1 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
4 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
5 but different source value.
6 (adjust_callers_for_value_intersection): New function.
7 (gather_edges_for_value): Adjust order of callers to let a
8 non-self-recursive caller be the first element.
9 (self_recursive_pass_through_p): Add a new parameter "simple", and
10 check generalized self-recursive pass-through jump function.
11 (self_recursive_agg_pass_through_p): Likewise.
12 (find_more_scalar_values_for_callers_subset): Compute value from
13 pass-through jump function for self-recursive.
14 (intersect_with_plats): Cleanup previous implementation code for value
15 itersection with self-recursive call edge.
16 (intersect_with_agg_replacements): Likewise.
17 (intersect_aggregates_with_edge): Deduce value from pass-through jump
18 function for self-recursive call edge. Cleanup previous implementation
19 code for value intersection with self-recursive call edge.
20 (decide_whether_version_node): Remove dead callers and adjust order
21 to let a non-self-recursive caller be the first element.
23 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
25 * recog.c: Move pass_split_before_sched2 code in front of
26 pass_split_before_regstack.
27 (pass_data_split_before_sched2): Rename pass to split3 from split4.
28 (pass_data_split_before_regstack): Rename pass to split4 from split3.
29 (rest_of_handle_split_before_sched2): Remove.
30 (pass_split_before_sched2::execute): Unconditionally call
32 (enable_split_before_sched2): New function.
33 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
34 (pass_split_before_regstack::gate): Ditto.
35 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
36 Update name check for renamed split4 pass.
37 * config/sh/sh.c (register_sh_passes): Update pass insertion
38 point for renamed split4 pass.
40 2020-02-09 Jakub Jelinek <jakub@redhat.com>
42 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
43 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
44 copying them around between host and target.
46 2020-02-08 Andrew Pinski <apinski@marvell.com>
49 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
50 STRICT_ALIGNMENT also.
52 2020-02-08 Jim Wilson <jimw@sifive.com>
55 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
57 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
58 Jakub Jelinek <jakub@redhat.com>
61 * config/i386/i386.h (CALL_USED_REGISTERS): Make
62 xmm16-xmm31 call-used even in 64-bit ms-abi.
64 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
66 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
67 (simd_ummla, simd_usmmla): Likewise.
68 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
69 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
72 2020-02-07 Richard Biener <rguenther@suse.de>
75 * tree-inline.c (fold_marked_statements): Do a PRE walk,
76 skipping unreachable regions.
77 (optimize_inline_calls): Skip folding stmts when we didn't
80 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
83 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
84 Don't return aggregates with only SFmode and DFmode in SSE
86 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
88 2020-02-07 Jakub Jelinek <jakub@redhat.com>
91 * config/rs6000/rs6000-logue.c
92 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
93 if it fails, move rs into end_addr and retry. Add
94 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
95 the insn pattern doesn't describe well what exactly happens to
99 * config/i386/predicates.md (avx_identity_operand): Remove.
100 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
101 (avx_<castmode><avxsizesuffix>_<castmode>,
102 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
103 a VEC_CONCAT of the operand and UNSPEC_CAST.
104 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
105 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
109 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
110 recog_data.insn if distance_non_agu_define changed it.
112 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
115 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
116 we only had X-FORM (reg+reg) addressing for vectors. Also before
117 ISA 3.0, we only had X-FORM addressing for scalars in the
118 traditional Altivec registers.
120 2020-02-06 <zhongyunde@huawei.com>
121 Vladimir Makarov <vmakarov@redhat.com>
123 PR rtl-optimization/93561
124 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
127 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
129 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
132 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
134 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
135 where the low and the high 32 bits are equal to each other specially,
136 with an rldimi instruction.
138 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
140 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
142 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
144 * config/arm/arm-tables.opt: Regenerate.
146 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
149 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
150 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
151 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
153 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
155 PR rtl-optimization/87763
156 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
158 2020-02-06 Delia Burduv <delia.burduv@arm.com>
160 * config/aarch64/aarch64-simd-builtins.def
161 (bfmlaq): New built-in function.
162 (bfmlalb): New built-in function.
163 (bfmlalt): New built-in function.
164 (bfmlalb_lane): New built-in function.
165 (bfmlalt_lane): New built-in function.
166 * config/aarch64/aarch64-simd.md
167 (aarch64_bfmmlaqv4sf): New pattern.
168 (aarch64_bfmlal<bt>v4sf): New pattern.
169 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
170 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
171 (vbfmlalbq_f32): New intrinsic.
172 (vbfmlaltq_f32): New intrinsic.
173 (vbfmlalbq_lane_f32): New intrinsic.
174 (vbfmlaltq_lane_f32): New intrinsic.
175 (vbfmlalbq_laneq_f32): New intrinsic.
176 (vbfmlaltq_laneq_f32): New intrinsic.
177 * config/aarch64/iterators.md (BF_MLA): New int iterator.
178 (bt): New int attribute.
180 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
182 * config/i386/i386.md (*pushtf): Emit "#" instead of
183 calling gcc_unreachable in insn output.
186 (*pushsf_rex64): Ditto for alternatives other than 1.
187 (*pushsf): Ditto for alternatives other than 1.
189 2020-02-06 Martin Liska <mliska@suse.cz>
191 PR gcov-profile/91971
192 PR gcov-profile/93466
193 * coverage.c (coverage_init): Revert mangling of
194 path into filename. It can lead to huge filename length.
195 Creation of subfolders seem more natural.
197 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
200 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
201 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
202 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
204 2020-02-06 Jakub Jelinek <jakub@redhat.com>
207 * config/i386/predicates.md (avx_identity_operand): New predicate.
208 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
209 define_insn_and_split.
212 * omp-low.c (use_pointer_for_field): For nested constructs, also
213 look for map clauses on target construct.
214 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
215 taskreg_nesting_level.
218 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
219 shared clause, call omp_notice_variable on outer context if any.
221 2020-02-05 Jason Merrill <jason@redhat.com>
224 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
225 non-zero address even if weak and not yet defined.
227 2020-02-05 Martin Sebor <msebor@redhat.com>
229 PR tree-optimization/92765
230 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
231 * tree-ssa-strlen.c (compute_string_length): Remove.
232 (determine_min_objsize): Remove.
233 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
234 Avoid using type size as the upper bound on string length.
235 (handle_builtin_string_cmp): Add an argument. Adjust.
236 (strlen_check_and_optimize_call): Pass additional argument to
237 handle_builtin_string_cmp.
239 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
241 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
242 (*pushdi2_rex64 peephole2): Unconditionally split after
244 (*ashl<mode>3_doubleword): Ditto.
245 (*<shift_insn><mode>3_doubleword): Ditto.
247 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
250 * config/rs6000/rs6000.c (get_vector_offset): Fix
252 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
254 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
256 2020-02-05 David Malcolm <dmalcolm@redhat.com>
259 (Special Functions for Debugging the Analyzer): Update description
260 of __analyzer_dump_exploded_nodes.
262 2020-02-05 Jakub Jelinek <jakub@redhat.com>
265 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
266 include sets and not clobbers in the vzeroupper pattern.
267 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
268 the parallel has 17 (64-bit) or 9 (32-bit) elts.
269 (*avx_vzeroupper_1): New define_insn_and_split.
272 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
273 don't run when !optimize.
274 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
277 2020-02-05 Richard Biener <rguenther@suse.de>
280 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
281 checks before matching calls.
283 2020-02-05 Jakub Jelinek <jakub@redhat.com>
285 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
286 function comment typo.
289 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
290 simd_clone_create failed when i == 0, adjust clone->nargs by
293 2020-02-05 Martin Liska <mliska@suse.cz>
296 * doc/invoke.texi: Document that one should
297 not combine ASLR and -fpch.
299 2020-02-04 Richard Biener <rguenther@suse.de>
301 PR tree-optimization/93538
302 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
304 2020-02-04 Richard Biener <rguenther@suse.de>
306 PR tree-optimization/91123
307 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
308 (vn_walk_cb_data::last_vuse): New member.
309 (vn_walk_cb_data::saved_operands): Likewsie.
310 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
311 (vn_walk_cb_data::push_partial_def): Use finish.
312 (vn_reference_lookup_2): Update last_vuse and use finish if
313 we've saved operands.
314 (vn_reference_lookup_3): Use finish and update calls to
315 push_partial_defs everywhere. When translating through
316 memcpy or aggregate copies save off operands and alias-set.
317 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
318 operation for redundant store removal.
320 2020-02-04 Richard Biener <rguenther@suse.de>
322 PR tree-optimization/92819
323 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
324 generating more stmts than before.
326 2020-02-04 Martin Liska <mliska@suse.cz>
328 * config/arm/arm.c (arm_gen_far_branch): Move the function
329 outside of selftests.
331 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
333 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
334 function to adjust PC-relative vector addresses.
335 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
336 handle vectors with PC-relative addresses.
338 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
340 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
342 (hard_reg_and_mode_to_addr_mask): Delete.
343 (rs6000_adjust_vec_address): If the original vector address
344 was REG+REG or REG+OFFSET and the element is not zero, do the add
345 of the elements in the original address before adding the offset
346 for the vector element. Use address_to_insn_form to validate the
347 address using the register being loaded, rather than guessing
348 whether the address is a DS-FORM or DQ-FORM address.
350 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
352 * config/rs6000/rs6000.c (get_vector_offset): New helper function
353 to calculate the offset in memory from the start of a vector of a
354 particular element. Add code to keep the element number in
355 bounds if the element number is variable.
356 (rs6000_adjust_vec_address): Move calculation of offset of the
357 vector element to get_vector_offset.
358 (rs6000_split_vec_extract_var): Do not do the initial AND of
359 element here, move the code to get_vector_offset.
361 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
363 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
366 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
368 * config/rs6000/constraints.md: Improve documentation.
370 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
373 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
374 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
376 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
378 * config.gcc: Remove "carrizo" support.
379 * config/gcn/gcn-opts.h (processor_type): Likewise.
380 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
381 * config/gcn/gcn.opt (gpu_type): Likewise.
382 * config/gcn/t-omp-device: Likewise.
384 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
387 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
388 * config/arm/arm.c (arm_gen_far_branch): New function
390 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
392 2020-02-03 Julian Brown <julian@codesourcery.com>
393 Tobias Burnus <tobias@codesourcery.com>
395 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
397 2020-02-03 Jakub Jelinek <jakub@redhat.com>
400 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
401 valid RTL to sum up the lowest and second lowest bytes of the popcnt
404 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
406 PR rtl-optimization/91333
407 * ira-color.c (struct allocno_color_data): Add member
409 (init_allocno_threads): Set the member up.
410 (bucket_allocno_compare_func): Add compare hard reg
413 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
415 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
417 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
418 * config.in: Regenerated.
419 * configure: Regenerated.
420 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
421 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
422 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
424 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
426 * configure: Regenerate.
428 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
430 PR rtl-optimization/91333
431 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
432 reg preferences comparison up.
434 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
436 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
437 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
438 aarch64-sve-builtins-base.h.
439 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
440 aarch64-sve-builtins-base.cc.
441 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
442 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
444 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
445 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
446 (svcvtnt): New functions.
447 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
448 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
449 (svcvtnt): New functions.
450 (svcvt): Add a form that converts f32 to bf16.
451 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
452 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
454 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
455 Treat B as bfloat16_t.
456 (ternary_bfloat_lane_base): New class.
457 (ternary_bfloat_def): Likewise.
458 (ternary_bfloat): New shape.
459 (ternary_bfloat_lane_def): New class.
460 (ternary_bfloat_lane): New shape.
461 (ternary_bfloat_lanex2_def): New class.
462 (ternary_bfloat_lanex2): New shape.
463 (ternary_bfloat_opt_n_def): New class.
464 (ternary_bfloat_opt_n): New shape.
465 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
466 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
467 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
468 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
469 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
470 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
471 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
472 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
473 the pattern off the narrow mode instead of the wider one.
474 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
475 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
476 (sve_fp_op): Handle them.
477 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
478 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
480 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
482 * config/aarch64/arm_sve.h: Include arm_bf16.h.
483 * config/aarch64/aarch64-modes.def (BF): Move definition before
484 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
485 (SVE_MODES): Handle BF modes.
486 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
488 (aarch64_full_sve_mode): Likewise.
489 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
491 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
492 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
493 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
494 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
496 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
498 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
499 (TYPES_all_data): Add bf16.
500 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
501 (register_tuple_type): Increase buffer size.
502 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
503 (bf16): New type suffix.
504 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
505 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
506 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
507 Change type from all_data to all_arith.
508 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
511 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
512 Matthew Malcomson <matthew.malcomson@arm.com>
513 Richard Sandiford <richard.sandiford@arm.com>
515 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
516 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
517 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
518 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
519 __ARM_FEATURE_MATMUL_FP64.
520 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
521 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
522 be disabled at the same time.
523 (f32mm): New extension.
524 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
525 (AARCH64_FL_F64MM): Bump to the next bit up.
526 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
527 (TARGET_SVE_F64MM): New macros.
528 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
529 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
530 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
531 (UNSPEC_ZIP2Q): New unspeccs.
532 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
533 (optab, sur, perm_insn): Handle the new unspecs.
534 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
535 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
536 TARGET_SVE_F64MM instead of separate tests.
537 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
538 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
539 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
540 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
541 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
542 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
543 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
544 (TYPES_s_signed): New macro.
545 (TYPES_s_integer): Use it.
546 (TYPES_d_float): New macro.
547 (TYPES_d_data): Use it.
548 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
549 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
550 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
551 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
553 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
555 (ternary_resize2_lane_base): Likewise.
556 (ternary_resize2_base): New class.
557 (ternary_qq_lane_base): Likewise.
558 (ternary_intq_uintq_lane_def): Likewise.
559 (ternary_intq_uintq_lane): New shape.
560 (ternary_intq_uintq_opt_n_def): New class
561 (ternary_intq_uintq_opt_n): New shape.
562 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
563 (ternary_uintq_intq_def): New class.
564 (ternary_uintq_intq): New shape.
565 (ternary_uintq_intq_lane_def): New class.
566 (ternary_uintq_intq_lane): New shape.
567 (ternary_uintq_intq_opt_n_def): New class.
568 (ternary_uintq_intq_opt_n): New shape.
569 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
570 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
571 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
572 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
574 (svdotprod_lane_impl): ...this new class.
575 (svmmla_impl, svusdot_impl): New classes.
576 (svdot_lane): Update to use svdotprod_lane_impl.
577 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
578 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
580 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
581 function, with no types defined.
582 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
583 AARCH64_FL_I8MM functions.
584 (svmmla): New AARCH64_FL_F32MM function.
585 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
586 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
587 AARCH64_FL_F64MM function.
588 (REQUIRED_EXTENSIONS):
590 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
592 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
595 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
597 * config/i386/i386.md (*movoi_internal_avx): Do not check for
598 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
599 (*movti_internal): Do not check for
600 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
601 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
602 just after check for TARGET_AVX.
603 (*movdf_internal): Ditto.
604 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
605 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
606 * config/i386/sse.md (mov<mode>_internal): Only check
607 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
608 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
609 (<sse>_andnot<mode>3<mask_name>): Move check for
610 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
611 (<code><mode>3<mask_name>): Ditto.
612 (*andnot<mode>3): Ditto.
614 (*<code><mode>3): Ditto.
616 (*andnot<VI:mode>3): Remove
617 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
618 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
619 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
620 (sse4_1_blendv<ssemodesuffix>): Ditto.
621 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
622 Explain that tune applies to 128bit instructions only.
624 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
626 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
627 to definition of hsa_kernel_description. Parse assembly to find SGPR
628 and VGPR count of kernel and store in hsa_kernel_description.
630 2020-01-31 Tamar Christina <tamar.christina@arm.com>
632 PR rtl-optimization/91838
633 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
634 to truncate if allowed or reject combination.
636 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
638 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
639 (find_inv_vars_cb): Likewise.
641 2020-01-31 David Malcolm <dmalcolm@redhat.com>
643 * calls.c (special_function_p): Split out the check for DECL_NAME
644 being non-NULL and fndecl being extern at file scope into a
645 new maybe_special_function_p and call it. Drop check for fndecl
646 being non-NULL that was after a usage of DECL_NAME (fndecl).
647 * tree.h (maybe_special_function_p): New inline function.
649 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
651 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
652 (mask_gather_load<mode>): ... here, and zero-initialize the
654 (maskload<mode>di): Zero-initialize the destination.
657 2020-01-30 David Malcolm <dmalcolm@redhat.com>
660 * doc/analyzer.texi (Limitations): Note that constraints on
661 floating-point values are currently ignored.
663 2020-01-30 Jakub Jelinek <jakub@redhat.com>
666 * symtab.c (symtab_node::noninterposable_alias): If localalias
667 already exists, but is not usable, append numbers after it until
668 a unique name is found. Formatting fix.
671 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
674 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
676 * config/gcn/gcn.c (print_operand): Handle LTGT.
677 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
679 2020-01-30 Richard Biener <rguenther@suse.de>
681 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
682 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
684 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
686 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
687 without a DECL in .data.rel.ro.local.
689 2020-01-30 Jakub Jelinek <jakub@redhat.com>
692 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
697 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
698 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
699 any_extend code iterator instead of always zero_extend.
700 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
701 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
702 Use any_extend code iterator instead of always zero_extend.
703 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
704 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
705 Use any_extend code iterator instead of always zero_extend.
706 (*sse2_pmovmskb_ext): New define_insn.
707 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
710 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
711 (*popcountsi2_zext_falsedep): New define_insn.
713 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
715 * config.in: Regenerated.
716 * configure: Regenerated.
718 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
721 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
722 LLVM's assembler changed the default in version 9.
724 2020-01-24 Jeff Law <law@redhat.com>
726 PR tree-optimization/89689
727 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
729 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
733 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
735 PR rtl-optimization/87763
736 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
737 simplification to handle subregs as well as bare regs.
738 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
740 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
743 * ira.c (ira): Revert use of simplified LRA algorithm.
745 2020-01-29 Martin Jambor <mjambor@suse.cz>
747 PR tree-optimization/92706
748 * tree-sra.c (struct access): Fields first_link, last_link,
749 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
750 next_rhs_queued and grp_rhs_queued respectively, new fields
751 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
752 (struct assign_link): Field next renamed to next_rhs, new field
753 next_lhs. Updated comment.
754 (work_queue_head): Renamed to rhs_work_queue_head.
755 (lhs_work_queue_head): New variable.
756 (add_link_to_lhs): New function.
757 (relink_to_new_repr): Also relink LHS lists.
758 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
759 (add_access_to_lhs_work_queue): New function.
760 (pop_access_from_work_queue): Renamed to
761 pop_access_from_rhs_work_queue.
762 (pop_access_from_lhs_work_queue): New function.
763 (build_accesses_from_assign): Also add links to LHS lists and to LHS
765 (child_would_conflict_in_lacc): Renamed to
766 child_would_conflict_in_acc. Adjusted parameter names.
767 (create_artificial_child_access): New parameter set_grp_read, use it.
768 (subtree_mark_written_and_enqueue): Renamed to
769 subtree_mark_written_and_rhs_enqueue.
770 (propagate_subaccesses_across_link): Renamed to
771 propagate_subaccesses_from_rhs.
772 (propagate_subaccesses_from_lhs): New function.
773 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
776 2020-01-29 Martin Jambor <mjambor@suse.cz>
778 PR tree-optimization/92706
779 * tree-sra.c (struct access): Adjust comment of
780 grp_total_scalarization.
781 (find_access_in_subtree): Look for single children spanning an entire
783 (scalarizable_type_p): Allow register accesses, adjust callers.
784 (completely_scalarize): Remove function.
785 (scalarize_elem): Likewise.
786 (create_total_scalarization_access): Likewise.
787 (sort_and_splice_var_accesses): Do not track total scalarization
789 (analyze_access_subtree): New parameter totally, adjust to new meaning
790 of grp_total_scalarization.
791 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
792 (can_totally_scalarize_forest_p): New function.
793 (create_total_scalarization_access): Likewise.
794 (create_total_access_and_reshape): Likewise.
795 (total_should_skip_creating_access): Likewise.
796 (totally_scalarize_subtree): Likewise.
797 (analyze_all_variable_accesses): Perform total scalarization after
798 subaccess propagation using the new functions above.
799 (initialize_constant_pool_replacements): Output initializers by
800 traversing the access tree.
802 2020-01-29 Martin Jambor <mjambor@suse.cz>
804 * tree-sra.c (verify_sra_access_forest): New function.
805 (verify_all_sra_access_forests): Likewise.
806 (create_artificial_child_access): Set parent.
807 (analyze_all_variable_accesses): Call the verifier.
809 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
811 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
812 if called on indirect edge.
813 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
814 speculative call if needed.
816 2020-01-29 Richard Biener <rguenther@suse.de>
818 PR tree-optimization/93428
819 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
820 permutation when the load node is created.
821 (vect_analyze_slp_instance): Re-use it here.
823 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
825 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
827 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
829 PR rtl-optimization/93272
830 * ira-lives.c (process_out_of_region_eh_regs): New function.
831 (process_bb_node_lives): Call it.
833 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
835 * coverage.c (read_counts_file): Make error message lowercase.
837 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
839 * profile-count.c (profile_quality_display_names): Fix ordering.
841 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
844 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
845 hash only when edge is first within the sequence.
846 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
847 (symbol_table::create_edge): Do not set target_prob.
848 (cgraph_edge::remove_caller): Watch for speculative calls when updating
850 (cgraph_edge::make_speculative): Drop target_prob parameter.
851 (cgraph_edge::speculative_call_info): Remove.
852 (cgraph_edge::first_speculative_call_target): New member function.
853 (update_call_stmt_hash_for_removing_direct_edge): New function.
854 (cgraph_edge::resolve_speculation): Rewrite to new API.
855 (cgraph_edge::speculative_call_for_target): New member function.
856 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
857 multiple speculation targets.
858 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
860 (verify_speculative_call): Verify that targets form an interval.
861 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
862 (cgraph_edge::first_speculative_call_target): New member function.
863 (cgraph_edge::next_speculative_call_target): New member function.
864 (cgraph_edge::speculative_call_target_ref): New member function.
865 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
866 (cgraph_edge): Remove target_prob.
867 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
868 Fix handling of speculative calls.
869 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
870 * ipa-fnsummary.c (analyze_function_body): Likewise.
871 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
872 * ipa-profile.c (dump_histogram): Fix formating.
873 (ipa_profile_generate_summary): Watch for overflows.
874 (ipa_profile): Do not require probablity to be 1/2; update to new API.
875 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
876 (update_indirect_edges_after_inlining): Update to new API.
877 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
879 * profile-count.h: (profile_probability::adjusted): New.
880 * tree-inline.c (copy_bb): Update to new speculative call API; fix
882 * value-prof.c (gimple_ic_transform): Rename to ...
883 (dump_ic_profile): ... this one; update dumping.
884 (stream_in_histogram_value): Fix formating.
885 (gimple_value_profile_transformations): Update.
887 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
890 * config/i386/i386.md (*movoi_internal_avx): Remove
891 TARGET_SSE_TYPELESS_STORES check.
892 (*movti_internal): Prefer TARGET_AVX over
893 TARGET_SSE_TYPELESS_STORES.
894 (*movtf_internal): Likewise.
895 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
896 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
897 from TARGET_SSE_TYPELESS_STORES.
899 2020-01-28 David Malcolm <dmalcolm@redhat.com>
901 * diagnostic-core.h (warning_at): Rename overload to...
902 (warning_meta): ...this.
903 (emit_diagnostic_valist): Delete decl of overload taking
905 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
906 (warning_at): Rename overload taking diagnostic_metadata to...
907 (warning_meta): ...this.
909 2020-01-28 Richard Biener <rguenther@suse.de>
911 PR tree-optimization/93439
912 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
913 * tree-cfg.c (move_sese_region_to_fn): ... here.
914 (verify_types_in_gimple_reference): Verify used cliques are
917 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
920 * config/i386/i386-options.c (set_ix86_tune_features): Add an
921 argument of a pointer to struct gcc_options and pass it to
922 parse_mtune_ctrl_str.
923 (ix86_function_specific_restore): Pass opts to
924 set_ix86_tune_features.
925 (ix86_option_override_internal): Likewise.
926 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
927 gcc_options and use it for x_ix86_tune_ctrl_string.
929 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
931 PR rtl-optimization/87763
932 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
933 simplification to handle subregs as well as bare regs.
934 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
936 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
938 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
939 for reduction chains that (now) include a call.
941 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
943 PR tree-optimization/92822
944 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
945 out the don't-care elements of a vector whose significant elements
946 are duplicates, make the don't-care elements duplicates too.
948 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
950 PR tree-optimization/93434
951 * tree-predcom.c (split_data_refs_to_components): Record which
952 components have had aliasing loads removed. Prevent store-store
953 commoning for all such components.
955 2020-01-28 Jakub Jelinek <jakub@redhat.com>
958 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
959 -1 or is_vshift is true, use new_vector with number of elts npatterns
960 rather than new_unary_operation.
962 PR tree-optimization/93454
963 * gimple-fold.c (fold_array_ctor_reference): Perform
964 elt_size.to_uhwi () just once, instead of calling it in every
965 iteration. Punt if that value is above size of the temporary
966 buffer. Decrease third native_encode_expr argument when
967 bufoff + elt_sz is above size of buf.
969 2020-01-27 Joseph Myers <joseph@codesourcery.com>
971 * config/mips/mips.c (mips_declare_object_name)
972 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
974 2020-01-27 Martin Liska <mliska@suse.cz>
976 PR gcov-profile/93403
977 * tree-profile.c (gimple_init_gcov_profiler): Generate
978 both __gcov_indirect_call_profiler_v4 and
979 __gcov_indirect_call_profiler_v4_atomic.
981 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
984 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
986 (@aarch64_split_simd_mov<mode>): Use it.
987 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
988 Leave the vec_extract patterns to handle 2-element vectors.
989 (aarch64_simd_mov_from_<mode>high): Likewise.
990 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
991 (vec_extractv2dfv1df): Likewise.
993 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
995 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
996 jump conditions for *compare_condjump<GPI:mode>.
998 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1001 * digraph.cc (test_edge::test_edge): Specify template for base
1004 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1006 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1008 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1010 * config/arc/arc-protos.h (gen_mlo): Remove.
1011 (gen_mhi): Likewise.
1012 * config/arc/arc.c (AUX_MULHI): Define.
1013 (arc_must_save_reister): Special handling for r58/59.
1014 (arc_compute_frame_size): Consider mlo/mhi registers.
1015 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1017 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1018 mlo/mhi name selection.
1019 (arc_restore_callee_saves): Don't early restore blink when ISR.
1020 (arc_expand_prologue): Add mlo/mhi saving.
1021 (arc_expand_epilogue): Add mlo/mhi restoring.
1024 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1025 numbering when MUL64 option is used.
1026 (DWARF2_FRAME_REG_OUT): Define.
1027 * config/arc/arc.md (arc600_stall): New pattern.
1028 (VUNSPEC_ARC_ARC600_STALL): Define.
1029 (mulsi64): Use correct mlo/mhi registers.
1030 (mulsi_600): Clean it up.
1031 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1033 (mhi_operand): Likewise.
1035 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1036 Petro Karashchenko <petro.karashchenko@ring.com>
1038 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1039 attributes if needed.
1040 (prepare_move_operands): Generate special unspec instruction for
1042 (arc_isuncached_mem_p): Propagate uncached attribute to each
1044 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1045 (VUNSPEC_ARC_STDI): Likewise.
1046 (ALLI): New mode iterator.
1047 (mALLI): New mode attribute.
1048 (lddi): New instruction pattern.
1050 (stdidi_split): Split instruction for architectures which are not
1051 supporting ll64 option.
1052 (lddidi_split): Likewise.
1054 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1056 PR rtl-optimization/92989
1057 * lra-lives.c (process_bb_lives): Update the live-in set before
1058 processing additional clobbers.
1060 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1062 PR rtl-optimization/93170
1063 * cselib.c (cselib_invalidate_regno_val): New function, split out
1065 (cselib_invalidate_regno): ...here.
1066 (cselib_invalidated_by_call_p): New function.
1067 (cselib_process_insn): Iterate over all the hard-register entries in
1068 REG_VALUES and invalidate any that cross call-clobbered registers.
1070 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1072 * dojump.c (split_comparison): Use HONOR_NANS rather than
1073 HONOR_SNANS when splitting LTGT.
1075 2020-01-27 Martin Liska <mliska@suse.cz>
1078 * opts.c (print_filtered_help): Exclude language-specific
1079 options from --help=common unless enabled in all FEs.
1081 2020-01-27 Martin Liska <mliska@suse.cz>
1083 * opts.c (print_help): Exclude params from
1084 all except --help=param.
1086 2020-01-27 Martin Liska <mliska@suse.cz>
1089 * config/i386/i386-features.c (make_resolver_func):
1090 Align the code with ppc64 target implementation.
1091 Do not generate a unique name for resolver function.
1093 2020-01-27 Richard Biener <rguenther@suse.de>
1095 PR tree-optimization/93397
1096 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1097 converted reduction chain SLP graph adjustment.
1099 2020-01-26 Marek Polacek <polacek@redhat.com>
1102 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1105 2020-01-26 Jason Merrill <jason@redhat.com>
1108 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1111 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1113 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1114 (rx_setmem): Likewise.
1116 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1119 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1120 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1121 drop <di> from constraint of last operand.
1124 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1125 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1126 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1128 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1131 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1134 2020-01-24 Jeff Law <law@redhat.com>
1136 PR tree-optimization/92788
1137 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1140 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1143 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1144 *avx_vperm_broadcast_<mode>,
1145 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1146 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1147 Move before avx2_perm<mode>/avx512f_perm<mode>.
1150 * simplify-rtx.c (simplify_const_unary_operation,
1151 simplify_const_binary_operation): Punt for mode precision above
1152 MAX_BITSIZE_MODE_ANY_INT.
1154 2020-01-24 Andrew Pinski <apinski@marvell.com>
1156 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1159 2020-01-24 Jeff Law <law@redhat.com>
1162 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1163 for REGs. Call output_operand_lossage to get more reasonable
1166 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1168 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1169 gcn_fp_compare_operator.
1170 (vec_cmpu<mode>di): Use gcn_compare_operator.
1171 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1172 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1173 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1174 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1175 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1176 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1177 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1178 gcn_fp_compare_operator.
1179 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1180 gcn_fp_compare_operator.
1181 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1182 gcn_fp_compare_operator.
1183 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1184 gcn_fp_compare_operator.
1186 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1188 * doc/install.texi (Cross-Compiler-Specific Options): Document
1189 `--with-toolexeclibdir' option.
1191 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1193 * target.def (flags_regnum): Also mention effect on delay slot filling.
1194 * doc/tm.texi: Regenerate.
1196 2020-01-23 Jeff Law <law@redhat.com>
1198 PR translation/90162
1199 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1201 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1204 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1207 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1209 PR rtl-optimization/93402
1210 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1213 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1215 * config.in: Regenerated.
1216 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1217 for TARGET_LIBC_GNUSTACK.
1218 * configure: Regenerated.
1219 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1220 found to be 2.31 or greater.
1222 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1224 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1226 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1227 (mips_asm_file_end): New function. Delegate to
1228 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1229 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1231 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1234 * config/i386/i386-modes.def (POImode): New mode.
1235 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1236 * config/i386/i386.md (DPWI): New mode attribute.
1237 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1239 (QPWI): ... this. Use POI instead of OI for TImode.
1240 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1241 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1244 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1247 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1249 (speculation_tracker_rev): New pattern.
1250 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1251 Use speculation_tracker_rev to track the inverse condition.
1253 2020-01-23 Richard Biener <rguenther@suse.de>
1255 PR tree-optimization/93381
1256 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1257 alias-set of the def as argument and record the first one.
1258 (vn_walk_cb_data::first_set): New member.
1259 (vn_reference_lookup_3): Pass the alias-set of the current def
1260 to push_partial_def. Fix alias-set used in the aggregate copy
1262 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1263 * real.c (clear_significand_below): Fix out-of-bound access.
1265 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1268 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1269 New define_insn patterns.
1271 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1273 * doc/sourcebuild.texi (check-function-bodies): Add an
1274 optional target/xfail selector.
1276 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1278 PR rtl-optimization/93124
1279 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1280 bare USE and CLOBBER insns.
1282 2020-01-22 Andrew Pinski <apinski@marvell.com>
1284 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1286 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1289 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1290 diagnostic_manager into "ana" namespace.
1291 * selftest-run-tests.c (selftest::run_tests): Update for move of
1292 selftest::run_analyzer_selftests to
1293 ana::selftest::run_analyzer_selftests.
1295 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1297 * cfgexpand.c (union_stack_vars): Update the size.
1299 2020-01-22 Richard Biener <rguenther@suse.de>
1301 PR tree-optimization/93381
1302 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1303 throughout, handle all conversions the same.
1305 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1308 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1309 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1310 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1311 Call force_reg on high_in2 unconditionally.
1313 2020-01-22 Martin Liska <mliska@suse.cz>
1315 PR tree-optimization/92924
1316 * profile.c (compute_value_histograms): Divide
1319 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1322 * output.h (assemble_name_resolve): Declare.
1323 * varasm.c (assemble_name_resolve): New function.
1324 (assemble_name): Use it.
1325 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1327 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1329 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1330 update_web_docs_git instead of update_web_docs_svn.
1332 2020-01-21 Andrew Pinski <apinski@marvell.com>
1335 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1336 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1337 (*tlsgd_small_<mode>): Likewise.
1338 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1339 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1340 register. Convert that register back to dest using convert_mode.
1342 2020-01-21 Jim Wilson <jimw@sifive.com>
1344 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1347 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1348 Uros Bizjak <ubizjak@gmail.com>
1351 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1353 (legitimize_tls_address): Do GNU2 TLS address computation in
1354 ptr_mode and zero-extend result to Pmode.
1355 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1356 :P with :PTR and Pmode with ptr_mode.
1357 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1358 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1359 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1361 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1364 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1365 the last two operands are CONST_INT_P before using them as such.
1367 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1369 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1370 to get the integer element types.
1372 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1374 * config/aarch64/aarch64-sve-builtins.h
1375 (function_expander::convert_to_pmode): Declare.
1376 * config/aarch64/aarch64-sve-builtins.cc
1377 (function_expander::convert_to_pmode): New function.
1378 (function_expander::get_contiguous_base): Use it.
1379 (function_expander::prepare_gather_address_operands): Likewise.
1380 * config/aarch64/aarch64-sve-builtins-sve2.cc
1381 (svwhilerw_svwhilewr_impl::expand): Likewise.
1383 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1386 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1387 cfun->machine->label_is_assembled.
1388 (aarch64_print_patchable_function_entry): New.
1389 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1390 * config/aarch64/aarch64.h (struct machine_function): New field,
1393 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1396 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1399 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1402 * cgraph.c (cgraph_edge::resolve_speculation,
1403 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1404 call_stmt_site_hash.
1406 2020-01-21 Martin Liska <mliska@suse.cz>
1408 * config/rs6000/rs6000.c (common_mode_defined): Remove
1411 2020-01-21 Richard Biener <rguenther@suse.de>
1413 PR tree-optimization/92328
1414 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1415 type when value-numbering same-sized store by inserting a
1417 (eliminate_dom_walker::eliminate_stmt): When eliminating
1418 a redundant store handle bit-reinterpretation of the same value.
1420 2020-01-21 Andrew Pinski <apinski@marvel.com>
1423 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1425 (prepare_block_for_update): This. Use a worklist instead of
1428 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1430 * gcc/config/arm/arm.c (clear_operation_p):
1431 Initialise last_regno, skip first iteration
1432 based on the first_set value and use ints instead
1433 of the unnecessary HOST_WIDE_INTs.
1435 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1438 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1439 compare_mode other than SFmode or DFmode.
1441 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1444 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1445 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1446 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1448 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1450 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1452 2020-01-20 Andrew Pinski <apinski@marvell.com>
1455 * targhooks.c (default_print_patchable_function_entry): Use
1456 output_asm_insn to emit the nop instruction.
1458 2020-01-20 Fangrui Song <maskray@google.com>
1461 * targhooks.c (default_print_patchable_function_entry): Align to
1464 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1467 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1468 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1469 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1470 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1471 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1472 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1473 Remove the {q} suffix from lea.
1474 (*tls_dynamic_gnu2_call_64): Renamed to ...
1475 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1476 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1477 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1478 Pass Pmode to gen_tls_dynamic_gnu2_64.
1480 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1482 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1484 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1486 * config/aarch64/aarch64-sve-builtins-base.cc
1487 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1489 2020-01-20 Richard Biener <rguenther@suse.de>
1492 * dwarf2out.c (prune_unused_types): Unconditionally mark
1493 called function DIEs.
1495 2020-01-20 Martin Liska <mliska@suse.cz>
1497 PR tree-optimization/93199
1498 * tree-eh.c (struct leh_state): Add
1499 new field outer_non_cleanup.
1500 (cleanup_is_dead_in): Pass leh_state instead
1501 of eh_region. Add a checking that state->outer_non_cleanup
1502 points to outer non-clean up region.
1503 (lower_try_finally): Record outer_non_cleanup
1505 (lower_catch): Likewise.
1506 (lower_eh_filter): Likewise.
1507 (lower_eh_must_not_throw): Likewise.
1508 (lower_cleanup): Likewise.
1510 2020-01-20 Richard Biener <rguenther@suse.de>
1512 PR tree-optimization/93094
1513 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1514 (vect_transform_loop): Likewise.
1515 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1516 loop_vectorized_call to vect_transform_loop.
1517 * tree-vect-loop.c (vect_transform_loop): Pass down
1518 loop_vectorized_call to vect_loop_versioning.
1519 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1520 the earlier discovered loop_vectorized_call.
1522 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1524 * doc/contribute.texi: Update for SVN -> Git transition.
1525 * doc/install.texi: Likewise.
1527 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1530 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1531 speculative targets.
1532 (verify_speculative_call): New function
1533 (cgraph_node::verify_node): Use it.
1534 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1537 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1540 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1541 (cgraph_edge::make_direct): Remove all indirect targets.
1542 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1543 (cgraph_node::verify_node): Verify that only one call_stmt or
1544 lto_stmt_uid is set.
1545 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1547 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1548 (lto_output_ref): Simplify streaming of stmt.
1549 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1551 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1553 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1554 Mark parameter unused.
1556 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1558 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1560 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1562 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1564 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1566 * Makefile.in: Add coroutine-passes.o.
1567 * builtin-types.def (BT_CONST_SIZE): New.
1568 (BT_FN_BOOL_PTR): New.
1569 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1570 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1571 * coroutine-builtins.def: New file.
1572 * coroutine-passes.cc: New file.
1573 * function.h (struct GTY function): Add a bit to indicate that the
1574 function is a coroutine component.
1575 * internal-fn.c (expand_CO_FRAME): New.
1576 (expand_CO_YIELD): New.
1577 (expand_CO_SUSPN): New.
1578 (expand_CO_ACTOR): New.
1579 * internal-fn.def (CO_ACTOR): New.
1583 * passes.def: Add pass_coroutine_lower_builtins,
1584 pass_coroutine_early_expand_ifns.
1585 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1586 (make_pass_coroutine_early_expand_ifns): New.
1587 * doc/invoke.texi: Document the fcoroutines command line
1590 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1592 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1595 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1596 after checking the argument is a REG. Don't use REGNO (reg)
1597 again to set last_regno, reuse regno variable instead.
1599 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1601 * doc/analyzer.texi (Limitations): Add note about NaN.
1603 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1604 Sudakshina Das <sudi.das@arm.com>
1606 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1607 and valid immediate.
1608 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1609 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1610 * config/arm/constraints.md (Pg): New.
1611 * config/arm/predicates.md (long_shift_imm): New.
1612 (arm_reg_or_long_shift_imm): Likewise.
1613 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1614 (thumb2_lsll): Likewise.
1617 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1618 Sudakshina Das <sudi.das@arm.com>
1620 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1621 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1622 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1623 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1624 * config/arm/thumb2.md (thumb2_asrl): New.
1625 (thumb2_lsll): Likewise.
1627 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1629 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1632 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1634 * gdbinit.in (help-gcc-hooks): New command.
1635 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1636 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1639 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1641 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1642 correct target macro.
1644 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1646 * config/aarch64/aarch64-protos.h
1647 (aarch64_sve_ld1ro_operand_p): New.
1648 * config/aarch64/aarch64-sve-builtins-base.cc
1649 (class load_replicate): New.
1650 (class svld1ro_impl): New.
1651 (class svld1rq_impl): Change to inherit from load_replicate.
1652 (svld1ro): New sve intrinsic function base.
1653 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1654 New DEF_SVE_FUNCTION.
1655 * config/aarch64/aarch64-sve-builtins-base.h
1656 (svld1ro): New decl.
1657 * config/aarch64/aarch64-sve-builtins.cc
1658 (function_expander::add_mem_operand): Modify assert to allow
1660 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1662 * config/aarch64/aarch64.c
1663 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1664 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1665 (aarch64_sve_ld1ro_operand_p): New.
1666 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1667 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1668 * config/aarch64/predicates.md
1669 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1671 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1673 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1674 Introduce this ACLE specified predefined macro.
1675 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1676 (fp): Disabling this disables f64mm.
1677 (simd): Disabling this disables f64mm.
1678 (fp16): Disabling this disables f64mm.
1679 (sve): Disabling this disables f64mm.
1680 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1681 (AARCH64_ISA_F64MM): New.
1682 (TARGET_F64MM): New.
1683 * doc/invoke.texi (f64mm): Document new option.
1685 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1687 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1688 (neoversen1_tunings): Likewise.
1690 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1693 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1694 Add assert to ensure prolog has been emitted.
1695 (aarch64_split_atomic_op): Likewise.
1696 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1697 Use epilogue_completed rather than reload_completed.
1698 (aarch64_atomic_exchange<mode>): Likewise.
1699 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1700 (atomic_nand<mode>): Likewise.
1701 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1702 (atomic_fetch_nand<mode>): Likewise.
1703 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1704 (atomic_nand_fetch<mode>): Likewise.
1706 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1709 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1711 (REVERSE_CONDITION): Delete.
1712 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1713 (CCFP_CCFPE): Likewise.
1714 (e): New mode attribute.
1715 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1716 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1717 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1718 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1719 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1720 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1721 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1722 name of generator from gen_ccmpdi to gen_ccmpccdi.
1723 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1724 the previous comparison but aren't able to, use the new ccmp_rev
1727 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1729 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1730 than testing directly for INTEGER_CST.
1731 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1733 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1735 PR tree-optimization/93292
1736 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1737 get_vectype_for_scalar_type returns NULL.
1739 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1741 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1742 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1744 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1746 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1748 * params.opt: (max-predicted-iterations): Set bounds.
1749 * predict.c (real_almost_one, real_br_prob_base,
1750 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1751 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1752 probabilities; do not truncate to reg_br_prob_bases.
1753 (estimate_loops_at_level): Pass max_cyclic_prob.
1754 (estimate_loops): Compute max_cyclic_prob.
1755 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1757 * profile-count.c (profile_probability::to_sreal): New.
1758 * profile-count.h (class sreal): Move up in file.
1759 (profile_probability::to_sreal): Declare.
1761 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1764 (arm_invalid_conversion): New function for target hook.
1765 (arm_invalid_unary_op): New function for target hook.
1766 (arm_invalid_binary_op): New function for target hook.
1768 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1770 * config.gcc: Add arm_bf16.h.
1771 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1772 (arm_simd_builtin_std_type): Add BFmode.
1773 (arm_init_simd_builtin_types): Define element types for vector types.
1774 (arm_init_bf16_types): New function.
1775 (arm_init_builtins): Add arm_init_bf16_types function call.
1776 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1777 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1778 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1779 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1780 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1781 (arm_mangle_type): Add __bf16.
1782 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1783 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1784 arm_bf16_ptr_type_node.
1785 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1786 define_split between ARM registers.
1787 * config/arm/arm_bf16.h: New file.
1788 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1789 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1791 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1792 * config/arm/vfp.md: Add BFmode to movhf patterns.
1794 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1795 Andre Vieira <andre.simoesdiasvieira@arm.com>
1797 * config/arm/arm-cpus.in (mve, mve_float): New features.
1798 (dsp, mve, mve.fp): New options.
1799 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1800 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1801 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1803 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1804 Thomas Preud'homme <thomas.preudhomme@arm.com>
1806 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1808 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1809 error for using -mcmse when targeting Armv8.1-M Mainline.
1811 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1812 Thomas Preud'homme <thomas.preudhomme@arm.com>
1814 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1815 address in r4 when targeting Armv8.1-M Mainline.
1816 (nonsecure_call_value_internal): Likewise.
1817 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1818 a register match_operand again. Emit BLXNS when targeting
1820 (nonsecure_call_value_reg_thumb2): Likewise.
1822 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1823 Thomas Preud'homme <thomas.preudhomme@arm.com>
1825 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1826 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1827 variable as true when floating-point ABI is not hard. Replace
1828 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1829 Generate VLSTM and VLLDM instruction respectively before and
1830 after a function call to cmse_nonsecure_call function.
1831 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1832 (VUNSPEC_VLLDM): Likewise.
1833 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1834 (lazy_load_multiple_insn): Likewise.
1836 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1837 Thomas Preud'homme <thomas.preudhomme@arm.com>
1839 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1840 (arm_emit_vfp_multi_reg_pop): Likewise.
1841 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1842 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1843 restore callee-saved VFP registers.
1845 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1846 Thomas Preud'homme <thomas.preudhomme@arm.com>
1848 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1849 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1850 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1851 callee-saved GPRs as well as clear ip register before doing a nonsecure
1852 call then restore callee-saved GPRs after it when targeting
1854 (arm_reorg): Adapt to function rename.
1856 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1857 Thomas Preud'homme <thomas.preudhomme@arm.com>
1859 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1860 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1861 clear_vfp_multiple pattern based on a new vfp parameter.
1862 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1863 targeting Armv8.1-M Mainline.
1864 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1865 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1866 whether VFP registers are available before looking call_used_regs for a
1868 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1869 of prototype of clear_operation_p.
1870 (clear_vfp_multiple_operation): New predicate.
1871 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1872 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1874 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1875 Thomas Preud'homme <thomas.preudhomme@arm.com>
1877 * config/arm/arm-protos.h (clear_operation_p): Declare.
1878 * config/arm/arm.c (clear_operation_p): New function.
1879 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1880 targeting Armv8.1-M Mainline or successor.
1881 (output_return_instruction): Only output APSR register clearing if
1882 Armv8.1-M Mainline instructions not available.
1883 (thumb_exit): Likewise.
1884 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1885 * config/arm/thumb2.md (clear_apsr): New define_insn.
1886 (clear_multiple): Likewise.
1887 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1889 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1890 Thomas Preud'homme <thomas.preudhomme@arm.com>
1892 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1893 (use_return_insn): Also return false for Armv8.1-M Mainline.
1894 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1895 Mainline instructions are available.
1896 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1897 when targeting Armv8.1-M Mainline Security Extensions.
1898 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1899 Mainline entry function.
1900 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1901 targeting Armv8.1-M Mainline or successor.
1902 (arm_expand_epilogue): Fix indentation of caller-saved register
1903 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1905 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1906 (FP_SYSREGS): Likewise.
1907 (enum vfp_sysregs_encoding): Define enum.
1908 (fp_sysreg_names): Declare.
1909 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1910 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1911 (pop_fpsysreg_insn): Likewise.
1913 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1914 Thomas Preud'homme <thomas.preudhomme@arm.com>
1916 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1917 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1918 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1919 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1920 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1921 (ARMv8_1m_main): New feature group.
1922 (armv8.1-m.main): New architecture.
1923 * config/arm/arm-tables.opt: Regenerate.
1924 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1925 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1926 (arm_options_perform_arch_sanity_checks): Error out when targeting
1927 Armv8.1-M Mainline Security Extensions.
1928 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1930 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1932 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1933 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1934 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1935 aarch64_bfdot_laneq): New.
1936 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1937 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1938 vbfdotq_laneq_f32): New.
1939 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1940 VBFMLA_W, VBF): New.
1941 (isquadop): Add V4BF, V8BF.
1943 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1945 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1946 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1947 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1948 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1949 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1950 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1951 usdot_laneq, sudot_lane,sudot_laneq): New.
1952 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1953 (aarch64_<sur>dot_lane): New.
1954 * config/aarch64/arm_neon.h (vusdot_s32): New.
1956 (vusdot_lane_s32): New.
1957 (vsudot_lane_s32): New.
1958 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1959 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1961 2020-01-16 Martin Liska <mliska@suse.cz>
1963 * value-prof.c (dump_histogram_value): Fix
1964 obvious spacing issue.
1966 2020-01-16 Andrew Pinski <apinski@marvell.com>
1968 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1969 !storage_order_barrier_p.
1971 2020-01-16 Andrew Pinski <apinski@marvell.com>
1973 * sched-int.h (_dep): Add unused bit-field field for the padding.
1974 * sched-deps.c (init_dep_1): Init unused field.
1976 2020-01-16 Andrew Pinski <apinski@marvell.com>
1978 * optabs.h (create_expand_operand): Initialize target field also.
1980 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
1982 PR tree-optimization/92429
1983 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1984 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1986 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1989 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
1991 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1992 aarch64_sve_int_mode to each mode.
1994 2020-01-15 David Malcolm <dmalcolm@redhat.com>
1996 * doc/analyzer.texi (Overview): Add note about
1997 -fdump-ipa-analyzer.
1999 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2001 PR tree-optimization/93231
2002 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2003 input_type is unsigned. Use tree_to_shwi for shift constant.
2004 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2005 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2006 inputs more efficiently.
2008 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2010 * config/i386/i386.md (*movsf_internal): Do not require
2011 SSE2 ISA for alternatives 14 and 15.
2013 2020-01-15 Richard Biener <rguenther@suse.de>
2016 * tree-eh.c (sink_clobbers): If we already visited the destination
2017 block do not defer insertion.
2018 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2019 the purpose of defered insertion.
2021 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2023 * BASE-VER: Bump to 10.0.1.
2025 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2027 PR tree-optimization/93247
2028 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2029 type of the stmt that we're going to vectorize.
2031 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2033 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2034 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2037 2020-01-15 Martin Liska <mliska@suse.cz>
2039 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2040 2 calls of streamer_read_hwi in a function call.
2042 2020-01-15 Richard Biener <rguenther@suse.de>
2044 * alias.c (record_alias_subset): Avoid redundant work when
2045 subset is already recorded.
2047 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2049 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2050 the analyzer options provide CWE identifiers.
2052 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2054 * tree-diagnostic-path.cc (path_summary::event_range::print):
2055 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2056 using get_pure_location.
2058 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2060 PR tree-optimization/93262
2061 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2062 perform head trimming only if the last argument is constant,
2063 either all ones, or larger or equal to head trim, in the latter
2064 case decrease the last argument by head_trim.
2066 PR tree-optimization/93249
2067 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2068 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2069 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2070 perform head trim unless we can prove there are no '\0' chars
2071 from the source among the first head_trim chars.
2073 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2075 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2077 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2080 * config/i386/sse.md
2081 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2082 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2083 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2084 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2085 just a single alternative instead of two, make operands 1 and 2
2088 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2091 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2094 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2096 * Makefile.in (lang_opt_files): Add analyzer.opt.
2097 (ANALYZER_OBJS): New.
2098 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2099 tristate.o and ANALYZER_OBJS.
2100 (TEXI_GCCINT_FILES): Add analyzer.texi.
2101 * common.opt (-fanalyzer): New driver option.
2102 * config.in: Regenerate.
2103 * configure: Regenerate.
2104 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2105 (gccdepdir): Also create depdir for "analyzer" subdir.
2106 * digraph.cc: New file.
2107 * digraph.h: New file.
2108 * doc/analyzer.texi: New file.
2109 * doc/gccint.texi ("Static Analyzer") New menu item.
2110 (analyzer.texi): Include it.
2111 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2112 ("Warning Options"): Add static analysis warnings to the list.
2113 (-Wno-analyzer-double-fclose): New option.
2114 (-Wno-analyzer-double-free): New option.
2115 (-Wno-analyzer-exposure-through-output-file): New option.
2116 (-Wno-analyzer-file-leak): New option.
2117 (-Wno-analyzer-free-of-non-heap): New option.
2118 (-Wno-analyzer-malloc-leak): New option.
2119 (-Wno-analyzer-possible-null-argument): New option.
2120 (-Wno-analyzer-possible-null-dereference): New option.
2121 (-Wno-analyzer-null-argument): New option.
2122 (-Wno-analyzer-null-dereference): New option.
2123 (-Wno-analyzer-stale-setjmp-buffer): New option.
2124 (-Wno-analyzer-tainted-array-index): New option.
2125 (-Wno-analyzer-use-after-free): New option.
2126 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2127 (-Wno-analyzer-use-of-uninitialized-value): New option.
2128 (-Wanalyzer-too-complex): New option.
2129 (-fanalyzer-call-summaries): New warning.
2130 (-fanalyzer-checker=): New warning.
2131 (-fanalyzer-fine-grained): New warning.
2132 (-fno-analyzer-state-merge): New warning.
2133 (-fno-analyzer-state-purge): New warning.
2134 (-fanalyzer-transitivity): New warning.
2135 (-fanalyzer-verbose-edges): New warning.
2136 (-fanalyzer-verbose-state-changes): New warning.
2137 (-fanalyzer-verbosity=): New warning.
2138 (-fdump-analyzer): New warning.
2139 (-fdump-analyzer-callgraph): New warning.
2140 (-fdump-analyzer-exploded-graph): New warning.
2141 (-fdump-analyzer-exploded-nodes): New warning.
2142 (-fdump-analyzer-exploded-nodes-2): New warning.
2143 (-fdump-analyzer-exploded-nodes-3): New warning.
2144 (-fdump-analyzer-supergraph): New warning.
2145 * doc/sourcebuild.texi (dg-require-dot): New.
2146 (dg-check-dot): New.
2147 * gdbinit.in (break-on-saved-diagnostic): New command.
2148 * graphviz.cc: New file.
2149 * graphviz.h: New file.
2150 * ordered-hash-map-tests.cc: New file.
2151 * ordered-hash-map.h: New file.
2152 * passes.def (pass_analyzer): Add before
2153 pass_ipa_whole_program_visibility.
2154 * selftest-run-tests.c (selftest::run_tests): Call
2155 selftest::ordered_hash_map_tests_cc_tests.
2156 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2158 * shortest-paths.h: New file.
2159 * timevar.def (TV_ANALYZER): New timevar.
2160 (TV_ANALYZER_SUPERGRAPH): Likewise.
2161 (TV_ANALYZER_STATE_PURGE): Likewise.
2162 (TV_ANALYZER_PLAN): Likewise.
2163 (TV_ANALYZER_SCC): Likewise.
2164 (TV_ANALYZER_WORKLIST): Likewise.
2165 (TV_ANALYZER_DUMP): Likewise.
2166 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2167 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2168 * tree-pass.h (make_pass_analyzer): New decl.
2169 * tristate.cc: New file.
2170 * tristate.h: New file.
2172 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2175 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2176 alternatives 9 and 10.
2178 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2180 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2181 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2182 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2183 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2184 (selftest::hash_map_tests_c_tests): Call it.
2185 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2186 New static constant, using the value of = H::empty_zero_p.
2187 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2188 from default_hash_traits <Value>.
2189 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2191 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2192 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2193 calls to mark_empty with !Descriptor::empty_zero_p.
2194 (hash_table::empty_slow): Conditionalize the memset call with a
2195 check that Descriptor::empty_zero_p; otherwise, loop through the
2196 entries calling mark_empty on them.
2197 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2198 (pointer_hash::empty_zero_p): Likewise.
2199 (pair_hash::empty_zero_p): Likewise.
2200 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2202 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2203 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2204 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2205 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2206 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2207 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2208 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2209 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2211 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2214 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2216 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2217 fix typo on return value.
2219 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2222 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2224 (cgraph_edge::make_speculative): Add param for setting speculative_id
2226 (cgraph_edge::speculative_call_info): Update comments and find reference
2227 by speculative_id for multiple indirect targets.
2228 (cgraph_edge::resolve_speculation): Decrease the speculations
2229 for indirect edge, drop it's speculative if not direct target
2230 left. Update comments.
2231 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2232 (cgraph_node::dump): Print num_speculative_call_targets.
2233 (cgraph_node::verify_node): Don't report error if speculative
2234 edge not include statement.
2235 (cgraph_edge::num_speculative_call_targets_p): New function.
2236 * cgraph.h (int common_target_id): Remove.
2237 (int common_target_probability): Remove.
2238 (num_speculative_call_targets): New variable.
2239 (make_speculative): Add param for setting speculative_id.
2240 (cgraph_edge::num_speculative_call_targets_p): New declare.
2241 (target_prob): New variable.
2242 (speculative_id): New variable.
2243 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2244 call summaries for multiple speculative call targets.
2245 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2246 * ipa-profile.c (struct speculative_call_target): New struct.
2247 (class speculative_call_summary): New class.
2248 (class speculative_call_summaries): New class.
2249 (call_sums): New variable.
2250 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2251 (ipa_profile_write_edge_summary): New function.
2252 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2253 (ipa_profile_dump_all_summaries): New function.
2254 (ipa_profile_read_edge_summary): New function.
2255 (ipa_profile_read_summary_section): New function.
2256 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2257 (ipa_profile): Generate num_speculative_call_targets from
2259 * ipa-ref.h (speculative_id): New variable.
2260 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2261 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2262 common_target_probability. Stream out speculative_id and
2263 num_speculative_call_targets.
2264 (input_edge): Likewise.
2265 * predict.c (dump_prediction): Remove edges count assert to be
2267 * symtab.c (symtab_node::create_reference): Init speculative_id.
2268 (symtab_node::clone_references): Clone speculative_id.
2269 (symtab_node::clone_referring): Clone speculative_id.
2270 (symtab_node::clone_reference): Clone speculative_id.
2271 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2272 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2273 if indirect call contains multiple speculative targets.
2274 * value-prof.h (check_ic_target): Remove.
2275 * value-prof.c (gimple_value_profile_transformations):
2276 Use void function gimple_ic_transform.
2277 * value-prof.c (gimple_ic_transform): Handle topn case.
2278 Fix comment typos. Change it to a void function.
2280 2020-01-13 Andrew Pinski <apinski@marvell.com>
2282 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2283 (octeontx2t98): New define.
2284 (octeontx2t96): New define.
2285 (octeontx2t93): New define.
2286 (octeontx2f95): New define.
2287 (octeontx2f95n): New define.
2288 (octeontx2f95mm): New define.
2289 * config/aarch64/aarch64-tune.md: Regenerate.
2290 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2292 2020-01-13 Jason Merrill <jason@redhat.com>
2294 PR c++/33799 - destroy return value if local cleanup throws.
2295 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2297 2020-01-13 Martin Liska <mliska@suse.cz>
2299 * ipa-cp.c (get_max_overall_size): Use newly
2300 renamed param param_ipa_cp_unit_growth.
2301 * params.opt: Remove legacy param name.
2303 2020-01-13 Martin Sebor <msebor@redhat.com>
2305 PR tree-optimization/93213
2306 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2307 stores to be eliminated.
2309 2020-01-13 Martin Liska <mliska@suse.cz>
2311 * opts.c (print_help): Do not print CL_PARAM
2312 and CL_WARNING for CL_OPTIMIZATION.
2314 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2317 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2318 depending on optimization settings.
2320 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2322 PR tree-optimization/90838
2323 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2324 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2325 argument rather than to initialize temporary for targets that
2326 don't use the mode argument at all. Initialize ctzval to avoid
2329 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2331 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2332 * tree-core.h: Document it.
2333 * gimplify.c (gimplify_omp_workshare): Set it.
2334 * omp-low.c (lower_omp_target): Use it.
2335 * tree-pretty-print.c (dump_omp_clause): Print it.
2337 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2338 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2340 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2342 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2343 * common.opt (fdiagnostics-path-format=): New option.
2344 (diagnostic_path_format): New enum.
2345 (fdiagnostics-show-path-depths): New option.
2346 * coretypes.h (diagnostic_event_id_t): New forward decl.
2347 * diagnostic-color.c (color_dict): Add "path".
2348 * diagnostic-event-id.h: New file.
2349 * diagnostic-format-json.cc (json_from_expanded_location): Make
2351 (json_end_diagnostic): Call context->make_json_for_path if it
2352 exists and the diagnostic has a path.
2353 (diagnostic_output_format_init): Clear context->print_path.
2354 * diagnostic-path.h: New file.
2355 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2356 when printing a run of events in a diagnostic_path so that they
2357 all get the same color.
2358 (layout::m_diagnostic_path_p): New field.
2359 (layout::layout): Initialize it.
2360 (layout::print_any_labels): Don't colorize the label text for an
2361 event in a diagnostic_path.
2362 (gcc_rich_location::add_location_if_nearby): Add
2363 "restrict_to_current_line_spans" and "label" params. Pass the
2364 former to layout.maybe_add_location_range; pass the latter
2365 when calling add_range.
2366 * diagnostic.c: Include "diagnostic-path.h".
2367 (diagnostic_initialize): Initialize context->path_format and
2368 context->show_path_depths.
2369 (diagnostic_show_any_path): New function.
2370 (diagnostic_path::interprocedural_p): New function.
2371 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2372 (simple_diagnostic_path::num_events): New function.
2373 (simple_diagnostic_path::get_event): New function.
2374 (simple_diagnostic_path::add_event): New function.
2375 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2376 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2377 (debug): New overload taking a diagnostic_path *.
2378 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2379 * diagnostic.h (enum diagnostic_path_format): New enum.
2380 (json::value): New forward decl.
2381 (diagnostic_context::path_format): New field.
2382 (diagnostic_context::show_path_depths): New field.
2383 (diagnostic_context::print_path): New callback field.
2384 (diagnostic_context::make_json_for_path): New callback field.
2385 (diagnostic_show_any_path): New decl.
2386 (json_from_expanded_location): New decl.
2387 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2388 (-fdiagnostics-show-path-depths): New option.
2389 (-fdiagnostics-color): Add "path" to description of default
2390 GCC_COLORS; describe it.
2391 (-fdiagnostics-format=json): Document how diagnostic paths are
2392 represented in the JSON output format.
2393 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2394 Add optional params "restrict_to_current_line_spans" and "label".
2395 * opts.c (common_handle_option): Handle
2396 OPT_fdiagnostics_path_format_ and
2397 OPT_fdiagnostics_show_path_depths.
2398 * pretty-print.c: Include "diagnostic-event-id.h".
2399 (pp_format): Implement "%@" format code for printing
2400 diagnostic_event_id_t *.
2401 (selftest::test_pp_format): Add tests for "%@".
2402 * selftest-run-tests.c (selftest::run_tests): Call
2403 selftest::tree_diagnostic_path_cc_tests.
2404 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2405 * toplev.c (general_init): Initialize global_dc->path_format and
2406 global_dc->show_path_depths.
2407 * tree-diagnostic-path.cc: New file.
2408 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2409 non-static. Drop "diagnostic" param in favor of storing the
2410 original value of "where" and re-using it.
2411 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2412 maybe_unwind_expanded_macro_loc.
2413 (tree_diagnostics_defaults): Initialize context->print_path and
2414 context->make_json_for_path.
2415 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2417 (default_tree_make_json_for_path): New decl.
2418 (maybe_unwind_expanded_macro_loc): New decl.
2420 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2422 PR tree-optimization/93210
2423 * fold-const.h (native_encode_initializer,
2424 can_native_interpret_type_p): Declare.
2425 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2427 (native_encode_initializer): New function, moved from dwarf2out.c.
2428 Adjust to native_encode_expr compatible arguments, including dry-run
2429 and partial extraction modes. Don't handle STRING_CST.
2430 (can_native_interpret_type_p): No longer static.
2431 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2432 offset / BITS_PER_UNIT fits into int and don't call it if
2433 can_native_interpret_type_p fails. If suboff is NULL and for
2434 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2435 native_encode_initializer.
2436 (fold_const_aggregate_ref_1): Formatting fix.
2437 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2438 (tree_add_const_value_attribute): Adjust caller.
2440 PR tree-optimization/90838
2441 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2442 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2443 CTZ_DEFINED_VALUE_AT_ZERO.
2445 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2448 * lra-constraints.c (match_reload): Permit input operands have the
2449 same mode as output while other input operands have a different
2452 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2454 PR tree-optimization/90838
2455 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2456 (check_ctz_string): Likewise.
2457 (optimize_count_trailing_zeroes): Likewise.
2458 (simplify_count_trailing_zeroes): Likewise.
2459 (pass_forwprop::execute): Try ctz simplification.
2460 * match.pd: Add matching for ctz idioms.
2462 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2464 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2466 (aarch64_invalid_unary_op): New function for target hook.
2467 (aarch64_invalid_binary_op): New function for target hook.
2469 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2471 * config.gcc: Add arm_bf16.h.
2472 * config/aarch64/aarch64-builtins.c
2473 (aarch64_simd_builtin_std_type): Add BFmode.
2474 (aarch64_init_simd_builtin_types): Define element types for vector
2476 (aarch64_init_bf16_types): New function.
2477 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2478 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2480 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2481 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2483 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2484 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2485 * config/aarch64/aarch64.c
2486 (aarch64_classify_vector_mode): Add support for BF types.
2487 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2488 (aarch64_vq_mode): Add support for BF types.
2489 (aarch64_simd_container_mode): Add support for BF types.
2490 (aarch64_mangle_type): Add support for BF scalar type.
2491 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2492 * config/aarch64/arm_bf16.h: New file.
2493 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2494 * config/aarch64/iterators.md: Add BF types to mode attributes.
2495 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2497 2020-01-10 Jason Merrill <jason@redhat.com>
2499 PR c++/93173 - incorrect tree sharing.
2500 * gimplify.c (copy_if_shared): No longer static.
2501 * gimplify.h: Declare it.
2503 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2505 * doc/invoke.texi (-msve-vector-bits=): Document that
2506 -msve-vector-bits=128 now generates VL-specific code for
2507 little-endian targets.
2508 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2509 build_vector_type_for_mode to construct the data vector types.
2510 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2511 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2512 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2513 for 128-bit vectors.
2515 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2517 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2520 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2522 * config/aarch64/aarch64-builtins.c
2523 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2524 rather than checking the number of elements and the element mode.
2526 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2528 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2529 get_related_vectype_for_scalar_type rather than build_vector_type
2530 to create the index type for a conditional reduction.
2532 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2534 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2535 for any type of gather or scatter, including strided accesses.
2537 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2539 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2542 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2544 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2546 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2547 parameter and its use to reset DR_OFFSET's.
2548 (vect_transform_loop): Remove orig_drs_init argument.
2549 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2550 member of dr_vec_info rather than the offset of the associated
2551 data_reference's innermost_loop_behavior.
2552 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2553 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2554 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2555 get_dr_vinfo_offset.
2556 (vectorizable_store): Likewise.
2557 (vectorizable_load): Likewise.
2559 2020-01-10 Richard Biener <rguenther@suse.de>
2561 * gimple-ssa-store-merging
2562 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2564 2020-01-10 Martin Liska <mliska@suse.cz>
2567 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2568 encapsulation that was there before r280040.
2570 2020-01-10 Richard Biener <rguenther@suse.de>
2573 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2574 sequences to avoid walking them again for secondary opportunities.
2575 (pass_lower_eh_dispatch::execute): Instead actually insert
2578 2020-01-10 Richard Biener <rguenther@suse.de>
2581 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2582 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2583 avoid quadraticness.
2585 2020-01-10 Martin Jambor <mjambor@suse.cz>
2587 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2588 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2589 to get param_ipa_sra_max_replacements.
2590 (param_splitting_across_edge): Pass the caller to
2591 pull_accesses_from_callee.
2593 2020-01-10 Martin Jambor <mjambor@suse.cz>
2595 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2596 * ipa-cp.c (max_new_size): Removed.
2597 (orig_overall_size): New variable.
2598 (get_max_overall_size): New function.
2599 (estimate_local_effects): Use it. Adjust dump.
2600 (decide_about_value): Likewise.
2601 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2602 orig_overall_size. Adjust dump.
2603 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2605 2020-01-10 Martin Jambor <mjambor@suse.cz>
2607 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2608 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2609 instead of param_ipa_max_agg_items.
2610 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2611 optimization info for the callee.
2613 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2615 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2616 markers if debug_inline_points is false.
2618 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2620 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2622 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2623 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2624 aarch64-sve-builtins-sve2.h.
2625 (aarch64-sve-builtins-sve2.o): New rule.
2626 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2627 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2628 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2629 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2630 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2631 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2633 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2634 instructions that are handled here.
2635 (@cond_asrd<mode>): Generalize to...
2636 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2637 (*cond_asrd<mode>_2): Generalize to...
2638 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2639 (*cond_asrd<mode>_z): Generalize to...
2640 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2641 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2642 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2643 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2644 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2646 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2647 (@aarch64_scatter_stnt<mode>): Likewise.
2648 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2649 (@aarch64_mul_lane_<mode>): Likewise.
2650 (@aarch64_sve_suqadd<mode>_const): Likewise.
2651 (*<sur>h<addsub><mode>): Generalize to...
2652 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2654 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2655 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2656 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2657 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2658 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2659 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2660 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2661 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2662 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2663 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2664 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2665 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2666 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2667 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2668 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2669 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2670 (@aarch64_sve2_xar<mode>): Likewise.
2671 (@aarch64_sve2_bcax<mode>): Likewise.
2672 (*aarch64_sve2_eor3<mode>): Rename to...
2673 (@aarch64_sve2_eor3<mode>): ...this.
2674 (@aarch64_sve2_bsl<mode>): New expander.
2675 (@aarch64_sve2_nbsl<mode>): Likewise.
2676 (@aarch64_sve2_bsl1n<mode>): Likewise.
2677 (@aarch64_sve2_bsl2n<mode>): Likewise.
2678 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2679 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2680 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2681 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2682 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2683 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2684 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2685 (<su>mull<bt><Vwide>): Generalize to...
2686 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2688 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2689 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2690 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2691 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2692 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2693 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2694 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2695 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2696 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2697 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2698 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2699 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2700 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2701 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2702 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2703 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2704 (<SHRNB:r>shrnb<mode>): Generalize to...
2705 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2707 (<SHRNT:r>shrnt<mode>): Generalize to...
2708 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2710 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2711 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2712 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2713 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2714 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2715 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2716 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2717 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2718 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2719 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2720 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2721 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2722 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2723 (@aarch64_sve2_cvtnt<mode>): Likewise.
2724 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2725 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2726 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2727 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2728 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2729 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2730 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2731 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2732 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2733 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2734 (@aarch64_sve2_pmul<mode>): Likewise.
2735 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2736 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2737 (@aarch64_sve2_tbl2<mode>): Likewise.
2738 (@aarch64_sve2_tbx<mode>): Likewise.
2739 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2740 (@aarch64_sve2_histcnt<mode>): Likewise.
2741 (@aarch64_sve2_histseg<mode>): Likewise.
2742 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2743 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2744 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2745 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2746 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2747 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2748 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2749 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2750 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2751 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2752 (SVE2_PMULL_PAIR_I): New mode iterators.
2753 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2754 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2755 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2756 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2757 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2758 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2759 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2760 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2761 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2762 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2763 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2764 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2765 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2766 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2767 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2768 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2769 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2770 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2771 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2772 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2773 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2774 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2775 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2776 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2777 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2778 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2779 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2780 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2781 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2782 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2783 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2785 (VNARROW, Ventype): New mode attributes.
2786 (Vewtype): Handle VNx2DI. Fix typo in comment.
2787 (VDOUBLE): New mode attribute.
2788 (sve_lane_con): Handle VNx8HI.
2789 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2790 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2791 (sve_int_op, sve_int_op_rev): Handle the above codes.
2792 (sve_pred_int_rhs2_operand): Likewise.
2793 (MULLBT, SHRNB, SHRNT): Delete.
2794 (SVE_INT_SHIFT_IMM): New int iterator.
2795 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2796 and UNSPEC_WHILEHS for TARGET_SVE2.
2797 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2798 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2799 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2800 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2801 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2802 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2803 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2804 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2805 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2806 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2807 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2808 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2809 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2810 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2811 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2812 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2813 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2814 (optab): Handle the new unspecs.
2815 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2817 (lr): Handle the new unspecs.
2819 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2820 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2821 (sve_int_qsub_op): New int attributes.
2822 (sve_fp_op, rot): Handle the new unspecs.
2823 * config/aarch64/aarch64-sve-builtins.h
2824 (function_resolver::require_matching_pointer_type): Declare.
2825 (function_resolver::resolve_unary): Add an optional boolean argument.
2826 (function_resolver::finish_opt_n_resolution): Add an optional
2827 type_suffix_index argument.
2828 (gimple_folder::redirect_call): Declare.
2829 (gimple_expander::prepare_gather_address_operands): Add an optional
2831 * config/aarch64/aarch64-sve-builtins.cc: Include
2832 aarch64-sve-builtins-sve2.h.
2833 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2834 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2835 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2836 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2837 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2838 (TYPES_s_unsigned): Likewise.
2839 (TYPES_s_integer): Use TYPES_s_unsigned.
2840 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2841 (TYPES_sd_integer): Use them.
2842 (TYPES_d_unsigned): New macro.
2843 (TYPES_d_integer): Use it.
2844 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2845 (TYPES_cvt_narrow): Likewise.
2846 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2847 (preds_mx): New variable.
2848 (function_builder::add_overloaded_function): Allow the new feature
2849 set to be more restrictive than the original one.
2850 (function_resolver::infer_pointer_type): Remove qualifiers from
2851 the pointer type before printing it.
2852 (function_resolver::require_matching_pointer_type): New function.
2853 (function_resolver::resolve_sv_displacement): Handle functions
2854 that don't support 32-bit vector indices or svint32_t vector offsets.
2855 (function_resolver::finish_opt_n_resolution): Take the inferred type
2856 as a separate argument.
2857 (function_resolver::resolve_unary): Optionally treat all forms in
2858 the same way as normal merging functions.
2859 (gimple_folder::redirect_call): New function.
2860 (function_expander::prepare_gather_address_operands): Add an argument
2861 that says whether scaled forms are available. If they aren't,
2862 handle scaling of vector indices and don't add the extension and
2864 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2865 fall back to using cond_* instead.
2866 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2867 Split out the member variables into...
2868 (rtx_code_function_base): ...this new base class.
2869 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2870 (unspec_based_function): Split out the member variables into...
2871 (unspec_based_function_base): ...this new base class.
2872 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2873 (unspec_based_function_exact_insn): New class.
2874 (unspec_based_add_function, unspec_based_add_lane_function)
2875 (unspec_based_lane_function, unspec_based_pred_function)
2876 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2877 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2878 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2880 (unspec_based_fused_function): New class.
2881 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2882 (unspec_based_fused_lane_function): New class.
2883 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2885 (CODE_FOR_MODE1): New macro.
2886 (fixed_insn_function): New class.
2887 (while_comparison): Likewise.
2888 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2889 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2890 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2891 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2892 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2893 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2894 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2895 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2896 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2897 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2898 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2899 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2900 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2901 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2902 Also add an initial argument for unary_convert_narrowt, regardless
2903 of the predication type.
2904 (build_32_64): Allow loads and stores to specify MODE_none.
2905 (build_sv_index64, build_sv_uint_offset): New functions.
2906 (long_type_suffix): New function.
2907 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2908 (binary_imm_long_base, load_gather_sv_base): Likewise.
2909 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2910 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2911 (unary_narrowb_base, unary_narrowt_base): Likewise.
2912 (binary_long_lane_def, binary_long_lane): New shape.
2913 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2914 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2915 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2916 (binary_to_uint_def, binary_to_uint): Likewise.
2917 (binary_wide_def, binary_wide): Likewise.
2918 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2919 (compare_def, compare): Likewise.
2920 (compare_ptr_def, compare_ptr): Likewise.
2921 (load_ext_gather_index_restricted_def,
2922 load_ext_gather_index_restricted): Likewise.
2923 (load_ext_gather_offset_restricted_def,
2924 load_ext_gather_offset_restricted): Likewise.
2925 (load_gather_sv_def): Inherit from load_gather_sv_base.
2926 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2927 (shift_left_imm_def, shift_left_imm): Likewise.
2928 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2929 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2930 (store_scatter_index_restricted_def,
2931 store_scatter_index_restricted): Likewise.
2932 (store_scatter_offset_restricted_def,
2933 store_scatter_offset_restricted): Likewise.
2934 (tbl_tuple_def, tbl_tuple): Likewise.
2935 (ternary_long_lane_def, ternary_long_lane): Likewise.
2936 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2937 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2938 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2939 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2940 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2941 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2942 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2943 (ternary_uint_def, ternary_uint): Likewise.
2944 (unary_convert): Fix typo in comment.
2945 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2946 (unary_long_def, unary_long): Likewise.
2947 (unary_narrowb_def, unary_narrowb): Likewise.
2948 (unary_narrowt_def, unary_narrowt): Likewise.
2949 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2950 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2951 (unary_to_int_def, unary_to_int): Likewise.
2952 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2953 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2954 (svasrd_impl): Delete.
2955 (svcadd_impl::expand): Handle integer operations too.
2956 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2957 new functions to derive the unspec numbers.
2958 (svmla_svmls_lane_impl): Replace with...
2959 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2960 integer operations too.
2961 (svwhile_impl): Rename to...
2962 (svwhilelx_impl): ...this and inherit from while_comparison.
2963 (svasrd): Use unspec_based_function.
2964 (svmla_lane): Use svmla_lane_impl.
2965 (svmls_lane): Use svmls_lane_impl.
2966 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2967 (svwhilele, svwhilelt): Use svwhilelx_impl.
2968 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2969 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2970 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2971 * config/aarch64/aarch64-sve-builtins.def: Include
2972 aarch64-sve-builtins-sve2.def.
2974 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2976 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2977 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2978 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2979 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
2980 immediates as well as vector ones.
2981 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2982 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2983 (aarch64_sve_qsub_immediate): Update calls accordingly.
2985 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2987 * config/aarch64/aarch64-sve2.md: Add banner comments.
2988 (<su>mulh<r>s<mode>3): Move further up file.
2989 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2990 (*aarch64_sve2_sra<mode>): Move further down file.
2991 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2993 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2995 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2997 (while_optab_cmp): Handle them.
2998 * config/aarch64/aarch64-sve.md
2999 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3000 and add a "@" marker.
3001 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3002 instead of gen_aarch64_sve2_while_ptest.
3003 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3005 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3007 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3008 (UNSPEC_WHILELE): ...this.
3009 (UNSPEC_WHILE_LO): Rename to...
3010 (UNSPEC_WHILELO): ...this.
3011 (UNSPEC_WHILE_LS): Rename to...
3012 (UNSPEC_WHILELS): ...this.
3013 (UNSPEC_WHILE_LT): Rename to...
3014 (UNSPEC_WHILELT): ...this.
3015 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3016 (cmp_op, while_optab_cmp): Likewise.
3017 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3018 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3019 (svwhilelt): Likewise.
3021 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3023 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3024 (unary_to_uint): Define.
3025 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3026 (unary_count): Rename to...
3027 (unary_to_uint_def, unary_to_uint): ...this.
3028 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3030 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3032 * config/aarch64/aarch64-sve-builtins-functions.h
3033 (code_for_mode_function): New class.
3034 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3035 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3036 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3037 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3038 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3040 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3042 * config/aarch64/iterators.md (addsub): New code attribute.
3043 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3045 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3046 in the asm string and attributes. Fix indentation.
3047 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3049 (@aarch64_sve_<optab><mode>): ...this.
3050 * config/aarch64/aarch64-sve-builtins.h
3051 (function_expander::expand_signed_unpred_op): Delete.
3052 * config/aarch64/aarch64-sve-builtins.cc
3053 (function_expander::expand_signed_unpred_op): Likewise.
3054 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3055 try using code_for_aarch64_sve instead.
3056 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3057 (svqsub_impl): Likewise.
3058 (svqadd, svqsub): Use rtx_code_function instead.
3060 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3062 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3063 (HADDSUB, sur, addsub): Remove them.
3065 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3067 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3068 functions the same way, rather than singling out those that
3069 aren't mapped directly to optabs.
3071 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3073 * target.def (compatible_vector_types_p): New target hook.
3074 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3075 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3076 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3077 * doc/tm.texi: Regenerate.
3078 * gimple-expr.c: Include target.h.
3079 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3080 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3082 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3083 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3084 Use the original predicate if it already has a suitable type.
3086 2020-01-09 Martin Jambor <mjambor@suse.cz>
3088 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3089 resolve_speculation and redirect_call_stmt_to_callee static. Change
3090 return type of set_call_stmt to cgraph_edge *.
3091 * auto-profile.c (afdo_indirect_call): Adjust call to
3092 redirect_call_stmt_to_callee.
3093 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3094 make the this pointer explicit, adjust self-recursive calls and the
3095 call top make_direct. Return the resulting edge.
3096 (cgraph_edge::remove): Make this pointer explicit.
3097 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3098 (cgraph_edge::make_direct): Likewise, adjust call to
3099 resolve_speculation.
3100 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3101 call to set_call_stmt.
3102 (cgraph_update_edges_for_call_stmt_node): Update call to
3103 set_call_stmt and remove.
3104 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3105 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3106 (cgraph_node::create_edge_including_clones): Moved "first" definition
3107 of edge to the block where it was used. Adjusted calls to
3109 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3110 cgraph_edge::remove.
3111 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3112 make_direct and redirect_call_stmt_to_callee.
3113 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3114 resolve_speculation and make_direct.
3115 * ipa-inline-transform.c (inline_transform): Adjust call to
3116 redirect_call_stmt_to_callee.
3117 (check_speculations_1):: Adjust call to resolve_speculation.
3118 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3119 resolve-speculation.
3120 (inline_small_functions): Adjust call to resolve_speculation.
3121 (ipa_inline): Likewise.
3122 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3124 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3125 safe with regards to edge removal, adjust calls to
3126 redirect_call_stmt_to_callee.
3127 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3128 and redirect_call_stmt_to_callee.
3129 * multiple_target.c (create_dispatcher_calls): Adjust call to
3130 redirect_call_stmt_to_callee
3131 (redirect_to_specific_clone): Likewise.
3132 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3133 Adjust calls to cgraph_edge::remove.
3134 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3135 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3136 (expand_call_inline): Adjust call to cgraph_edge::remove.
3138 2020-01-09 Martin Liska <mliska@suse.cz>
3140 * params.opt: Set Optimization for
3141 param_max_speculative_devirt_maydefs.
3143 2020-01-09 Martin Sebor <msebor@redhat.com>
3147 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3149 2020-01-09 Martin Liska <mliska@suse.cz>
3151 * auto-profile.c (auto_profile): Use opt_for_fn
3153 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3154 (propagate_vals_across_arith_jfunc): Likewise.
3155 (hint_time_bonus): Likewise.
3156 (incorporate_penalties): Likewise.
3157 (good_cloning_opportunity_p): Likewise.
3158 (perform_estimation_of_a_value): Likewise.
3159 (estimate_local_effects): Likewise.
3160 (ipcp_propagate_stage): Likewise.
3161 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3162 (set_switch_stmt_execution_predicate): Likewise.
3163 (analyze_function_body): Likewise.
3164 * ipa-inline-analysis.c (offline_size): Likewise.
3165 * ipa-inline.c (early_inliner): Likewise.
3166 * ipa-prop.c (ipa_analyze_node): Likewise.
3167 (ipcp_transform_function): Likewise.
3168 * ipa-sra.c (process_scan_results): Likewise.
3169 (ipa_sra_summarize_function): Likewise.
3170 * params.opt: Rename ipcp-unit-growth to
3171 ipa-cp-unit-growth. Add Optimization for various
3172 IPA-related parameters.
3174 2020-01-09 Richard Biener <rguenther@suse.de>
3177 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3179 2020-01-09 Richard Biener <rguenther@suse.de>
3181 PR tree-optimization/93040
3182 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3184 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3186 * common/config/avr/avr-common.c (avr_option_optimization_table)
3187 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3189 2020-01-09 Martin Liska <mliska@suse.cz>
3191 * cgraphclones.c (symbol_table::materialize_all_clones):
3192 Use cgraph_node::dump_name.
3194 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3197 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3198 output_operand_lossage instead of gcc_unreachable.
3199 * doc/md.texi (riscv f constraint): Fix typo.
3202 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3203 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3204 CONST_SCALAR_INT_P instead of CONST_INT_P.
3205 (*subv<mode>4_1): Rename to ...
3206 (subv<mode>4_1): ... this.
3207 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3208 define_insn_and_split patterns.
3209 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3212 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3214 * vec.c (class selftest::count_dtor): New class.
3215 (selftest::test_auto_delete_vec): New test.
3216 (selftest::vec_c_tests): Call it.
3217 * vec.h (class auto_delete_vec): New class template.
3218 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3220 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3222 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3224 2020-01-08 Jim Wilson <jimw@sifive.com>
3226 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3227 use of TLS_MODEL_LOCAL_EXEC when not pic.
3229 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3231 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3234 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3237 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3238 *stack_protect_set_3 peephole2): Also check that the second
3239 insns source is general_operand.
3242 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3243 predicate for output operand instead of register_operand.
3244 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3245 memory destination and non-memory operands[2].
3247 2020-01-08 Martin Liska <mliska@suse.cz>
3249 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3250 ::dump_asm_name instead of (::name or ::asm_name).
3251 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3252 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3253 (analyze_functions): Likewise.
3254 (expand_all_functions): Likewise.
3255 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3256 (propagate_bits_across_jump_function): Likewise.
3257 (dump_profile_updates): Likewise.
3258 (ipcp_store_bits_results): Likewise.
3259 (ipcp_store_vr_results): Likewise.
3260 * ipa-devirt.c (dump_targets): Likewise.
3261 * ipa-fnsummary.c (analyze_function_body): Likewise.
3262 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3263 (process_hsa_functions): Likewise.
3264 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3265 (set_alias_uids): Likewise.
3266 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3267 * ipa-inline.c (recursive_inlining): Likewise.
3268 (inline_to_all_callers_1): Likewise.
3269 (ipa_inline): Likewise.
3270 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3271 (ipa_propagate_frequency): Likewise.
3272 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3273 (remove_described_reference): Likewise.
3274 * ipa-pure-const.c (worse_state): Likewise.
3275 (check_retval_uses): Likewise.
3276 (analyze_function): Likewise.
3277 (propagate_pure_const): Likewise.
3278 (propagate_nothrow): Likewise.
3279 (dump_malloc_lattice): Likewise.
3280 (propagate_malloc): Likewise.
3281 (pass_local_pure_const::execute): Likewise.
3282 * ipa-visibility.c (optimize_weakref): Likewise.
3283 (function_and_variable_visibility): Likewise.
3284 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3285 (ipa_discover_variable_flags): Likewise.
3286 * lto-streamer-out.c (output_function): Likewise.
3287 (output_constructor): Likewise.
3288 * tree-inline.c (copy_bb): Likewise.
3289 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3290 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3292 2020-01-08 Richard Biener <rguenther@suse.de>
3295 * tree-eh.c (sink_clobbers): Update virtual operands for
3296 the first and last stmt only. Add a dry-run capability.
3297 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3298 after CFG manipulations and in RPO order to catch all
3299 secondary opportunities reliably.
3301 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3304 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3306 2019-01-08 Richard Biener <rguenther@suse.de>
3309 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3310 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3311 virtual operand, also updating SSA use.
3312 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3313 Update stmt after resetting virtual operand.
3314 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3315 * gimple-iterator.c (gsi_remove): When not removing the stmt
3316 permanently do not delink immediate uses or mark the stmt modified.
3318 2020-01-08 Martin Liska <mliska@suse.cz>
3320 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3321 (ipa_call_context::estimate_size_and_time): Likewise.
3322 (inline_analyze_function): Likewise.
3324 2020-01-08 Martin Liska <mliska@suse.cz>
3326 * cgraph.c (cgraph_node::dump): Use systematically
3329 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3331 Add -nodevicespecs option for avr.
3334 * config/avr/avr.opt (-nodevicespecs): New driver option.
3335 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3336 "-specs=device-specs/..." if that option is not set.
3337 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3339 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3341 Implement 64-bit double functions for avr.
3344 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3345 --with-double-comparison.
3346 * doc/install.texi: Document them.
3347 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3348 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3349 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3350 * doc/invoke.texi (AVR Built-in Macros): Document them.
3351 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3352 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3353 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3355 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3358 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3359 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3360 when only building rm-profile multilibs.
3362 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3365 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3366 lattice for a value to check.
3367 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3368 finite propagation in self-recursive scc.
3370 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3372 * ipa-inline.c (caller_growth_limits): Restore the AND.
3374 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3376 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3377 (VEC_ALLREG_ALT): New iterator.
3378 (VEC_ALLREG_INT_MODE): New iterator.
3379 (VCMP_MODE): New iterator.
3380 (VCMP_MODE_INT): New iterator.
3381 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3382 (vec_cmp<u>v64qidi): New define_expand.
3383 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3384 (vec_cmpu<mode>di_exec): New define_expand.
3385 (vec_cmp<u>v64qidi_exec): New define_expand.
3386 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3387 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3388 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3389 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3390 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3391 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3392 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3393 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3394 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3395 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3397 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3398 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3400 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3402 * config/gcn/constraints.md (DA): Update description and match.
3404 (Db): New constraint.
3405 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3407 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3408 Implement 'Db' mixed immediate type.
3409 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3410 (addcv64si3_dup<exec_vcc>): Delete.
3411 (subcv64si3<exec_vcc>): Rework constraints.
3412 (addv64di3): Rework constraints.
3413 (addv64di3_exec): Rework constraints.
3414 (subv64di3): Rework constraints.
3415 (addv64di3_dup): Delete.
3416 (addv64di3_dup_exec): Delete.
3417 (addv64di3_zext): Rework constraints.
3418 (addv64di3_zext_exec): Rework constraints.
3419 (addv64di3_zext_dup): Rework constraints.
3420 (addv64di3_zext_dup_exec): Rework constraints.
3421 (addv64di3_zext_dup2): Rework constraints.
3422 (addv64di3_zext_dup2_exec): Rework constraints.
3423 (addv64di3_sext_dup2): Rework constraints.
3424 (addv64di3_sext_dup2_exec): Rework constraints.
3426 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3428 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3429 existing target checks.
3431 2020-01-07 Richard Biener <rguenther@suse.de>
3433 * doc/install.texi: Bump minimal supported MPC version.
3435 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3437 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3438 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3439 * langhooks.c: Include stor-layout.h.
3440 (lhd_simulate_enum_decl): New function.
3441 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3442 handle_arm_sve_h for the LTO frontend.
3443 (register_vector_type): Cope with null returns from pushdecl.
3445 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3447 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3448 (aarch64_sve::nvectors_if_data_type): Replace with...
3449 (aarch64_sve::builtin_type_p): ...this.
3450 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3451 (find_vector_type): Delete.
3452 (add_sve_type_attribute): New function.
3453 (lookup_sve_type_attribute): Likewise.
3454 (register_builtin_types): Add an "SVE type" attribute to each type.
3455 (register_tuple_type): Likewise.
3456 (svbool_type_p, nvectors_if_data_type): Delete.
3457 (mangle_builtin_type): Use lookup_sve_type_attribute.
3458 (builtin_type_p): Likewise. Add an overload that returns the
3459 number of constituent vector and predicate registers.
3460 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3461 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3462 instead of aarch64_sve_argument_p.
3463 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3464 (aarch64_pass_by_reference): Likewise.
3465 (aarch64_function_value_1): Likewise.
3466 (aarch64_return_in_memory): Likewise.
3467 (aarch64_layout_arg): Likewise.
3469 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3471 PR tree-optimization/93156
3472 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3473 least significant bit is always clear.
3475 PR tree-optimization/93118
3476 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3477 simplifier with two intermediate conversions.
3479 2020-01-07 Martin Liska <mliska@suse.cz>
3481 * params.opt: Add Optimization for various parameters.
3483 2020-01-07 Martin Liska <mliska@suse.cz>
3486 * doc/extend.texi: Explain cloning for target_clone
3489 2020-01-07 Martin Liska <mliska@suse.cz>
3491 PR tree-optimization/92860
3492 * common.opt: Make in Optimization option
3493 as it is affected by -O0, which is an Optimization
3495 * tree-inline.c (tree_inlinable_function_p):
3496 Use opt_for_fn for warn_inline.
3497 (expand_call_inline): Likewise.
3499 2020-01-07 Martin Liska <mliska@suse.cz>
3501 PR tree-optimization/92860
3502 * common.opt: Make flag_ree as optimization
3505 2020-01-07 Martin Liska <mliska@suse.cz>
3507 PR optimization/92860
3508 * params.opt: Mark param_min_crossjump_insns with Optimization
3511 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3513 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3514 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3516 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3518 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3519 helper function to return the valid addressing formats for a given
3520 hard register and mode.
3521 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3523 * config/rs6000/constraints.md (Q constraint): Update
3525 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3528 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3529 Use 'Q' for doing vector extract from memory.
3530 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3532 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3533 doing vector extract from memory.
3534 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3535 extract from memory.
3537 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3538 for the offset being 34-bits when -mcpu=future is used.
3540 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3542 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3543 instead of cmpib_comparison_operator in cmpib patterns.
3544 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3545 of cmpib_comparison_operator. Revise comment.
3547 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3549 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3550 in an IFN_DIV_POW2 node to be equal.
3552 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3554 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3555 (vect_check_scalar_mask): ...this.
3556 (vectorizable_store, vectorizable_load): Update call accordingly.
3557 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3558 argument in calls to conditional internal functions.
3560 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3562 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3563 '0' matching inputs.
3564 (subv64di3_exec): Likewise.
3566 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3568 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3569 * doc/md.texi (movstr): Likewise.
3571 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3573 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3576 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3578 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3580 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3581 to a temporary file and use move-if-change to update the real
3582 file where necessary.
3584 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3586 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3587 rather than Upa for CPY /M.
3589 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3591 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3594 2020-01-06 Martin Liska <mliska@suse.cz>
3596 PR tree-optimization/92860
3597 * params.opt: Mark param_max_combine_insns with Optimization
3600 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3603 * config/i386/i386.md (SWIDWI): New mode iterator.
3604 (DWI, dwi): Add TImode variants.
3605 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3606 <general_hilo_operand> instead of <general_operand>. Use
3607 CONST_SCALAR_INT_P instead of CONST_INT_P.
3608 (*addv<mode>4_1): Rename to ...
3609 (addv<mode>4_1): ... this.
3610 (QWI): New mode attribute.
3611 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3612 define_insn_and_split patterns.
3613 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3615 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3616 <general_hilo_operand> instead of <general_operand>.
3617 (*addcarry<mode>_1): New define_insn.
3618 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3620 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3622 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3623 Use "call" instead of "set".
3625 2020-01-03 Martin Jambor <mjambor@suse.cz>
3628 * ipa-cp.c (print_all_lattices): Skip functions without info.
3630 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3633 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3634 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3635 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3636 for 'e' simd clones.
3639 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3641 (mprefer-vector-width=): Add Save.
3642 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3643 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3644 (ix86_debug_options, ix86_function_specific_print): Adjust
3645 ix86_target_string callers.
3646 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3647 (ix86_valid_target_attribute_tree): Likewise.
3648 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3649 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3650 ix86_target_string caller.
3653 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3654 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3655 instead of gen_int_shift_amount + convert_modes.
3657 PR rtl-optimization/93088
3658 * loop-iv.c (find_single_def_src): Punt after looking through
3659 128 reg copies for regs with single definitions. Move definitions
3662 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3664 * config/arm/arm-c.c (arm_cpu_builtins): Define
3665 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3666 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3667 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3668 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3669 * config/arm/arm-tables.opt: Regenerated.
3670 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3671 arm_arch_i8mm and arm_arch_bf16 when enabled.
3672 * config/arm/arm.h (TARGET_I8MM): New macro.
3673 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3674 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3675 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3676 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3677 (v8_6_a_simd_variants): New.
3678 (v8_*_a_simd_variants): Add i8mm and bf16.
3679 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3681 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3684 * predict.c (compute_function_frequency): Don't call
3685 warn_function_cold on functions that already have cold attribute.
3687 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3690 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3691 COMDAT group function labels in .data.rel.ro.local section.
3692 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3695 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3696 comparison_operator in B and S integer comparisons. Likewise, use
3697 ordered_comparison_operator instead of cmpib_comparison_operator in
3699 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3701 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3703 Update copyright years.
3705 * gcc.c (process_command): Update copyright notice dates.
3706 * gcov-dump.c (print_version): Ditto.
3707 * gcov.c (print_version): Ditto.
3708 * gcov-tool.c (print_version): Ditto.
3709 * gengtype.c (create_file): Ditto.
3710 * doc/cpp.texi: Bump @copying's copyright year.
3711 * doc/cppinternals.texi: Ditto.
3712 * doc/gcc.texi: Ditto.
3713 * doc/gccint.texi: Ditto.
3714 * doc/gcov.texi: Ditto.
3715 * doc/install.texi: Ditto.
3716 * doc/invoke.texi: Ditto.
3718 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3720 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3723 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3725 PR tree-optimization/93098
3726 * match.pd (popcount): For shift amounts, use integer_onep
3727 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3728 tests. Make sure that precision is power of two larger than or equal
3729 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3730 instead of ULL suffixed constants. Formatting fixes.
3732 Copyright (C) 2020 Free Software Foundation, Inc.
3734 Copying and distribution of this file, with or without modification,
3735 are permitted in any medium without royalty provided the copyright
3736 notice and this notice are preserved.