq_lane): New shape.
592 (ternary_intq_uintq_opt_n_def): New class
593 (ternary_intq_uintq_opt_n): New shape.
594 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
595 (ternary_uintq_intq_def): New class.
596 (ternary_uintq_intq): New shape.
597 (ternary_uintq_intq_lane_def): New class.
598 (ternary_uintq_intq_lane): New shape.
599 (ternary_uintq_intq_opt_n_def): New class.
600 (ternary_uintq_intq_opt_n): New shape.
601 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
602 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
603 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
604 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
606 (svdotprod_lane_impl): ...this new class.
607 (svmmla_impl, svusdot_impl): New classes.
608 (svdot_lane): Update to use svdotprod_lane_impl.
609 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
610 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
612 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
613 function, with no types defined.
614 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
615 AARCH64_FL_I8MM functions.
616 (svmmla): New AARCH64_FL_F32MM function.
617 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
618 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
619 AARCH64_FL_F64MM function.
620 (REQUIRED_EXTENSIONS):
622 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
624 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
627 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
629 * config/i386/i386.md (*movoi_internal_avx): Do not check for
630 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
631 (*movti_internal): Do not check for
632 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
633 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
634 just after check for TARGET_AVX.
635 (*movdf_internal): Ditto.
636 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
637 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
638 * config/i386/sse.md (mov<mode>_internal): Only check
639 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
640 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
641 (<sse>_andnot<mode>3<mask_name>): Move check for
642 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
643 (<code><mode>3<mask_name>): Ditto.
644 (*andnot<mode>3): Ditto.
646 (*<code><mode>3): Ditto.
648 (*andnot<VI:mode>3): Remove
649 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
650 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
651 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
652 (sse4_1_blendv<ssemodesuffix>): Ditto.
653 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
654 Explain that tune applies to 128bit instructions only.
656 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
658 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
659 to definition of hsa_kernel_description. Parse assembly to find SGPR
660 and VGPR count of kernel and store in hsa_kernel_description.
662 2020-01-31 Tamar Christina <tamar.christina@arm.com>
664 PR rtl-optimization/91838
665 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
666 to truncate if allowed or reject combination.
668 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
670 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
671 (find_inv_vars_cb): Likewise.
673 2020-01-31 David Malcolm <dmalcolm@redhat.com>
675 * calls.c (special_function_p): Split out the check for DECL_NAME
676 being non-NULL and fndecl being extern at file scope into a
677 new maybe_special_function_p and call it. Drop check for fndecl
678 being non-NULL that was after a usage of DECL_NAME (fndecl).
679 * tree.h (maybe_special_function_p): New inline function.
681 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
683 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
684 (mask_gather_load<mode>): ... here, and zero-initialize the
686 (maskload<mode>di): Zero-initialize the destination.
689 2020-01-30 David Malcolm <dmalcolm@redhat.com>
692 * doc/analyzer.texi (Limitations): Note that constraints on
693 floating-point values are currently ignored.
695 2020-01-30 Jakub Jelinek <jakub@redhat.com>
698 * symtab.c (symtab_node::noninterposable_alias): If localalias
699 already exists, but is not usable, append numbers after it until
700 a unique name is found. Formatting fix.
703 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
706 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
708 * config/gcn/gcn.c (print_operand): Handle LTGT.
709 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
711 2020-01-30 Richard Biener <rguenther@suse.de>
713 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
714 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
716 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
718 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
719 without a DECL in .data.rel.ro.local.
721 2020-01-30 Jakub Jelinek <jakub@redhat.com>
724 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
729 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
730 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
731 any_extend code iterator instead of always zero_extend.
732 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
733 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
734 Use any_extend code iterator instead of always zero_extend.
735 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
736 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
737 Use any_extend code iterator instead of always zero_extend.
738 (*sse2_pmovmskb_ext): New define_insn.
739 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
742 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
743 (*popcountsi2_zext_falsedep): New define_insn.
745 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
747 * config.in: Regenerated.
748 * configure: Regenerated.
750 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
753 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
754 LLVM's assembler changed the default in version 9.
756 2020-01-24 Jeff Law <law@redhat.com>
758 PR tree-optimization/89689
759 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
761 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
765 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
767 PR rtl-optimization/87763
768 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
769 simplification to handle subregs as well as bare regs.
770 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
772 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
775 * ira.c (ira): Revert use of simplified LRA algorithm.
777 2020-01-29 Martin Jambor <mjambor@suse.cz>
779 PR tree-optimization/92706
780 * tree-sra.c (struct access): Fields first_link, last_link,
781 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
782 next_rhs_queued and grp_rhs_queued respectively, new fields
783 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
784 (struct assign_link): Field next renamed to next_rhs, new field
785 next_lhs. Updated comment.
786 (work_queue_head): Renamed to rhs_work_queue_head.
787 (lhs_work_queue_head): New variable.
788 (add_link_to_lhs): New function.
789 (relink_to_new_repr): Also relink LHS lists.
790 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
791 (add_access_to_lhs_work_queue): New function.
792 (pop_access_from_work_queue): Renamed to
793 pop_access_from_rhs_work_queue.
794 (pop_access_from_lhs_work_queue): New function.
795 (build_accesses_from_assign): Also add links to LHS lists and to LHS
797 (child_would_conflict_in_lacc): Renamed to
798 child_would_conflict_in_acc. Adjusted parameter names.
799 (create_artificial_child_access): New parameter set_grp_read, use it.
800 (subtree_mark_written_and_enqueue): Renamed to
801 subtree_mark_written_and_rhs_enqueue.
802 (propagate_subaccesses_across_link): Renamed to
803 propagate_subaccesses_from_rhs.
804 (propagate_subaccesses_from_lhs): New function.
805 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
808 2020-01-29 Martin Jambor <mjambor@suse.cz>
810 PR tree-optimization/92706
811 * tree-sra.c (struct access): Adjust comment of
812 grp_total_scalarization.
813 (find_access_in_subtree): Look for single children spanning an entire
815 (scalarizable_type_p): Allow register accesses, adjust callers.
816 (completely_scalarize): Remove function.
817 (scalarize_elem): Likewise.
818 (create_total_scalarization_access): Likewise.
819 (sort_and_splice_var_accesses): Do not track total scalarization
821 (analyze_access_subtree): New parameter totally, adjust to new meaning
822 of grp_total_scalarization.
823 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
824 (can_totally_scalarize_forest_p): New function.
825 (create_total_scalarization_access): Likewise.
826 (create_total_access_and_reshape): Likewise.
827 (total_should_skip_creating_access): Likewise.
828 (totally_scalarize_subtree): Likewise.
829 (analyze_all_variable_accesses): Perform total scalarization after
830 subaccess propagation using the new functions above.
831 (initialize_constant_pool_replacements): Output initializers by
832 traversing the access tree.
834 2020-01-29 Martin Jambor <mjambor@suse.cz>
836 * tree-sra.c (verify_sra_access_forest): New function.
837 (verify_all_sra_access_forests): Likewise.
838 (create_artificial_child_access): Set parent.
839 (analyze_all_variable_accesses): Call the verifier.
841 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
843 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
844 if called on indirect edge.
845 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
846 speculative call if needed.
848 2020-01-29 Richard Biener <rguenther@suse.de>
850 PR tree-optimization/93428
851 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
852 permutation when the load node is created.
853 (vect_analyze_slp_instance): Re-use it here.
855 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
857 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
859 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
861 PR rtl-optimization/93272
862 * ira-lives.c (process_out_of_region_eh_regs): New function.
863 (process_bb_node_lives): Call it.
865 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
867 * coverage.c (read_counts_file): Make error message lowercase.
869 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
871 * profile-count.c (profile_quality_display_names): Fix ordering.
873 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
876 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
877 hash only when edge is first within the sequence.
878 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
879 (symbol_table::create_edge): Do not set target_prob.
880 (cgraph_edge::remove_caller): Watch for speculative calls when updating
882 (cgraph_edge::make_speculative): Drop target_prob parameter.
883 (cgraph_edge::speculative_call_info): Remove.
884 (cgraph_edge::first_speculative_call_target): New member function.
885 (update_call_stmt_hash_for_removing_direct_edge): New function.
886 (cgraph_edge::resolve_speculation): Rewrite to new API.
887 (cgraph_edge::speculative_call_for_target): New member function.
888 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
889 multiple speculation targets.
890 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
892 (verify_speculative_call): Verify that targets form an interval.
893 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
894 (cgraph_edge::first_speculative_call_target): New member function.
895 (cgraph_edge::next_speculative_call_target): New member function.
896 (cgraph_edge::speculative_call_target_ref): New member function.
897 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
898 (cgraph_edge): Remove target_prob.
899 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
900 Fix handling of speculative calls.
901 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
902 * ipa-fnsummary.c (analyze_function_body): Likewise.
903 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
904 * ipa-profile.c (dump_histogram): Fix formating.
905 (ipa_profile_generate_summary): Watch for overflows.
906 (ipa_profile): Do not require probablity to be 1/2; update to new API.
907 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
908 (update_indirect_edges_after_inlining): Update to new API.
909 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
911 * profile-count.h: (profile_probability::adjusted): New.
912 * tree-inline.c (copy_bb): Update to new speculative call API; fix
914 * value-prof.c (gimple_ic_transform): Rename to ...
915 (dump_ic_profile): ... this one; update dumping.
916 (stream_in_histogram_value): Fix formating.
917 (gimple_value_profile_transformations): Update.
919 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
922 * config/i386/i386.md (*movoi_internal_avx): Remove
923 TARGET_SSE_TYPELESS_STORES check.
924 (*movti_internal): Prefer TARGET_AVX over
925 TARGET_SSE_TYPELESS_STORES.
926 (*movtf_internal): Likewise.
927 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
928 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
929 from TARGET_SSE_TYPELESS_STORES.
931 2020-01-28 David Malcolm <dmalcolm@redhat.com>
933 * diagnostic-core.h (warning_at): Rename overload to...
934 (warning_meta): ...this.
935 (emit_diagnostic_valist): Delete decl of overload taking
937 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
938 (warning_at): Rename overload taking diagnostic_metadata to...
939 (warning_meta): ...this.
941 2020-01-28 Richard Biener <rguenther@suse.de>
943 PR tree-optimization/93439
944 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
945 * tree-cfg.c (move_sese_region_to_fn): ... here.
946 (verify_types_in_gimple_reference): Verify used cliques are
949 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
952 * config/i386/i386-options.c (set_ix86_tune_features): Add an
953 argument of a pointer to struct gcc_options and pass it to
954 parse_mtune_ctrl_str.
955 (ix86_function_specific_restore): Pass opts to
956 set_ix86_tune_features.
957 (ix86_option_override_internal): Likewise.
958 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
959 gcc_options and use it for x_ix86_tune_ctrl_string.
961 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
963 PR rtl-optimization/87763
964 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
965 simplification to handle subregs as well as bare regs.
966 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
968 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
970 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
971 for reduction chains that (now) include a call.
973 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
975 PR tree-optimization/92822
976 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
977 out the don't-care elements of a vector whose significant elements
978 are duplicates, make the don't-care elements duplicates too.
980 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
982 PR tree-optimization/93434
983 * tree-predcom.c (split_data_refs_to_components): Record which
984 components have had aliasing loads removed. Prevent store-store
985 commoning for all such components.
987 2020-01-28 Jakub Jelinek <jakub@redhat.com>
990 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
991 -1 or is_vshift is true, use new_vector with number of elts npatterns
992 rather than new_unary_operation.
994 PR tree-optimization/93454
995 * gimple-fold.c (fold_array_ctor_reference): Perform
996 elt_size.to_uhwi () just once, instead of calling it in every
997 iteration. Punt if that value is above size of the temporary
998 buffer. Decrease third native_encode_expr argument when
999 bufoff + elt_sz is above size of buf.
1001 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1003 * config/mips/mips.c (mips_declare_object_name)
1004 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1006 2020-01-27 Martin Liska <mliska@suse.cz>
1008 PR gcov-profile/93403
1009 * tree-profile.c (gimple_init_gcov_profiler): Generate
1010 both __gcov_indirect_call_profiler_v4 and
1011 __gcov_indirect_call_profiler_v4_atomic.
1013 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1016 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1018 (@aarch64_split_simd_mov<mode>): Use it.
1019 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1020 Leave the vec_extract patterns to handle 2-element vectors.
1021 (aarch64_simd_mov_from_<mode>high): Likewise.
1022 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1023 (vec_extractv2dfv1df): Likewise.
1025 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1027 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1028 jump conditions for *compare_condjump<GPI:mode>.
1030 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1033 * digraph.cc (test_edge::test_edge): Specify template for base
1036 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1038 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1040 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1042 * config/arc/arc-protos.h (gen_mlo): Remove.
1043 (gen_mhi): Likewise.
1044 * config/arc/arc.c (AUX_MULHI): Define.
1045 (arc_must_save_reister): Special handling for r58/59.
1046 (arc_compute_frame_size): Consider mlo/mhi registers.
1047 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1049 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1050 mlo/mhi name selection.
1051 (arc_restore_callee_saves): Don't early restore blink when ISR.
1052 (arc_expand_prologue): Add mlo/mhi saving.
1053 (arc_expand_epilogue): Add mlo/mhi restoring.
1056 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1057 numbering when MUL64 option is used.
1058 (DWARF2_FRAME_REG_OUT): Define.
1059 * config/arc/arc.md (arc600_stall): New pattern.
1060 (VUNSPEC_ARC_ARC600_STALL): Define.
1061 (mulsi64): Use correct mlo/mhi registers.
1062 (mulsi_600): Clean it up.
1063 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1065 (mhi_operand): Likewise.
1067 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1068 Petro Karashchenko <petro.karashchenko@ring.com>
1070 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1071 attributes if needed.
1072 (prepare_move_operands): Generate special unspec instruction for
1074 (arc_isuncached_mem_p): Propagate uncached attribute to each
1076 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1077 (VUNSPEC_ARC_STDI): Likewise.
1078 (ALLI): New mode iterator.
1079 (mALLI): New mode attribute.
1080 (lddi): New instruction pattern.
1082 (stdidi_split): Split instruction for architectures which are not
1083 supporting ll64 option.
1084 (lddidi_split): Likewise.
1086 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1088 PR rtl-optimization/92989
1089 * lra-lives.c (process_bb_lives): Update the live-in set before
1090 processing additional clobbers.
1092 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1094 PR rtl-optimization/93170
1095 * cselib.c (cselib_invalidate_regno_val): New function, split out
1097 (cselib_invalidate_regno): ...here.
1098 (cselib_invalidated_by_call_p): New function.
1099 (cselib_process_insn): Iterate over all the hard-register entries in
1100 REG_VALUES and invalidate any that cross call-clobbered registers.
1102 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1104 * dojump.c (split_comparison): Use HONOR_NANS rather than
1105 HONOR_SNANS when splitting LTGT.
1107 2020-01-27 Martin Liska <mliska@suse.cz>
1110 * opts.c (print_filtered_help): Exclude language-specific
1111 options from --help=common unless enabled in all FEs.
1113 2020-01-27 Martin Liska <mliska@suse.cz>
1115 * opts.c (print_help): Exclude params from
1116 all except --help=param.
1118 2020-01-27 Martin Liska <mliska@suse.cz>
1121 * config/i386/i386-features.c (make_resolver_func):
1122 Align the code with ppc64 target implementation.
1123 Do not generate a unique name for resolver function.
1125 2020-01-27 Richard Biener <rguenther@suse.de>
1127 PR tree-optimization/93397
1128 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1129 converted reduction chain SLP graph adjustment.
1131 2020-01-26 Marek Polacek <polacek@redhat.com>
1134 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1137 2020-01-26 Jason Merrill <jason@redhat.com>
1140 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1143 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1145 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1146 (rx_setmem): Likewise.
1148 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1151 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1152 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1153 drop <di> from constraint of last operand.
1156 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1157 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1158 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1160 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1163 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1166 2020-01-24 Jeff Law <law@redhat.com>
1168 PR tree-optimization/92788
1169 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1172 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1175 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1176 *avx_vperm_broadcast_<mode>,
1177 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1178 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1179 Move before avx2_perm<mode>/avx512f_perm<mode>.
1182 * simplify-rtx.c (simplify_const_unary_operation,
1183 simplify_const_binary_operation): Punt for mode precision above
1184 MAX_BITSIZE_MODE_ANY_INT.
1186 2020-01-24 Andrew Pinski <apinski@marvell.com>
1188 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1191 2020-01-24 Jeff Law <law@redhat.com>
1194 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1195 for REGs. Call output_operand_lossage to get more reasonable
1198 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1200 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1201 gcn_fp_compare_operator.
1202 (vec_cmpu<mode>di): Use gcn_compare_operator.
1203 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1204 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1205 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1206 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1207 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1208 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1209 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1210 gcn_fp_compare_operator.
1211 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1212 gcn_fp_compare_operator.
1213 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1214 gcn_fp_compare_operator.
1215 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1216 gcn_fp_compare_operator.
1218 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1220 * doc/install.texi (Cross-Compiler-Specific Options): Document
1221 `--with-toolexeclibdir' option.
1223 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1225 * target.def (flags_regnum): Also mention effect on delay slot filling.
1226 * doc/tm.texi: Regenerate.
1228 2020-01-23 Jeff Law <law@redhat.com>
1230 PR translation/90162
1231 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1233 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1236 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1239 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1241 PR rtl-optimization/93402
1242 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1245 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1247 * config.in: Regenerated.
1248 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1249 for TARGET_LIBC_GNUSTACK.
1250 * configure: Regenerated.
1251 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1252 found to be 2.31 or greater.
1254 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1256 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1258 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1259 (mips_asm_file_end): New function. Delegate to
1260 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1261 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1263 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1266 * config/i386/i386-modes.def (POImode): New mode.
1267 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1268 * config/i386/i386.md (DPWI): New mode attribute.
1269 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1271 (QPWI): ... this. Use POI instead of OI for TImode.
1272 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1273 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1276 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1279 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1281 (speculation_tracker_rev): New pattern.
1282 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1283 Use speculation_tracker_rev to track the inverse condition.
1285 2020-01-23 Richard Biener <rguenther@suse.de>
1287 PR tree-optimization/93381
1288 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1289 alias-set of the def as argument and record the first one.
1290 (vn_walk_cb_data::first_set): New member.
1291 (vn_reference_lookup_3): Pass the alias-set of the current def
1292 to push_partial_def. Fix alias-set used in the aggregate copy
1294 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1295 * real.c (clear_significand_below): Fix out-of-bound access.
1297 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1300 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1301 New define_insn patterns.
1303 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1305 * doc/sourcebuild.texi (check-function-bodies): Add an
1306 optional target/xfail selector.
1308 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1310 PR rtl-optimization/93124
1311 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1312 bare USE and CLOBBER insns.
1314 2020-01-22 Andrew Pinski <apinski@marvell.com>
1316 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1318 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1321 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1322 diagnostic_manager into "ana" namespace.
1323 * selftest-run-tests.c (selftest::run_tests): Update for move of
1324 selftest::run_analyzer_selftests to
1325 ana::selftest::run_analyzer_selftests.
1327 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1329 * cfgexpand.c (union_stack_vars): Update the size.
1331 2020-01-22 Richard Biener <rguenther@suse.de>
1333 PR tree-optimization/93381
1334 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1335 throughout, handle all conversions the same.
1337 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1340 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1341 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1342 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1343 Call force_reg on high_in2 unconditionally.
1345 2020-01-22 Martin Liska <mliska@suse.cz>
1347 PR tree-optimization/92924
1348 * profile.c (compute_value_histograms): Divide
1351 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1354 * output.h (assemble_name_resolve): Declare.
1355 * varasm.c (assemble_name_resolve): New function.
1356 (assemble_name): Use it.
1357 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1359 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1361 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1362 update_web_docs_git instead of update_web_docs_svn.
1364 2020-01-21 Andrew Pinski <apinski@marvell.com>
1367 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1368 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1369 (*tlsgd_small_<mode>): Likewise.
1370 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1371 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1372 register. Convert that register back to dest using convert_mode.
1374 2020-01-21 Jim Wilson <jimw@sifive.com>
1376 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1379 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1380 Uros Bizjak <ubizjak@gmail.com>
1383 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1385 (legitimize_tls_address): Do GNU2 TLS address computation in
1386 ptr_mode and zero-extend result to Pmode.
1387 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1388 :P with :PTR and Pmode with ptr_mode.
1389 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1390 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1391 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1393 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1396 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1397 the last two operands are CONST_INT_P before using them as such.
1399 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1401 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1402 to get the integer element types.
1404 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1406 * config/aarch64/aarch64-sve-builtins.h
1407 (function_expander::convert_to_pmode): Declare.
1408 * config/aarch64/aarch64-sve-builtins.cc
1409 (function_expander::convert_to_pmode): New function.
1410 (function_expander::get_contiguous_base): Use it.
1411 (function_expander::prepare_gather_address_operands): Likewise.
1412 * config/aarch64/aarch64-sve-builtins-sve2.cc
1413 (svwhilerw_svwhilewr_impl::expand): Likewise.
1415 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1418 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1419 cfun->machine->label_is_assembled.
1420 (aarch64_print_patchable_function_entry): New.
1421 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1422 * config/aarch64/aarch64.h (struct machine_function): New field,
1425 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1428 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1431 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1434 * cgraph.c (cgraph_edge::resolve_speculation,
1435 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1436 call_stmt_site_hash.
1438 2020-01-21 Martin Liska <mliska@suse.cz>
1440 * config/rs6000/rs6000.c (common_mode_defined): Remove
1443 2020-01-21 Richard Biener <rguenther@suse.de>
1445 PR tree-optimization/92328
1446 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1447 type when value-numbering same-sized store by inserting a
1449 (eliminate_dom_walker::eliminate_stmt): When eliminating
1450 a redundant store handle bit-reinterpretation of the same value.
1452 2020-01-21 Andrew Pinski <apinski@marvel.com>
1455 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1457 (prepare_block_for_update): This. Use a worklist instead of
1460 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1462 * gcc/config/arm/arm.c (clear_operation_p):
1463 Initialise last_regno, skip first iteration
1464 based on the first_set value and use ints instead
1465 of the unnecessary HOST_WIDE_INTs.
1467 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1470 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1471 compare_mode other than SFmode or DFmode.
1473 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1476 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1477 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1478 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1480 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1482 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1484 2020-01-20 Andrew Pinski <apinski@marvell.com>
1487 * targhooks.c (default_print_patchable_function_entry): Use
1488 output_asm_insn to emit the nop instruction.
1490 2020-01-20 Fangrui Song <maskray@google.com>
1493 * targhooks.c (default_print_patchable_function_entry): Align to
1496 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1499 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1500 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1501 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1502 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1503 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1504 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1505 Remove the {q} suffix from lea.
1506 (*tls_dynamic_gnu2_call_64): Renamed to ...
1507 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1508 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1509 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1510 Pass Pmode to gen_tls_dynamic_gnu2_64.
1512 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1514 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1516 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1518 * config/aarch64/aarch64-sve-builtins-base.cc
1519 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1521 2020-01-20 Richard Biener <rguenther@suse.de>
1524 * dwarf2out.c (prune_unused_types): Unconditionally mark
1525 called function DIEs.
1527 2020-01-20 Martin Liska <mliska@suse.cz>
1529 PR tree-optimization/93199
1530 * tree-eh.c (struct leh_state): Add
1531 new field outer_non_cleanup.
1532 (cleanup_is_dead_in): Pass leh_state instead
1533 of eh_region. Add a checking that state->outer_non_cleanup
1534 points to outer non-clean up region.
1535 (lower_try_finally): Record outer_non_cleanup
1537 (lower_catch): Likewise.
1538 (lower_eh_filter): Likewise.
1539 (lower_eh_must_not_throw): Likewise.
1540 (lower_cleanup): Likewise.
1542 2020-01-20 Richard Biener <rguenther@suse.de>
1544 PR tree-optimization/93094
1545 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1546 (vect_transform_loop): Likewise.
1547 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1548 loop_vectorized_call to vect_transform_loop.
1549 * tree-vect-loop.c (vect_transform_loop): Pass down
1550 loop_vectorized_call to vect_loop_versioning.
1551 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1552 the earlier discovered loop_vectorized_call.
1554 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1556 * doc/contribute.texi: Update for SVN -> Git transition.
1557 * doc/install.texi: Likewise.
1559 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1562 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1563 speculative targets.
1564 (verify_speculative_call): New function
1565 (cgraph_node::verify_node): Use it.
1566 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1569 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1572 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1573 (cgraph_edge::make_direct): Remove all indirect targets.
1574 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1575 (cgraph_node::verify_node): Verify that only one call_stmt or
1576 lto_stmt_uid is set.
1577 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1579 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1580 (lto_output_ref): Simplify streaming of stmt.
1581 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1583 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1585 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1586 Mark parameter unused.
1588 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1590 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1592 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1594 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1596 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1598 * Makefile.in: Add coroutine-passes.o.
1599 * builtin-types.def (BT_CONST_SIZE): New.
1600 (BT_FN_BOOL_PTR): New.
1601 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1602 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1603 * coroutine-builtins.def: New file.
1604 * coroutine-passes.cc: New file.
1605 * function.h (struct GTY function): Add a bit to indicate that the
1606 function is a coroutine component.
1607 * internal-fn.c (expand_CO_FRAME): New.
1608 (expand_CO_YIELD): New.
1609 (expand_CO_SUSPN): New.
1610 (expand_CO_ACTOR): New.
1611 * internal-fn.def (CO_ACTOR): New.
1615 * passes.def: Add pass_coroutine_lower_builtins,
1616 pass_coroutine_early_expand_ifns.
1617 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1618 (make_pass_coroutine_early_expand_ifns): New.
1619 * doc/invoke.texi: Document the fcoroutines command line
1622 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1624 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1627 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1628 after checking the argument is a REG. Don't use REGNO (reg)
1629 again to set last_regno, reuse regno variable instead.
1631 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1633 * doc/analyzer.texi (Limitations): Add note about NaN.
1635 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1636 Sudakshina Das <sudi.das@arm.com>
1638 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1639 and valid immediate.
1640 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1641 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1642 * config/arm/constraints.md (Pg): New.
1643 * config/arm/predicates.md (long_shift_imm): New.
1644 (arm_reg_or_long_shift_imm): Likewise.
1645 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1646 (thumb2_lsll): Likewise.
1649 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1650 Sudakshina Das <sudi.das@arm.com>
1652 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1653 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1654 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1655 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1656 * config/arm/thumb2.md (thumb2_asrl): New.
1657 (thumb2_lsll): Likewise.
1659 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1661 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1664 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1666 * gdbinit.in (help-gcc-hooks): New command.
1667 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1668 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1671 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1673 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1674 correct target macro.
1676 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1678 * config/aarch64/aarch64-protos.h
1679 (aarch64_sve_ld1ro_operand_p): New.
1680 * config/aarch64/aarch64-sve-builtins-base.cc
1681 (class load_replicate): New.
1682 (class svld1ro_impl): New.
1683 (class svld1rq_impl): Change to inherit from load_replicate.
1684 (svld1ro): New sve intrinsic function base.
1685 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1686 New DEF_SVE_FUNCTION.
1687 * config/aarch64/aarch64-sve-builtins-base.h
1688 (svld1ro): New decl.
1689 * config/aarch64/aarch64-sve-builtins.cc
1690 (function_expander::add_mem_operand): Modify assert to allow
1692 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1694 * config/aarch64/aarch64.c
1695 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1696 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1697 (aarch64_sve_ld1ro_operand_p): New.
1698 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1699 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1700 * config/aarch64/predicates.md
1701 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1703 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1705 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1706 Introduce this ACLE specified predefined macro.
1707 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1708 (fp): Disabling this disables f64mm.
1709 (simd): Disabling this disables f64mm.
1710 (fp16): Disabling this disables f64mm.
1711 (sve): Disabling this disables f64mm.
1712 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1713 (AARCH64_ISA_F64MM): New.
1714 (TARGET_F64MM): New.
1715 * doc/invoke.texi (f64mm): Document new option.
1717 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1719 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1720 (neoversen1_tunings): Likewise.
1722 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1725 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1726 Add assert to ensure prolog has been emitted.
1727 (aarch64_split_atomic_op): Likewise.
1728 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1729 Use epilogue_completed rather than reload_completed.
1730 (aarch64_atomic_exchange<mode>): Likewise.
1731 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1732 (atomic_nand<mode>): Likewise.
1733 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1734 (atomic_fetch_nand<mode>): Likewise.
1735 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1736 (atomic_nand_fetch<mode>): Likewise.
1738 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1741 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1743 (REVERSE_CONDITION): Delete.
1744 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1745 (CCFP_CCFPE): Likewise.
1746 (e): New mode attribute.
1747 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1748 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1749 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1750 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1751 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1752 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1753 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1754 name of generator from gen_ccmpdi to gen_ccmpccdi.
1755 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1756 the previous comparison but aren't able to, use the new ccmp_rev
1759 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1761 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1762 than testing directly for INTEGER_CST.
1763 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1765 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1767 PR tree-optimization/93292
1768 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1769 get_vectype_for_scalar_type returns NULL.
1771 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1773 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1774 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1776 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1778 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1780 * params.opt: (max-predicted-iterations): Set bounds.
1781 * predict.c (real_almost_one, real_br_prob_base,
1782 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1783 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1784 probabilities; do not truncate to reg_br_prob_bases.
1785 (estimate_loops_at_level): Pass max_cyclic_prob.
1786 (estimate_loops): Compute max_cyclic_prob.
1787 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1789 * profile-count.c (profile_probability::to_sreal): New.
1790 * profile-count.h (class sreal): Move up in file.
1791 (profile_probability::to_sreal): Declare.
1793 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1796 (arm_invalid_conversion): New function for target hook.
1797 (arm_invalid_unary_op): New function for target hook.
1798 (arm_invalid_binary_op): New function for target hook.
1800 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1802 * config.gcc: Add arm_bf16.h.
1803 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1804 (arm_simd_builtin_std_type): Add BFmode.
1805 (arm_init_simd_builtin_types): Define element types for vector types.
1806 (arm_init_bf16_types): New function.
1807 (arm_init_builtins): Add arm_init_bf16_types function call.
1808 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1809 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1810 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1811 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1812 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1813 (arm_mangle_type): Add __bf16.
1814 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1815 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1816 arm_bf16_ptr_type_node.
1817 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1818 define_split between ARM registers.
1819 * config/arm/arm_bf16.h: New file.
1820 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1821 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1823 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1824 * config/arm/vfp.md: Add BFmode to movhf patterns.
1826 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1827 Andre Vieira <andre.simoesdiasvieira@arm.com>
1829 * config/arm/arm-cpus.in (mve, mve_float): New features.
1830 (dsp, mve, mve.fp): New options.
1831 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1832 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1833 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1835 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1836 Thomas Preud'homme <thomas.preudhomme@arm.com>
1838 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1840 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1841 error for using -mcmse when targeting Armv8.1-M Mainline.
1843 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1844 Thomas Preud'homme <thomas.preudhomme@arm.com>
1846 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1847 address in r4 when targeting Armv8.1-M Mainline.
1848 (nonsecure_call_value_internal): Likewise.
1849 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1850 a register match_operand again. Emit BLXNS when targeting
1852 (nonsecure_call_value_reg_thumb2): Likewise.
1854 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1855 Thomas Preud'homme <thomas.preudhomme@arm.com>
1857 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1858 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1859 variable as true when floating-point ABI is not hard. Replace
1860 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1861 Generate VLSTM and VLLDM instruction respectively before and
1862 after a function call to cmse_nonsecure_call function.
1863 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1864 (VUNSPEC_VLLDM): Likewise.
1865 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1866 (lazy_load_multiple_insn): Likewise.
1868 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1869 Thomas Preud'homme <thomas.preudhomme@arm.com>
1871 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1872 (arm_emit_vfp_multi_reg_pop): Likewise.
1873 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1874 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1875 restore callee-saved VFP registers.
1877 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1878 Thomas Preud'homme <thomas.preudhomme@arm.com>
1880 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1881 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1882 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1883 callee-saved GPRs as well as clear ip register before doing a nonsecure
1884 call then restore callee-saved GPRs after it when targeting
1886 (arm_reorg): Adapt to function rename.
1888 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1889 Thomas Preud'homme <thomas.preudhomme@arm.com>
1891 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1892 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1893 clear_vfp_multiple pattern based on a new vfp parameter.
1894 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1895 targeting Armv8.1-M Mainline.
1896 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1897 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1898 whether VFP registers are available before looking call_used_regs for a
1900 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1901 of prototype of clear_operation_p.
1902 (clear_vfp_multiple_operation): New predicate.
1903 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1904 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1906 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1907 Thomas Preud'homme <thomas.preudhomme@arm.com>
1909 * config/arm/arm-protos.h (clear_operation_p): Declare.
1910 * config/arm/arm.c (clear_operation_p): New function.
1911 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1912 targeting Armv8.1-M Mainline or successor.
1913 (output_return_instruction): Only output APSR register clearing if
1914 Armv8.1-M Mainline instructions not available.
1915 (thumb_exit): Likewise.
1916 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1917 * config/arm/thumb2.md (clear_apsr): New define_insn.
1918 (clear_multiple): Likewise.
1919 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1921 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1922 Thomas Preud'homme <thomas.preudhomme@arm.com>
1924 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1925 (use_return_insn): Also return false for Armv8.1-M Mainline.
1926 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1927 Mainline instructions are available.
1928 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1929 when targeting Armv8.1-M Mainline Security Extensions.
1930 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1931 Mainline entry function.
1932 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1933 targeting Armv8.1-M Mainline or successor.
1934 (arm_expand_epilogue): Fix indentation of caller-saved register
1935 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1937 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1938 (FP_SYSREGS): Likewise.
1939 (enum vfp_sysregs_encoding): Define enum.
1940 (fp_sysreg_names): Declare.
1941 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1942 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1943 (pop_fpsysreg_insn): Likewise.
1945 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1946 Thomas Preud'homme <thomas.preudhomme@arm.com>
1948 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1949 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1950 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1951 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1952 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1953 (ARMv8_1m_main): New feature group.
1954 (armv8.1-m.main): New architecture.
1955 * config/arm/arm-tables.opt: Regenerate.
1956 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1957 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1958 (arm_options_perform_arch_sanity_checks): Error out when targeting
1959 Armv8.1-M Mainline Security Extensions.
1960 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1962 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1964 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1965 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1966 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1967 aarch64_bfdot_laneq): New.
1968 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1969 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1970 vbfdotq_laneq_f32): New.
1971 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1972 VBFMLA_W, VBF): New.
1973 (isquadop): Add V4BF, V8BF.
1975 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1977 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1978 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1979 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1980 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1981 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1982 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1983 usdot_laneq, sudot_lane,sudot_laneq): New.
1984 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1985 (aarch64_<sur>dot_lane): New.
1986 * config/aarch64/arm_neon.h (vusdot_s32): New.
1988 (vusdot_lane_s32): New.
1989 (vsudot_lane_s32): New.
1990 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1991 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1993 2020-01-16 Martin Liska <mliska@suse.cz>
1995 * value-prof.c (dump_histogram_value): Fix
1996 obvious spacing issue.
1998 2020-01-16 Andrew Pinski <apinski@marvell.com>
2000 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2001 !storage_order_barrier_p.
2003 2020-01-16 Andrew Pinski <apinski@marvell.com>
2005 * sched-int.h (_dep): Add unused bit-field field for the padding.
2006 * sched-deps.c (init_dep_1): Init unused field.
2008 2020-01-16 Andrew Pinski <apinski@marvell.com>
2010 * optabs.h (create_expand_operand): Initialize target field also.
2012 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2014 PR tree-optimization/92429
2015 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2016 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2018 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2021 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2023 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2024 aarch64_sve_int_mode to each mode.
2026 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2028 * doc/analyzer.texi (Overview): Add note about
2029 -fdump-ipa-analyzer.
2031 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2033 PR tree-optimization/93231
2034 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2035 input_type is unsigned. Use tree_to_shwi for shift constant.
2036 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2037 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2038 inputs more efficiently.
2040 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2042 * config/i386/i386.md (*movsf_internal): Do not require
2043 SSE2 ISA for alternatives 14 and 15.
2045 2020-01-15 Richard Biener <rguenther@suse.de>
2048 * tree-eh.c (sink_clobbers): If we already visited the destination
2049 block do not defer insertion.
2050 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2051 the purpose of defered insertion.
2053 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2055 * BASE-VER: Bump to 10.0.1.
2057 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2059 PR tree-optimization/93247
2060 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2061 type of the stmt that we're going to vectorize.
2063 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2065 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2066 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2069 2020-01-15 Martin Liska <mliska@suse.cz>
2071 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2072 2 calls of streamer_read_hwi in a function call.
2074 2020-01-15 Richard Biener <rguenther@suse.de>
2076 * alias.c (record_alias_subset): Avoid redundant work when
2077 subset is already recorded.
2079 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2081 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2082 the analyzer options provide CWE identifiers.
2084 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2086 * tree-diagnostic-path.cc (path_summary::event_range::print):
2087 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2088 using get_pure_location.
2090 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2092 PR tree-optimization/93262
2093 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2094 perform head trimming only if the last argument is constant,
2095 either all ones, or larger or equal to head trim, in the latter
2096 case decrease the last argument by head_trim.
2098 PR tree-optimization/93249
2099 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2100 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2101 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2102 perform head trim unless we can prove there are no '\0' chars
2103 from the source among the first head_trim chars.
2105 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2107 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2109 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2112 * config/i386/sse.md
2113 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2114 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2115 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2116 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2117 just a single alternative instead of two, make operands 1 and 2
2120 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2123 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2126 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2128 * Makefile.in (lang_opt_files): Add analyzer.opt.
2129 (ANALYZER_OBJS): New.
2130 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2131 tristate.o and ANALYZER_OBJS.
2132 (TEXI_GCCINT_FILES): Add analyzer.texi.
2133 * common.opt (-fanalyzer): New driver option.
2134 * config.in: Regenerate.
2135 * configure: Regenerate.
2136 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2137 (gccdepdir): Also create depdir for "analyzer" subdir.
2138 * digraph.cc: New file.
2139 * digraph.h: New file.
2140 * doc/analyzer.texi: New file.
2141 * doc/gccint.texi ("Static Analyzer") New menu item.
2142 (analyzer.texi): Include it.
2143 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2144 ("Warning Options"): Add static analysis warnings to the list.
2145 (-Wno-analyzer-double-fclose): New option.
2146 (-Wno-analyzer-double-free): New option.
2147 (-Wno-analyzer-exposure-through-output-file): New option.
2148 (-Wno-analyzer-file-leak): New option.
2149 (-Wno-analyzer-free-of-non-heap): New option.
2150 (-Wno-analyzer-malloc-leak): New option.
2151 (-Wno-analyzer-possible-null-argument): New option.
2152 (-Wno-analyzer-possible-null-dereference): New option.
2153 (-Wno-analyzer-null-argument): New option.
2154 (-Wno-analyzer-null-dereference): New option.
2155 (-Wno-analyzer-stale-setjmp-buffer): New option.
2156 (-Wno-analyzer-tainted-array-index): New option.
2157 (-Wno-analyzer-use-after-free): New option.
2158 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2159 (-Wno-analyzer-use-of-uninitialized-value): New option.
2160 (-Wanalyzer-too-complex): New option.
2161 (-fanalyzer-call-summaries): New warning.
2162 (-fanalyzer-checker=): New warning.
2163 (-fanalyzer-fine-grained): New warning.
2164 (-fno-analyzer-state-merge): New warning.
2165 (-fno-analyzer-state-purge): New warning.
2166 (-fanalyzer-transitivity): New warning.
2167 (-fanalyzer-verbose-edges): New warning.
2168 (-fanalyzer-verbose-state-changes): New warning.
2169 (-fanalyzer-verbosity=): New warning.
2170 (-fdump-analyzer): New warning.
2171 (-fdump-analyzer-callgraph): New warning.
2172 (-fdump-analyzer-exploded-graph): New warning.
2173 (-fdump-analyzer-exploded-nodes): New warning.
2174 (-fdump-analyzer-exploded-nodes-2): New warning.
2175 (-fdump-analyzer-exploded-nodes-3): New warning.
2176 (-fdump-analyzer-supergraph): New warning.
2177 * doc/sourcebuild.texi (dg-require-dot): New.
2178 (dg-check-dot): New.
2179 * gdbinit.in (break-on-saved-diagnostic): New command.
2180 * graphviz.cc: New file.
2181 * graphviz.h: New file.
2182 * ordered-hash-map-tests.cc: New file.
2183 * ordered-hash-map.h: New file.
2184 * passes.def (pass_analyzer): Add before
2185 pass_ipa_whole_program_visibility.
2186 * selftest-run-tests.c (selftest::run_tests): Call
2187 selftest::ordered_hash_map_tests_cc_tests.
2188 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2190 * shortest-paths.h: New file.
2191 * timevar.def (TV_ANALYZER): New timevar.
2192 (TV_ANALYZER_SUPERGRAPH): Likewise.
2193 (TV_ANALYZER_STATE_PURGE): Likewise.
2194 (TV_ANALYZER_PLAN): Likewise.
2195 (TV_ANALYZER_SCC): Likewise.
2196 (TV_ANALYZER_WORKLIST): Likewise.
2197 (TV_ANALYZER_DUMP): Likewise.
2198 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2199 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2200 * tree-pass.h (make_pass_analyzer): New decl.
2201 * tristate.cc: New file.
2202 * tristate.h: New file.
2204 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2207 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2208 alternatives 9 and 10.
2210 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2212 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2213 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2214 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2215 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2216 (selftest::hash_map_tests_c_tests): Call it.
2217 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2218 New static constant, using the value of = H::empty_zero_p.
2219 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2220 from default_hash_traits <Value>.
2221 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2223 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2224 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2225 calls to mark_empty with !Descriptor::empty_zero_p.
2226 (hash_table::empty_slow): Conditionalize the memset call with a
2227 check that Descriptor::empty_zero_p; otherwise, loop through the
2228 entries calling mark_empty on them.
2229 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2230 (pointer_hash::empty_zero_p): Likewise.
2231 (pair_hash::empty_zero_p): Likewise.
2232 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2234 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2235 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2236 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2237 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2238 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2239 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2240 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2241 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2243 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2246 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2248 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2249 fix typo on return value.
2251 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2254 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2256 (cgraph_edge::make_speculative): Add param for setting speculative_id
2258 (cgraph_edge::speculative_call_info): Update comments and find reference
2259 by speculative_id for multiple indirect targets.
2260 (cgraph_edge::resolve_speculation): Decrease the speculations
2261 for indirect edge, drop it's speculative if not direct target
2262 left. Update comments.
2263 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2264 (cgraph_node::dump): Print num_speculative_call_targets.
2265 (cgraph_node::verify_node): Don't report error if speculative
2266 edge not include statement.
2267 (cgraph_edge::num_speculative_call_targets_p): New function.
2268 * cgraph.h (int common_target_id): Remove.
2269 (int common_target_probability): Remove.
2270 (num_speculative_call_targets): New variable.
2271 (make_speculative): Add param for setting speculative_id.
2272 (cgraph_edge::num_speculative_call_targets_p): New declare.
2273 (target_prob): New variable.
2274 (speculative_id): New variable.
2275 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2276 call summaries for multiple speculative call targets.
2277 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2278 * ipa-profile.c (struct speculative_call_target): New struct.
2279 (class speculative_call_summary): New class.
2280 (class speculative_call_summaries): New class.
2281 (call_sums): New variable.
2282 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2283 (ipa_profile_write_edge_summary): New function.
2284 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2285 (ipa_profile_dump_all_summaries): New function.
2286 (ipa_profile_read_edge_summary): New function.
2287 (ipa_profile_read_summary_section): New function.
2288 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2289 (ipa_profile): Generate num_speculative_call_targets from
2291 * ipa-ref.h (speculative_id): New variable.
2292 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2293 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2294 common_target_probability. Stream out speculative_id and
2295 num_speculative_call_targets.
2296 (input_edge): Likewise.
2297 * predict.c (dump_prediction): Remove edges count assert to be
2299 * symtab.c (symtab_node::create_reference): Init speculative_id.
2300 (symtab_node::clone_references): Clone speculative_id.
2301 (symtab_node::clone_referring): Clone speculative_id.
2302 (symtab_node::clone_reference): Clone speculative_id.
2303 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2304 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2305 if indirect call contains multiple speculative targets.
2306 * value-prof.h (check_ic_target): Remove.
2307 * value-prof.c (gimple_value_profile_transformations):
2308 Use void function gimple_ic_transform.
2309 * value-prof.c (gimple_ic_transform): Handle topn case.
2310 Fix comment typos. Change it to a void function.
2312 2020-01-13 Andrew Pinski <apinski@marvell.com>
2314 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2315 (octeontx2t98): New define.
2316 (octeontx2t96): New define.
2317 (octeontx2t93): New define.
2318 (octeontx2f95): New define.
2319 (octeontx2f95n): New define.
2320 (octeontx2f95mm): New define.
2321 * config/aarch64/aarch64-tune.md: Regenerate.
2322 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2324 2020-01-13 Jason Merrill <jason@redhat.com>
2326 PR c++/33799 - destroy return value if local cleanup throws.
2327 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2329 2020-01-13 Martin Liska <mliska@suse.cz>
2331 * ipa-cp.c (get_max_overall_size): Use newly
2332 renamed param param_ipa_cp_unit_growth.
2333 * params.opt: Remove legacy param name.
2335 2020-01-13 Martin Sebor <msebor@redhat.com>
2337 PR tree-optimization/93213
2338 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2339 stores to be eliminated.
2341 2020-01-13 Martin Liska <mliska@suse.cz>
2343 * opts.c (print_help): Do not print CL_PARAM
2344 and CL_WARNING for CL_OPTIMIZATION.
2346 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2349 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2350 depending on optimization settings.
2352 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2354 PR tree-optimization/90838
2355 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2356 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2357 argument rather than to initialize temporary for targets that
2358 don't use the mode argument at all. Initialize ctzval to avoid
2361 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2363 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2364 * tree-core.h: Document it.
2365 * gimplify.c (gimplify_omp_workshare): Set it.
2366 * omp-low.c (lower_omp_target): Use it.
2367 * tree-pretty-print.c (dump_omp_clause): Print it.
2369 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2370 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2372 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2374 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2375 * common.opt (fdiagnostics-path-format=): New option.
2376 (diagnostic_path_format): New enum.
2377 (fdiagnostics-show-path-depths): New option.
2378 * coretypes.h (diagnostic_event_id_t): New forward decl.
2379 * diagnostic-color.c (color_dict): Add "path".
2380 * diagnostic-event-id.h: New file.
2381 * diagnostic-format-json.cc (json_from_expanded_location): Make
2383 (json_end_diagnostic): Call context->make_json_for_path if it
2384 exists and the diagnostic has a path.
2385 (diagnostic_output_format_init): Clear context->print_path.
2386 * diagnostic-path.h: New file.
2387 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2388 when printing a run of events in a diagnostic_path so that they
2389 all get the same color.
2390 (layout::m_diagnostic_path_p): New field.
2391 (layout::layout): Initialize it.
2392 (layout::print_any_labels): Don't colorize the label text for an
2393 event in a diagnostic_path.
2394 (gcc_rich_location::add_location_if_nearby): Add
2395 "restrict_to_current_line_spans" and "label" params. Pass the
2396 former to layout.maybe_add_location_range; pass the latter
2397 when calling add_range.
2398 * diagnostic.c: Include "diagnostic-path.h".
2399 (diagnostic_initialize): Initialize context->path_format and
2400 context->show_path_depths.
2401 (diagnostic_show_any_path): New function.
2402 (diagnostic_path::interprocedural_p): New function.
2403 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2404 (simple_diagnostic_path::num_events): New function.
2405 (simple_diagnostic_path::get_event): New function.
2406 (simple_diagnostic_path::add_event): New function.
2407 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2408 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2409 (debug): New overload taking a diagnostic_path *.
2410 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2411 * diagnostic.h (enum diagnostic_path_format): New enum.
2412 (json::value): New forward decl.
2413 (diagnostic_context::path_format): New field.
2414 (diagnostic_context::show_path_depths): New field.
2415 (diagnostic_context::print_path): New callback field.
2416 (diagnostic_context::make_json_for_path): New callback field.
2417 (diagnostic_show_any_path): New decl.
2418 (json_from_expanded_location): New decl.
2419 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2420 (-fdiagnostics-show-path-depths): New option.
2421 (-fdiagnostics-color): Add "path" to description of default
2422 GCC_COLORS; describe it.
2423 (-fdiagnostics-format=json): Document how diagnostic paths are
2424 represented in the JSON output format.
2425 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2426 Add optional params "restrict_to_current_line_spans" and "label".
2427 * opts.c (common_handle_option): Handle
2428 OPT_fdiagnostics_path_format_ and
2429 OPT_fdiagnostics_show_path_depths.
2430 * pretty-print.c: Include "diagnostic-event-id.h".
2431 (pp_format): Implement "%@" format code for printing
2432 diagnostic_event_id_t *.
2433 (selftest::test_pp_format): Add tests for "%@".
2434 * selftest-run-tests.c (selftest::run_tests): Call
2435 selftest::tree_diagnostic_path_cc_tests.
2436 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2437 * toplev.c (general_init): Initialize global_dc->path_format and
2438 global_dc->show_path_depths.
2439 * tree-diagnostic-path.cc: New file.
2440 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2441 non-static. Drop "diagnostic" param in favor of storing the
2442 original value of "where" and re-using it.
2443 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2444 maybe_unwind_expanded_macro_loc.
2445 (tree_diagnostics_defaults): Initialize context->print_path and
2446 context->make_json_for_path.
2447 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2449 (default_tree_make_json_for_path): New decl.
2450 (maybe_unwind_expanded_macro_loc): New decl.
2452 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2454 PR tree-optimization/93210
2455 * fold-const.h (native_encode_initializer,
2456 can_native_interpret_type_p): Declare.
2457 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2459 (native_encode_initializer): New function, moved from dwarf2out.c.
2460 Adjust to native_encode_expr compatible arguments, including dry-run
2461 and partial extraction modes. Don't handle STRING_CST.
2462 (can_native_interpret_type_p): No longer static.
2463 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2464 offset / BITS_PER_UNIT fits into int and don't call it if
2465 can_native_interpret_type_p fails. If suboff is NULL and for
2466 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2467 native_encode_initializer.
2468 (fold_const_aggregate_ref_1): Formatting fix.
2469 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2470 (tree_add_const_value_attribute): Adjust caller.
2472 PR tree-optimization/90838
2473 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2474 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2475 CTZ_DEFINED_VALUE_AT_ZERO.
2477 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2480 * lra-constraints.c (match_reload): Permit input operands have the
2481 same mode as output while other input operands have a different
2484 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2486 PR tree-optimization/90838
2487 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2488 (check_ctz_string): Likewise.
2489 (optimize_count_trailing_zeroes): Likewise.
2490 (simplify_count_trailing_zeroes): Likewise.
2491 (pass_forwprop::execute): Try ctz simplification.
2492 * match.pd: Add matching for ctz idioms.
2494 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2496 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2498 (aarch64_invalid_unary_op): New function for target hook.
2499 (aarch64_invalid_binary_op): New function for target hook.
2501 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2503 * config.gcc: Add arm_bf16.h.
2504 * config/aarch64/aarch64-builtins.c
2505 (aarch64_simd_builtin_std_type): Add BFmode.
2506 (aarch64_init_simd_builtin_types): Define element types for vector
2508 (aarch64_init_bf16_types): New function.
2509 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2510 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2512 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2513 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2515 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2516 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2517 * config/aarch64/aarch64.c
2518 (aarch64_classify_vector_mode): Add support for BF types.
2519 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2520 (aarch64_vq_mode): Add support for BF types.
2521 (aarch64_simd_container_mode): Add support for BF types.
2522 (aarch64_mangle_type): Add support for BF scalar type.
2523 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2524 * config/aarch64/arm_bf16.h: New file.
2525 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2526 * config/aarch64/iterators.md: Add BF types to mode attributes.
2527 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2529 2020-01-10 Jason Merrill <jason@redhat.com>
2531 PR c++/93173 - incorrect tree sharing.
2532 * gimplify.c (copy_if_shared): No longer static.
2533 * gimplify.h: Declare it.
2535 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2537 * doc/invoke.texi (-msve-vector-bits=): Document that
2538 -msve-vector-bits=128 now generates VL-specific code for
2539 little-endian targets.
2540 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2541 build_vector_type_for_mode to construct the data vector types.
2542 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2543 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2544 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2545 for 128-bit vectors.
2547 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2549 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2552 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2554 * config/aarch64/aarch64-builtins.c
2555 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2556 rather than checking the number of elements and the element mode.
2558 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2560 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2561 get_related_vectype_for_scalar_type rather than build_vector_type
2562 to create the index type for a conditional reduction.
2564 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2566 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2567 for any type of gather or scatter, including strided accesses.
2569 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2571 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2574 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2576 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2578 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2579 parameter and its use to reset DR_OFFSET's.
2580 (vect_transform_loop): Remove orig_drs_init argument.
2581 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2582 member of dr_vec_info rather than the offset of the associated
2583 data_reference's innermost_loop_behavior.
2584 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2585 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2586 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2587 get_dr_vinfo_offset.
2588 (vectorizable_store): Likewise.
2589 (vectorizable_load): Likewise.
2591 2020-01-10 Richard Biener <rguenther@suse.de>
2593 * gimple-ssa-store-merging
2594 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2596 2020-01-10 Martin Liska <mliska@suse.cz>
2599 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2600 encapsulation that was there before r280040.
2602 2020-01-10 Richard Biener <rguenther@suse.de>
2605 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2606 sequences to avoid walking them again for secondary opportunities.
2607 (pass_lower_eh_dispatch::execute): Instead actually insert
2610 2020-01-10 Richard Biener <rguenther@suse.de>
2613 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2614 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2615 avoid quadraticness.
2617 2020-01-10 Martin Jambor <mjambor@suse.cz>
2619 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2620 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2621 to get param_ipa_sra_max_replacements.
2622 (param_splitting_across_edge): Pass the caller to
2623 pull_accesses_from_callee.
2625 2020-01-10 Martin Jambor <mjambor@suse.cz>
2627 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2628 * ipa-cp.c (max_new_size): Removed.
2629 (orig_overall_size): New variable.
2630 (get_max_overall_size): New function.
2631 (estimate_local_effects): Use it. Adjust dump.
2632 (decide_about_value): Likewise.
2633 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2634 orig_overall_size. Adjust dump.
2635 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2637 2020-01-10 Martin Jambor <mjambor@suse.cz>
2639 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2640 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2641 instead of param_ipa_max_agg_items.
2642 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2643 optimization info for the callee.
2645 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2647 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2648 markers if debug_inline_points is false.
2650 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2652 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2654 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2655 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2656 aarch64-sve-builtins-sve2.h.
2657 (aarch64-sve-builtins-sve2.o): New rule.
2658 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2659 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2660 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2661 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2662 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2663 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2665 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2666 instructions that are handled here.
2667 (@cond_asrd<mode>): Generalize to...
2668 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2669 (*cond_asrd<mode>_2): Generalize to...
2670 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2671 (*cond_asrd<mode>_z): Generalize to...
2672 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2673 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2674 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2675 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2676 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2678 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2679 (@aarch64_scatter_stnt<mode>): Likewise.
2680 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2681 (@aarch64_mul_lane_<mode>): Likewise.
2682 (@aarch64_sve_suqadd<mode>_const): Likewise.
2683 (*<sur>h<addsub><mode>): Generalize to...
2684 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2686 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2687 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2688 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2689 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2690 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2691 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2692 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2693 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2694 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2695 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2696 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2697 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2698 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2699 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2700 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2701 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2702 (@aarch64_sve2_xar<mode>): Likewise.
2703 (@aarch64_sve2_bcax<mode>): Likewise.
2704 (*aarch64_sve2_eor3<mode>): Rename to...
2705 (@aarch64_sve2_eor3<mode>): ...this.
2706 (@aarch64_sve2_bsl<mode>): New expander.
2707 (@aarch64_sve2_nbsl<mode>): Likewise.
2708 (@aarch64_sve2_bsl1n<mode>): Likewise.
2709 (@aarch64_sve2_bsl2n<mode>): Likewise.
2710 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2711 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2712 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2713 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2714 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2715 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2716 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2717 (<su>mull<bt><Vwide>): Generalize to...
2718 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2720 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2721 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2722 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2723 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2724 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2725 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2726 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2727 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2728 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2729 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2730 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2731 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2732 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2733 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2734 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2735 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2736 (<SHRNB:r>shrnb<mode>): Generalize to...
2737 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2739 (<SHRNT:r>shrnt<mode>): Generalize to...
2740 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2742 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2743 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2744 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2745 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2746 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2747 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2748 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2749 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2750 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2751 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2752 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2753 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2754 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2755 (@aarch64_sve2_cvtnt<mode>): Likewise.
2756 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2757 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2758 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2759 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2760 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2761 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2762 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2763 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2764 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2765 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2766 (@aarch64_sve2_pmul<mode>): Likewise.
2767 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2768 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2769 (@aarch64_sve2_tbl2<mode>): Likewise.
2770 (@aarch64_sve2_tbx<mode>): Likewise.
2771 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2772 (@aarch64_sve2_histcnt<mode>): Likewise.
2773 (@aarch64_sve2_histseg<mode>): Likewise.
2774 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2775 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2776 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2777 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2778 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2779 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2780 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2781 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2782 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2783 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2784 (SVE2_PMULL_PAIR_I): New mode iterators.
2785 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2786 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2787 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2788 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2789 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2790 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2791 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2792 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2793 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2794 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2795 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2796 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2797 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2798 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2799 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2800 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2801 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2802 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2803 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2804 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2805 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2806 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2807 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2808 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2809 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2810 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2811 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2812 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2813 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2814 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2815 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2817 (VNARROW, Ventype): New mode attributes.
2818 (Vewtype): Handle VNx2DI. Fix typo in comment.
2819 (VDOUBLE): New mode attribute.
2820 (sve_lane_con): Handle VNx8HI.
2821 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2822 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2823 (sve_int_op, sve_int_op_rev): Handle the above codes.
2824 (sve_pred_int_rhs2_operand): Likewise.
2825 (MULLBT, SHRNB, SHRNT): Delete.
2826 (SVE_INT_SHIFT_IMM): New int iterator.
2827 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2828 and UNSPEC_WHILEHS for TARGET_SVE2.
2829 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2830 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2831 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2832 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2833 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2834 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2835 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2836 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2837 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2838 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2839 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2840 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2841 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2842 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2843 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2844 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2845 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2846 (optab): Handle the new unspecs.
2847 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2849 (lr): Handle the new unspecs.
2851 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2852 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2853 (sve_int_qsub_op): New int attributes.
2854 (sve_fp_op, rot): Handle the new unspecs.
2855 * config/aarch64/aarch64-sve-builtins.h
2856 (function_resolver::require_matching_pointer_type): Declare.
2857 (function_resolver::resolve_unary): Add an optional boolean argument.
2858 (function_resolver::finish_opt_n_resolution): Add an optional
2859 type_suffix_index argument.
2860 (gimple_folder::redirect_call): Declare.
2861 (gimple_expander::prepare_gather_address_operands): Add an optional
2863 * config/aarch64/aarch64-sve-builtins.cc: Include
2864 aarch64-sve-builtins-sve2.h.
2865 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2866 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2867 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2868 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2869 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2870 (TYPES_s_unsigned): Likewise.
2871 (TYPES_s_integer): Use TYPES_s_unsigned.
2872 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2873 (TYPES_sd_integer): Use them.
2874 (TYPES_d_unsigned): New macro.
2875 (TYPES_d_integer): Use it.
2876 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2877 (TYPES_cvt_narrow): Likewise.
2878 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2879 (preds_mx): New variable.
2880 (function_builder::add_overloaded_function): Allow the new feature
2881 set to be more restrictive than the original one.
2882 (function_resolver::infer_pointer_type): Remove qualifiers from
2883 the pointer type before printing it.
2884 (function_resolver::require_matching_pointer_type): New function.
2885 (function_resolver::resolve_sv_displacement): Handle functions
2886 that don't support 32-bit vector indices or svint32_t vector offsets.
2887 (function_resolver::finish_opt_n_resolution): Take the inferred type
2888 as a separate argument.
2889 (function_resolver::resolve_unary): Optionally treat all forms in
2890 the same way as normal merging functions.
2891 (gimple_folder::redirect_call): New function.
2892 (function_expander::prepare_gather_address_operands): Add an argument
2893 that says whether scaled forms are available. If they aren't,
2894 handle scaling of vector indices and don't add the extension and
2896 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2897 fall back to using cond_* instead.
2898 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2899 Split out the member variables into...
2900 (rtx_code_function_base): ...this new base class.
2901 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2902 (unspec_based_function): Split out the member variables into...
2903 (unspec_based_function_base): ...this new base class.
2904 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2905 (unspec_based_function_exact_insn): New class.
2906 (unspec_based_add_function, unspec_based_add_lane_function)
2907 (unspec_based_lane_function, unspec_based_pred_function)
2908 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2909 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2910 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2912 (unspec_based_fused_function): New class.
2913 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2914 (unspec_based_fused_lane_function): New class.
2915 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2917 (CODE_FOR_MODE1): New macro.
2918 (fixed_insn_function): New class.
2919 (while_comparison): Likewise.
2920 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2921 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2922 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2923 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2924 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2925 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2926 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2927 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2928 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2929 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2930 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2931 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2932 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2933 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2934 Also add an initial argument for unary_convert_narrowt, regardless
2935 of the predication type.
2936 (build_32_64): Allow loads and stores to specify MODE_none.
2937 (build_sv_index64, build_sv_uint_offset): New functions.
2938 (long_type_suffix): New function.
2939 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2940 (binary_imm_long_base, load_gather_sv_base): Likewise.
2941 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2942 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2943 (unary_narrowb_base, unary_narrowt_base): Likewise.
2944 (binary_long_lane_def, binary_long_lane): New shape.
2945 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2946 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2947 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2948 (binary_to_uint_def, binary_to_uint): Likewise.
2949 (binary_wide_def, binary_wide): Likewise.
2950 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2951 (compare_def, compare): Likewise.
2952 (compare_ptr_def, compare_ptr): Likewise.
2953 (load_ext_gather_index_restricted_def,
2954 load_ext_gather_index_restricted): Likewise.
2955 (load_ext_gather_offset_restricted_def,
2956 load_ext_gather_offset_restricted): Likewise.
2957 (load_gather_sv_def): Inherit from load_gather_sv_base.
2958 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2959 (shift_left_imm_def, shift_left_imm): Likewise.
2960 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2961 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2962 (store_scatter_index_restricted_def,
2963 store_scatter_index_restricted): Likewise.
2964 (store_scatter_offset_restricted_def,
2965 store_scatter_offset_restricted): Likewise.
2966 (tbl_tuple_def, tbl_tuple): Likewise.
2967 (ternary_long_lane_def, ternary_long_lane): Likewise.
2968 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2969 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2970 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2971 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2972 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2973 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2974 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2975 (ternary_uint_def, ternary_uint): Likewise.
2976 (unary_convert): Fix typo in comment.
2977 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2978 (unary_long_def, unary_long): Likewise.
2979 (unary_narrowb_def, unary_narrowb): Likewise.
2980 (unary_narrowt_def, unary_narrowt): Likewise.
2981 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2982 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2983 (unary_to_int_def, unary_to_int): Likewise.
2984 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2985 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2986 (svasrd_impl): Delete.
2987 (svcadd_impl::expand): Handle integer operations too.
2988 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2989 new functions to derive the unspec numbers.
2990 (svmla_svmls_lane_impl): Replace with...
2991 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2992 integer operations too.
2993 (svwhile_impl): Rename to...
2994 (svwhilelx_impl): ...this and inherit from while_comparison.
2995 (svasrd): Use unspec_based_function.
2996 (svmla_lane): Use svmla_lane_impl.
2997 (svmls_lane): Use svmls_lane_impl.
2998 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2999 (svwhilele, svwhilelt): Use svwhilelx_impl.
3000 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3001 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3002 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3003 * config/aarch64/aarch64-sve-builtins.def: Include
3004 aarch64-sve-builtins-sve2.def.
3006 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3008 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3009 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3010 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3011 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3012 immediates as well as vector ones.
3013 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3014 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3015 (aarch64_sve_qsub_immediate): Update calls accordingly.
3017 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3019 * config/aarch64/aarch64-sve2.md: Add banner comments.
3020 (<su>mulh<r>s<mode>3): Move further up file.
3021 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3022 (*aarch64_sve2_sra<mode>): Move further down file.
3023 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3025 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3027 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3029 (while_optab_cmp): Handle them.
3030 * config/aarch64/aarch64-sve.md
3031 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3032 and add a "@" marker.
3033 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3034 instead of gen_aarch64_sve2_while_ptest.
3035 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3037 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3039 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3040 (UNSPEC_WHILELE): ...this.
3041 (UNSPEC_WHILE_LO): Rename to...
3042 (UNSPEC_WHILELO): ...this.
3043 (UNSPEC_WHILE_LS): Rename to...
3044 (UNSPEC_WHILELS): ...this.
3045 (UNSPEC_WHILE_LT): Rename to...
3046 (UNSPEC_WHILELT): ...this.
3047 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3048 (cmp_op, while_optab_cmp): Likewise.
3049 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3050 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3051 (svwhilelt): Likewise.
3053 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3055 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3056 (unary_to_uint): Define.
3057 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3058 (unary_count): Rename to...
3059 (unary_to_uint_def, unary_to_uint): ...this.
3060 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3062 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3064 * config/aarch64/aarch64-sve-builtins-functions.h
3065 (code_for_mode_function): New class.
3066 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3067 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3068 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3069 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3070 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3072 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3074 * config/aarch64/iterators.md (addsub): New code attribute.
3075 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3077 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3078 in the asm string and attributes. Fix indentation.
3079 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3081 (@aarch64_sve_<optab><mode>): ...this.
3082 * config/aarch64/aarch64-sve-builtins.h
3083 (function_expander::expand_signed_unpred_op): Delete.
3084 * config/aarch64/aarch64-sve-builtins.cc
3085 (function_expander::expand_signed_unpred_op): Likewise.
3086 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3087 try using code_for_aarch64_sve instead.
3088 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3089 (svqsub_impl): Likewise.
3090 (svqadd, svqsub): Use rtx_code_function instead.
3092 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3094 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3095 (HADDSUB, sur, addsub): Remove them.
3097 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3099 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3100 functions the same way, rather than singling out those that
3101 aren't mapped directly to optabs.
3103 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3105 * target.def (compatible_vector_types_p): New target hook.
3106 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3107 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3108 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3109 * doc/tm.texi: Regenerate.
3110 * gimple-expr.c: Include target.h.
3111 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3112 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3114 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3115 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3116 Use the original predicate if it already has a suitable type.
3118 2020-01-09 Martin Jambor <mjambor@suse.cz>
3120 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3121 resolve_speculation and redirect_call_stmt_to_callee static. Change
3122 return type of set_call_stmt to cgraph_edge *.
3123 * auto-profile.c (afdo_indirect_call): Adjust call to
3124 redirect_call_stmt_to_callee.
3125 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3126 make the this pointer explicit, adjust self-recursive calls and the
3127 call top make_direct. Return the resulting edge.
3128 (cgraph_edge::remove): Make this pointer explicit.
3129 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3130 (cgraph_edge::make_direct): Likewise, adjust call to
3131 resolve_speculation.
3132 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3133 call to set_call_stmt.
3134 (cgraph_update_edges_for_call_stmt_node): Update call to
3135 set_call_stmt and remove.
3136 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3137 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3138 (cgraph_node::create_edge_including_clones): Moved "first" definition
3139 of edge to the block where it was used. Adjusted calls to
3141 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3142 cgraph_edge::remove.
3143 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3144 make_direct and redirect_call_stmt_to_callee.
3145 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3146 resolve_speculation and make_direct.
3147 * ipa-inline-transform.c (inline_transform): Adjust call to
3148 redirect_call_stmt_to_callee.
3149 (check_speculations_1):: Adjust call to resolve_speculation.
3150 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3151 resolve-speculation.
3152 (inline_small_functions): Adjust call to resolve_speculation.
3153 (ipa_inline): Likewise.
3154 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3156 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3157 safe with regards to edge removal, adjust calls to
3158 redirect_call_stmt_to_callee.
3159 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3160 and redirect_call_stmt_to_callee.
3161 * multiple_target.c (create_dispatcher_calls): Adjust call to
3162 redirect_call_stmt_to_callee
3163 (redirect_to_specific_clone): Likewise.
3164 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3165 Adjust calls to cgraph_edge::remove.
3166 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3167 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3168 (expand_call_inline): Adjust call to cgraph_edge::remove.
3170 2020-01-09 Martin Liska <mliska@suse.cz>
3172 * params.opt: Set Optimization for
3173 param_max_speculative_devirt_maydefs.
3175 2020-01-09 Martin Sebor <msebor@redhat.com>
3179 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3181 2020-01-09 Martin Liska <mliska@suse.cz>
3183 * auto-profile.c (auto_profile): Use opt_for_fn
3185 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3186 (propagate_vals_across_arith_jfunc): Likewise.
3187 (hint_time_bonus): Likewise.
3188 (incorporate_penalties): Likewise.
3189 (good_cloning_opportunity_p): Likewise.
3190 (perform_estimation_of_a_value): Likewise.
3191 (estimate_local_effects): Likewise.
3192 (ipcp_propagate_stage): Likewise.
3193 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3194 (set_switch_stmt_execution_predicate): Likewise.
3195 (analyze_function_body): Likewise.
3196 * ipa-inline-analysis.c (offline_size): Likewise.
3197 * ipa-inline.c (early_inliner): Likewise.
3198 * ipa-prop.c (ipa_analyze_node): Likewise.
3199 (ipcp_transform_function): Likewise.
3200 * ipa-sra.c (process_scan_results): Likewise.
3201 (ipa_sra_summarize_function): Likewise.
3202 * params.opt: Rename ipcp-unit-growth to
3203 ipa-cp-unit-growth. Add Optimization for various
3204 IPA-related parameters.
3206 2020-01-09 Richard Biener <rguenther@suse.de>
3209 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3211 2020-01-09 Richard Biener <rguenther@suse.de>
3213 PR tree-optimization/93040
3214 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3216 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3218 * common/config/avr/avr-common.c (avr_option_optimization_table)
3219 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3221 2020-01-09 Martin Liska <mliska@suse.cz>
3223 * cgraphclones.c (symbol_table::materialize_all_clones):
3224 Use cgraph_node::dump_name.
3226 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3229 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3230 output_operand_lossage instead of gcc_unreachable.
3231 * doc/md.texi (riscv f constraint): Fix typo.
3234 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3235 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3236 CONST_SCALAR_INT_P instead of CONST_INT_P.
3237 (*subv<mode>4_1): Rename to ...
3238 (subv<mode>4_1): ... this.
3239 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3240 define_insn_and_split patterns.
3241 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3244 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3246 * vec.c (class selftest::count_dtor): New class.
3247 (selftest::test_auto_delete_vec): New test.
3248 (selftest::vec_c_tests): Call it.
3249 * vec.h (class auto_delete_vec): New class template.
3250 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3252 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3254 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3256 2020-01-08 Jim Wilson <jimw@sifive.com>
3258 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3259 use of TLS_MODEL_LOCAL_EXEC when not pic.
3261 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3263 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3266 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3269 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3270 *stack_protect_set_3 peephole2): Also check that the second
3271 insns source is general_operand.
3274 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3275 predicate for output operand instead of register_operand.
3276 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3277 memory destination and non-memory operands[2].
3279 2020-01-08 Martin Liska <mliska@suse.cz>
3281 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3282 ::dump_asm_name instead of (::name or ::asm_name).
3283 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3284 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3285 (analyze_functions): Likewise.
3286 (expand_all_functions): Likewise.
3287 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3288 (propagate_bits_across_jump_function): Likewise.
3289 (dump_profile_updates): Likewise.
3290 (ipcp_store_bits_results): Likewise.
3291 (ipcp_store_vr_results): Likewise.
3292 * ipa-devirt.c (dump_targets): Likewise.
3293 * ipa-fnsummary.c (analyze_function_body): Likewise.
3294 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3295 (process_hsa_functions): Likewise.
3296 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3297 (set_alias_uids): Likewise.
3298 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3299 * ipa-inline.c (recursive_inlining): Likewise.
3300 (inline_to_all_callers_1): Likewise.
3301 (ipa_inline): Likewise.
3302 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3303 (ipa_propagate_frequency): Likewise.
3304 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3305 (remove_described_reference): Likewise.
3306 * ipa-pure-const.c (worse_state): Likewise.
3307 (check_retval_uses): Likewise.
3308 (analyze_function): Likewise.
3309 (propagate_pure_const): Likewise.
3310 (propagate_nothrow): Likewise.
3311 (dump_malloc_lattice): Likewise.
3312 (propagate_malloc): Likewise.
3313 (pass_local_pure_const::execute): Likewise.
3314 * ipa-visibility.c (optimize_weakref): Likewise.
3315 (function_and_variable_visibility): Likewise.
3316 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3317 (ipa_discover_variable_flags): Likewise.
3318 * lto-streamer-out.c (output_function): Likewise.
3319 (output_constructor): Likewise.
3320 * tree-inline.c (copy_bb): Likewise.
3321 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3322 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3324 2020-01-08 Richard Biener <rguenther@suse.de>
3327 * tree-eh.c (sink_clobbers): Update virtual operands for
3328 the first and last stmt only. Add a dry-run capability.
3329 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3330 after CFG manipulations and in RPO order to catch all
3331 secondary opportunities reliably.
3333 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3336 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3338 2019-01-08 Richard Biener <rguenther@suse.de>
3341 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3342 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3343 virtual operand, also updating SSA use.
3344 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3345 Update stmt after resetting virtual operand.
3346 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3347 * gimple-iterator.c (gsi_remove): When not removing the stmt
3348 permanently do not delink immediate uses or mark the stmt modified.
3350 2020-01-08 Martin Liska <mliska@suse.cz>
3352 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3353 (ipa_call_context::estimate_size_and_time): Likewise.
3354 (inline_analyze_function): Likewise.
3356 2020-01-08 Martin Liska <mliska@suse.cz>
3358 * cgraph.c (cgraph_node::dump): Use systematically
3361 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3363 Add -nodevicespecs option for avr.
3366 * config/avr/avr.opt (-nodevicespecs): New driver option.
3367 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3368 "-specs=device-specs/..." if that option is not set.
3369 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3371 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3373 Implement 64-bit double functions for avr.
3376 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3377 --with-double-comparison.
3378 * doc/install.texi: Document them.
3379 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3380 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3381 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3382 * doc/invoke.texi (AVR Built-in Macros): Document them.
3383 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3384 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3385 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3387 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3390 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3391 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3392 when only building rm-profile multilibs.
3394 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3397 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3398 lattice for a value to check.
3399 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3400 finite propagation in self-recursive scc.
3402 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3404 * ipa-inline.c (caller_growth_limits): Restore the AND.
3406 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3408 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3409 (VEC_ALLREG_ALT): New iterator.
3410 (VEC_ALLREG_INT_MODE): New iterator.
3411 (VCMP_MODE): New iterator.
3412 (VCMP_MODE_INT): New iterator.
3413 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3414 (vec_cmp<u>v64qidi): New define_expand.
3415 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3416 (vec_cmpu<mode>di_exec): New define_expand.
3417 (vec_cmp<u>v64qidi_exec): New define_expand.
3418 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3419 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3420 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3421 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3422 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3423 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3424 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3425 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3426 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3427 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3429 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3430 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3432 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3434 * config/gcn/constraints.md (DA): Update description and match.
3436 (Db): New constraint.
3437 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3439 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3440 Implement 'Db' mixed immediate type.
3441 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3442 (addcv64si3_dup<exec_vcc>): Delete.
3443 (subcv64si3<exec_vcc>): Rework constraints.
3444 (addv64di3): Rework constraints.
3445 (addv64di3_exec): Rework constraints.
3446 (subv64di3): Rework constraints.
3447 (addv64di3_dup): Delete.
3448 (addv64di3_dup_exec): Delete.
3449 (addv64di3_zext): Rework constraints.
3450 (addv64di3_zext_exec): Rework constraints.
3451 (addv64di3_zext_dup): Rework constraints.
3452 (addv64di3_zext_dup_exec): Rework constraints.
3453 (addv64di3_zext_dup2): Rework constraints.
3454 (addv64di3_zext_dup2_exec): Rework constraints.
3455 (addv64di3_sext_dup2): Rework constraints.
3456 (addv64di3_sext_dup2_exec): Rework constraints.
3458 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3460 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3461 existing target checks.
3463 2020-01-07 Richard Biener <rguenther@suse.de>
3465 * doc/install.texi: Bump minimal supported MPC version.
3467 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3469 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3470 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3471 * langhooks.c: Include stor-layout.h.
3472 (lhd_simulate_enum_decl): New function.
3473 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3474 handle_arm_sve_h for the LTO frontend.
3475 (register_vector_type): Cope with null returns from pushdecl.
3477 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3479 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3480 (aarch64_sve::nvectors_if_data_type): Replace with...
3481 (aarch64_sve::builtin_type_p): ...this.
3482 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3483 (find_vector_type): Delete.
3484 (add_sve_type_attribute): New function.
3485 (lookup_sve_type_attribute): Likewise.
3486 (register_builtin_types): Add an "SVE type" attribute to each type.
3487 (register_tuple_type): Likewise.
3488 (svbool_type_p, nvectors_if_data_type): Delete.
3489 (mangle_builtin_type): Use lookup_sve_type_attribute.
3490 (builtin_type_p): Likewise. Add an overload that returns the
3491 number of constituent vector and predicate registers.
3492 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3493 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3494 instead of aarch64_sve_argument_p.
3495 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3496 (aarch64_pass_by_reference): Likewise.
3497 (aarch64_function_value_1): Likewise.
3498 (aarch64_return_in_memory): Likewise.
3499 (aarch64_layout_arg): Likewise.
3501 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3503 PR tree-optimization/93156
3504 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3505 least significant bit is always clear.
3507 PR tree-optimization/93118
3508 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3509 simplifier with two intermediate conversions.
3511 2020-01-07 Martin Liska <mliska@suse.cz>
3513 * params.opt: Add Optimization for various parameters.
3515 2020-01-07 Martin Liska <mliska@suse.cz>
3518 * doc/extend.texi: Explain cloning for target_clone
3521 2020-01-07 Martin Liska <mliska@suse.cz>
3523 PR tree-optimization/92860
3524 * common.opt: Make in Optimization option
3525 as it is affected by -O0, which is an Optimization
3527 * tree-inline.c (tree_inlinable_function_p):
3528 Use opt_for_fn for warn_inline.
3529 (expand_call_inline): Likewise.
3531 2020-01-07 Martin Liska <mliska@suse.cz>
3533 PR tree-optimization/92860
3534 * common.opt: Make flag_ree as optimization
3537 2020-01-07 Martin Liska <mliska@suse.cz>
3539 PR optimization/92860
3540 * params.opt: Mark param_min_crossjump_insns with Optimization
3543 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3545 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3546 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3548 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3550 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3551 helper function to return the valid addressing formats for a given
3552 hard register and mode.
3553 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3555 * config/rs6000/constraints.md (Q constraint): Update
3557 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3560 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3561 Use 'Q' for doing vector extract from memory.
3562 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3564 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3565 doing vector extract from memory.
3566 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3567 extract from memory.
3569 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3570 for the offset being 34-bits when -mcpu=future is used.
3572 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3574 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3575 instead of cmpib_comparison_operator in cmpib patterns.
3576 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3577 of cmpib_comparison_operator. Revise comment.
3579 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3581 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3582 in an IFN_DIV_POW2 node to be equal.
3584 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3586 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3587 (vect_check_scalar_mask): ...this.
3588 (vectorizable_store, vectorizable_load): Update call accordingly.
3589 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3590 argument in calls to conditional internal functions.
3592 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3594 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3595 '0' matching inputs.
3596 (subv64di3_exec): Likewise.
3598 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3600 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3601 * doc/md.texi (movstr): Likewise.
3603 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3605 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3608 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3610 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3612 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3613 to a temporary file and use move-if-change to update the real
3614 file where necessary.
3616 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3618 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3619 rather than Upa for CPY /M.
3621 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3623 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3626 2020-01-06 Martin Liska <mliska@suse.cz>
3628 PR tree-optimization/92860
3629 * params.opt: Mark param_max_combine_insns with Optimization
3632 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3635 * config/i386/i386.md (SWIDWI): New mode iterator.
3636 (DWI, dwi): Add TImode variants.
3637 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3638 <general_hilo_operand> instead of <general_operand>. Use
3639 CONST_SCALAR_INT_P instead of CONST_INT_P.
3640 (*addv<mode>4_1): Rename to ...
3641 (addv<mode>4_1): ... this.
3642 (QWI): New mode attribute.
3643 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3644 define_insn_and_split patterns.
3645 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3647 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3648 <general_hilo_operand> instead of <general_operand>.
3649 (*addcarry<mode>_1): New define_insn.
3650 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3652 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3654 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3655 Use "call" instead of "set".
3657 2020-01-03 Martin Jambor <mjambor@suse.cz>
3660 * ipa-cp.c (print_all_lattices): Skip functions without info.
3662 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3665 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3666 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3667 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3668 for 'e' simd clones.
3671 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3673 (mprefer-vector-width=): Add Save.
3674 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3675 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3676 (ix86_debug_options, ix86_function_specific_print): Adjust
3677 ix86_target_string callers.
3678 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3679 (ix86_valid_target_attribute_tree): Likewise.
3680 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3681 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3682 ix86_target_string caller.
3685 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3686 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3687 instead of gen_int_shift_amount + convert_modes.
3689 PR rtl-optimization/93088
3690 * loop-iv.c (find_single_def_src): Punt after looking through
3691 128 reg copies for regs with single definitions. Move definitions
3694 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3696 * config/arm/arm-c.c (arm_cpu_builtins): Define
3697 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3698 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3699 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3700 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3701 * config/arm/arm-tables.opt: Regenerated.
3702 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3703 arm_arch_i8mm and arm_arch_bf16 when enabled.
3704 * config/arm/arm.h (TARGET_I8MM): New macro.
3705 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3706 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3707 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3708 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3709 (v8_6_a_simd_variants): New.
3710 (v8_*_a_simd_variants): Add i8mm and bf16.
3711 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3713 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3716 * predict.c (compute_function_frequency): Don't call
3717 warn_function_cold on functions that already have cold attribute.
3719 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3722 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3723 COMDAT group function labels in .data.rel.ro.local section.
3724 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3727 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3728 comparison_operator in B and S integer comparisons. Likewise, use
3729 ordered_comparison_operator instead of cmpib_comparison_operator in
3731 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3733 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3735 Update copyright years.
3737 * gcc.c (process_command): Update copyright notice dates.
3738 * gcov-dump.c (print_version): Ditto.
3739 * gcov.c (print_version): Ditto.
3740 * gcov-tool.c (print_version): Ditto.
3741 * gengtype.c (create_file): Ditto.
3742 * doc/cpp.texi: Bump @copying's copyright year.
3743 * doc/cppinternals.texi: Ditto.
3744 * doc/gcc.texi: Ditto.
3745 * doc/gccint.texi: Ditto.
3746 * doc/gcov.texi: Ditto.
3747 * doc/install.texi: Ditto.
3748 * doc/invoke.texi: Ditto.
3750 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3752 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3755 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3757 PR tree-optimization/93098
3758 * match.pd (popcount): For shift amounts, use integer_onep
3759 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3760 tests. Make sure that precision is power of two larger than or equal
3761 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3762 instead of ULL suffixed constants. Formatting fixes.
3764 Copyright (C) 2020 Free Software Foundation, Inc.
3766 Copying and distribution of this file, with or without modification,
3767 are permitted in any medium without royalty provided the copyright
3768 notice and this notice are preserved.