1 2020-02-07 Richard Biener <rguenther@suse.de>
4 * tree-inline.c (fold_marked_statements): Do a PRE walk,
5 skipping unreachable regions.
6 (optimize_inline_calls): Skip folding stmts when we didn't
9 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
12 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
13 Don't return aggregates with only SFmode and DFmode in SSE
15 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
17 2020-02-07 Jakub Jelinek <jakub@redhat.com>
20 * config/rs6000/rs6000-logue.c
21 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
22 if it fails, move rs into end_addr and retry. Add
23 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
24 the insn pattern doesn't describe well what exactly happens to
28 * config/i386/predicates.md (avx_identity_operand): Remove.
29 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
30 (avx_<castmode><avxsizesuffix>_<castmode>,
31 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
32 a VEC_CONCAT of the operand and UNSPEC_CAST.
33 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
34 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
38 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
39 recog_data.insn if distance_non_agu_define changed it.
41 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
44 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
45 we only had X-FORM (reg+reg) addressing for vectors. Also before
46 ISA 3.0, we only had X-FORM addressing for scalars in the
47 traditional Altivec registers.
49 2020-02-06 <zhongyunde@huawei.com>
50 Vladimir Makarov <vmakarov@redhat.com>
52 PR rtl-optimization/93561
53 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
56 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
58 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
61 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
63 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
64 where the low and the high 32 bits are equal to each other specially,
65 with an rldimi instruction.
67 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
69 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
71 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
73 * config/arm/arm-tables.opt: Regenerate.
75 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
78 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
79 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
80 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
82 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
84 PR rtl-optimization/87763
85 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
87 2020-02-06 Delia Burduv <delia.burduv@arm.com>
89 * config/aarch64/aarch64-simd-builtins.def
90 (bfmlaq): New built-in function.
91 (bfmlalb): New built-in function.
92 (bfmlalt): New built-in function.
93 (bfmlalb_lane): New built-in function.
94 (bfmlalt_lane): New built-in function.
95 * config/aarch64/aarch64-simd.md
96 (aarch64_bfmmlaqv4sf): New pattern.
97 (aarch64_bfmlal<bt>v4sf): New pattern.
98 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
99 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
100 (vbfmlalbq_f32): New intrinsic.
101 (vbfmlaltq_f32): New intrinsic.
102 (vbfmlalbq_lane_f32): New intrinsic.
103 (vbfmlaltq_lane_f32): New intrinsic.
104 (vbfmlalbq_laneq_f32): New intrinsic.
105 (vbfmlaltq_laneq_f32): New intrinsic.
106 * config/aarch64/iterators.md (BF_MLA): New int iterator.
107 (bt): New int attribute.
109 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
111 * config/i386/i386.md (*pushtf): Emit "#" instead of
112 calling gcc_unreachable in insn output.
115 (*pushsf_rex64): Ditto for alternatives other than 1.
116 (*pushsf): Ditto for alternatives other than 1.
118 2020-02-06 Martin Liska <mliska@suse.cz>
120 PR gcov-profile/91971
121 PR gcov-profile/93466
122 * coverage.c (coverage_init): Revert mangling of
123 path into filename. It can lead to huge filename length.
124 Creation of subfolders seem more natural.
126 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
129 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
130 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
131 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
133 2020-02-06 Jakub Jelinek <jakub@redhat.com>
136 * config/i386/predicates.md (avx_identity_operand): New predicate.
137 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
138 define_insn_and_split.
141 * omp-low.c (use_pointer_for_field): For nested constructs, also
142 look for map clauses on target construct.
143 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
144 taskreg_nesting_level.
147 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
148 shared clause, call omp_notice_variable on outer context if any.
150 2020-02-05 Jason Merrill <jason@redhat.com>
153 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
154 non-zero address even if weak and not yet defined.
156 2020-02-05 Martin Sebor <msebor@redhat.com>
158 PR tree-optimization/92765
159 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
160 * tree-ssa-strlen.c (compute_string_length): Remove.
161 (determine_min_objsize): Remove.
162 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
163 Avoid using type size as the upper bound on string length.
164 (handle_builtin_string_cmp): Add an argument. Adjust.
165 (strlen_check_and_optimize_call): Pass additional argument to
166 handle_builtin_string_cmp.
168 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
170 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
171 (*pushdi2_rex64 peephole2): Unconditionally split after
173 (*ashl<mode>3_doubleword): Ditto.
174 (*<shift_insn><mode>3_doubleword): Ditto.
176 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
179 * config/rs6000/rs6000.c (get_vector_offset): Fix
181 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
183 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
185 2020-02-05 David Malcolm <dmalcolm@redhat.com>
188 (Special Functions for Debugging the Analyzer): Update description
189 of __analyzer_dump_exploded_nodes.
191 2020-02-05 Jakub Jelinek <jakub@redhat.com>
194 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
195 include sets and not clobbers in the vzeroupper pattern.
196 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
197 the parallel has 17 (64-bit) or 9 (32-bit) elts.
198 (*avx_vzeroupper_1): New define_insn_and_split.
201 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
202 don't run when !optimize.
203 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
206 2020-02-05 Richard Biener <rguenther@suse.de>
209 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
210 checks before matching calls.
212 2020-02-05 Jakub Jelinek <jakub@redhat.com>
214 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
215 function comment typo.
218 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
219 simd_clone_create failed when i == 0, adjust clone->nargs by
222 2020-02-05 Martin Liska <mliska@suse.cz>
225 * doc/invoke.texi: Document that one should
226 not combine ASLR and -fpch.
228 2020-02-04 Richard Biener <rguenther@suse.de>
230 PR tree-optimization/93538
231 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
233 2020-02-04 Richard Biener <rguenther@suse.de>
235 PR tree-optimization/91123
236 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
237 (vn_walk_cb_data::last_vuse): New member.
238 (vn_walk_cb_data::saved_operands): Likewsie.
239 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
240 (vn_walk_cb_data::push_partial_def): Use finish.
241 (vn_reference_lookup_2): Update last_vuse and use finish if
242 we've saved operands.
243 (vn_reference_lookup_3): Use finish and update calls to
244 push_partial_defs everywhere. When translating through
245 memcpy or aggregate copies save off operands and alias-set.
246 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
247 operation for redundant store removal.
249 2020-02-04 Richard Biener <rguenther@suse.de>
251 PR tree-optimization/92819
252 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
253 generating more stmts than before.
255 2020-02-04 Martin Liska <mliska@suse.cz>
257 * config/arm/arm.c (arm_gen_far_branch): Move the function
258 outside of selftests.
260 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
262 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
263 function to adjust PC-relative vector addresses.
264 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
265 handle vectors with PC-relative addresses.
267 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
269 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
271 (hard_reg_and_mode_to_addr_mask): Delete.
272 (rs6000_adjust_vec_address): If the original vector address
273 was REG+REG or REG+OFFSET and the element is not zero, do the add
274 of the elements in the original address before adding the offset
275 for the vector element. Use address_to_insn_form to validate the
276 address using the register being loaded, rather than guessing
277 whether the address is a DS-FORM or DQ-FORM address.
279 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
281 * config/rs6000/rs6000.c (get_vector_offset): New helper function
282 to calculate the offset in memory from the start of a vector of a
283 particular element. Add code to keep the element number in
284 bounds if the element number is variable.
285 (rs6000_adjust_vec_address): Move calculation of offset of the
286 vector element to get_vector_offset.
287 (rs6000_split_vec_extract_var): Do not do the initial AND of
288 element here, move the code to get_vector_offset.
290 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
292 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
295 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
297 * config/rs6000/constraints.md: Improve documentation.
299 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
302 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
303 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
305 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
307 * config.gcc: Remove "carrizo" support.
308 * config/gcn/gcn-opts.h (processor_type): Likewise.
309 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
310 * config/gcn/gcn.opt (gpu_type): Likewise.
311 * config/gcn/t-omp-device: Likewise.
313 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
316 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
317 * config/arm/arm.c (arm_gen_far_branch): New function
319 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
321 2020-02-03 Julian Brown <julian@codesourcery.com>
322 Tobias Burnus <tobias@codesourcery.com>
324 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
326 2020-02-03 Jakub Jelinek <jakub@redhat.com>
329 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
330 valid RTL to sum up the lowest and second lowest bytes of the popcnt
333 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
335 PR rtl-optimization/91333
336 * ira-color.c (struct allocno_color_data): Add member
338 (init_allocno_threads): Set the member up.
339 (bucket_allocno_compare_func): Add compare hard reg
342 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
344 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
346 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
347 * config.in: Regenerated.
348 * configure: Regenerated.
349 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
350 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
351 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
353 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
355 * configure: Regenerate.
357 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
359 PR rtl-optimization/91333
360 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
361 reg preferences comparison up.
363 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
365 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
366 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
367 aarch64-sve-builtins-base.h.
368 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
369 aarch64-sve-builtins-base.cc.
370 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
371 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
373 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
374 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
375 (svcvtnt): New functions.
376 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
377 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
378 (svcvtnt): New functions.
379 (svcvt): Add a form that converts f32 to bf16.
380 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
381 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
383 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
384 Treat B as bfloat16_t.
385 (ternary_bfloat_lane_base): New class.
386 (ternary_bfloat_def): Likewise.
387 (ternary_bfloat): New shape.
388 (ternary_bfloat_lane_def): New class.
389 (ternary_bfloat_lane): New shape.
390 (ternary_bfloat_lanex2_def): New class.
391 (ternary_bfloat_lanex2): New shape.
392 (ternary_bfloat_opt_n_def): New class.
393 (ternary_bfloat_opt_n): New shape.
394 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
395 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
396 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
397 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
398 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
399 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
400 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
401 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
402 the pattern off the narrow mode instead of the wider one.
403 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
404 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
405 (sve_fp_op): Handle them.
406 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
407 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
409 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
411 * config/aarch64/arm_sve.h: Include arm_bf16.h.
412 * config/aarch64/aarch64-modes.def (BF): Move definition before
413 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
414 (SVE_MODES): Handle BF modes.
415 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
417 (aarch64_full_sve_mode): Likewise.
418 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
420 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
421 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
422 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
423 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
425 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
427 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
428 (TYPES_all_data): Add bf16.
429 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
430 (register_tuple_type): Increase buffer size.
431 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
432 (bf16): New type suffix.
433 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
434 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
435 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
436 Change type from all_data to all_arith.
437 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
440 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
441 Matthew Malcomson <matthew.malcomson@arm.com>
442 Richard Sandiford <richard.sandiford@arm.com>
444 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
445 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
446 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
447 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
448 __ARM_FEATURE_MATMUL_FP64.
449 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
450 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
451 be disabled at the same time.
452 (f32mm): New extension.
453 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
454 (AARCH64_FL_F64MM): Bump to the next bit up.
455 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
456 (TARGET_SVE_F64MM): New macros.
457 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
458 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
459 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
460 (UNSPEC_ZIP2Q): New unspeccs.
461 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
462 (optab, sur, perm_insn): Handle the new unspecs.
463 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
464 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
465 TARGET_SVE_F64MM instead of separate tests.
466 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
467 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
468 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
469 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
470 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
471 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
472 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
473 (TYPES_s_signed): New macro.
474 (TYPES_s_integer): Use it.
475 (TYPES_d_float): New macro.
476 (TYPES_d_data): Use it.
477 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
478 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
479 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
480 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
482 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
484 (ternary_resize2_lane_base): Likewise.
485 (ternary_resize2_base): New class.
486 (ternary_qq_lane_base): Likewise.
487 (ternary_intq_uintq_lane_def): Likewise.
488 (ternary_intq_uintq_lane): New shape.
489 (ternary_intq_uintq_opt_n_def): New class
490 (ternary_intq_uintq_opt_n): New shape.
491 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
492 (ternary_uintq_intq_def): New class.
493 (ternary_uintq_intq): New shape.
494 (ternary_uintq_intq_lane_def): New class.
495 (ternary_uintq_intq_lane): New shape.
496 (ternary_uintq_intq_opt_n_def): New class.
497 (ternary_uintq_intq_opt_n): New shape.
498 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
499 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
500 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
501 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
503 (svdotprod_lane_impl): ...this new class.
504 (svmmla_impl, svusdot_impl): New classes.
505 (svdot_lane): Update to use svdotprod_lane_impl.
506 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
507 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
509 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
510 function, with no types defined.
511 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
512 AARCH64_FL_I8MM functions.
513 (svmmla): New AARCH64_FL_F32MM function.
514 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
515 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
516 AARCH64_FL_F64MM function.
517 (REQUIRED_EXTENSIONS):
519 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
521 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
524 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
526 * config/i386/i386.md (*movoi_internal_avx): Do not check for
527 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
528 (*movti_internal): Do not check for
529 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
530 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
531 just after check for TARGET_AVX.
532 (*movdf_internal): Ditto.
533 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
534 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
535 * config/i386/sse.md (mov<mode>_internal): Only check
536 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
537 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
538 (<sse>_andnot<mode>3<mask_name>): Move check for
539 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
540 (<code><mode>3<mask_name>): Ditto.
541 (*andnot<mode>3): Ditto.
543 (*<code><mode>3): Ditto.
545 (*andnot<VI:mode>3): Remove
546 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
547 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
548 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
549 (sse4_1_blendv<ssemodesuffix>): Ditto.
550 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
551 Explain that tune applies to 128bit instructions only.
553 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
555 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
556 to definition of hsa_kernel_description. Parse assembly to find SGPR
557 and VGPR count of kernel and store in hsa_kernel_description.
559 2020-01-31 Tamar Christina <tamar.christina@arm.com>
561 PR rtl-optimization/91838
562 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
563 to truncate if allowed or reject combination.
565 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
567 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
568 (find_inv_vars_cb): Likewise.
570 2020-01-31 David Malcolm <dmalcolm@redhat.com>
572 * calls.c (special_function_p): Split out the check for DECL_NAME
573 being non-NULL and fndecl being extern at file scope into a
574 new maybe_special_function_p and call it. Drop check for fndecl
575 being non-NULL that was after a usage of DECL_NAME (fndecl).
576 * tree.h (maybe_special_function_p): New inline function.
578 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
580 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
581 (mask_gather_load<mode>): ... here, and zero-initialize the
583 (maskload<mode>di): Zero-initialize the destination.
586 2020-01-30 David Malcolm <dmalcolm@redhat.com>
589 * doc/analyzer.texi (Limitations): Note that constraints on
590 floating-point values are currently ignored.
592 2020-01-30 Jakub Jelinek <jakub@redhat.com>
595 * symtab.c (symtab_node::noninterposable_alias): If localalias
596 already exists, but is not usable, append numbers after it until
597 a unique name is found. Formatting fix.
600 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
603 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
605 * config/gcn/gcn.c (print_operand): Handle LTGT.
606 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
608 2020-01-30 Richard Biener <rguenther@suse.de>
610 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
611 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
613 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
615 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
616 without a DECL in .data.rel.ro.local.
618 2020-01-30 Jakub Jelinek <jakub@redhat.com>
621 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
626 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
627 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
628 any_extend code iterator instead of always zero_extend.
629 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
630 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
631 Use any_extend code iterator instead of always zero_extend.
632 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
633 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
634 Use any_extend code iterator instead of always zero_extend.
635 (*sse2_pmovmskb_ext): New define_insn.
636 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
639 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
640 (*popcountsi2_zext_falsedep): New define_insn.
642 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
644 * config.in: Regenerated.
645 * configure: Regenerated.
647 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
650 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
651 LLVM's assembler changed the default in version 9.
653 2020-01-24 Jeff Law <law@redhat.com>
655 PR tree-optimization/89689
656 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
658 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
662 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
664 PR rtl-optimization/87763
665 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
666 simplification to handle subregs as well as bare regs.
667 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
669 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
672 * ira.c (ira): Revert use of simplified LRA algorithm.
674 2020-01-29 Martin Jambor <mjambor@suse.cz>
676 PR tree-optimization/92706
677 * tree-sra.c (struct access): Fields first_link, last_link,
678 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
679 next_rhs_queued and grp_rhs_queued respectively, new fields
680 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
681 (struct assign_link): Field next renamed to next_rhs, new field
682 next_lhs. Updated comment.
683 (work_queue_head): Renamed to rhs_work_queue_head.
684 (lhs_work_queue_head): New variable.
685 (add_link_to_lhs): New function.
686 (relink_to_new_repr): Also relink LHS lists.
687 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
688 (add_access_to_lhs_work_queue): New function.
689 (pop_access_from_work_queue): Renamed to
690 pop_access_from_rhs_work_queue.
691 (pop_access_from_lhs_work_queue): New function.
692 (build_accesses_from_assign): Also add links to LHS lists and to LHS
694 (child_would_conflict_in_lacc): Renamed to
695 child_would_conflict_in_acc. Adjusted parameter names.
696 (create_artificial_child_access): New parameter set_grp_read, use it.
697 (subtree_mark_written_and_enqueue): Renamed to
698 subtree_mark_written_and_rhs_enqueue.
699 (propagate_subaccesses_across_link): Renamed to
700 propagate_subaccesses_from_rhs.
701 (propagate_subaccesses_from_lhs): New function.
702 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
705 2020-01-29 Martin Jambor <mjambor@suse.cz>
707 PR tree-optimization/92706
708 * tree-sra.c (struct access): Adjust comment of
709 grp_total_scalarization.
710 (find_access_in_subtree): Look for single children spanning an entire
712 (scalarizable_type_p): Allow register accesses, adjust callers.
713 (completely_scalarize): Remove function.
714 (scalarize_elem): Likewise.
715 (create_total_scalarization_access): Likewise.
716 (sort_and_splice_var_accesses): Do not track total scalarization
718 (analyze_access_subtree): New parameter totally, adjust to new meaning
719 of grp_total_scalarization.
720 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
721 (can_totally_scalarize_forest_p): New function.
722 (create_total_scalarization_access): Likewise.
723 (create_total_access_and_reshape): Likewise.
724 (total_should_skip_creating_access): Likewise.
725 (totally_scalarize_subtree): Likewise.
726 (analyze_all_variable_accesses): Perform total scalarization after
727 subaccess propagation using the new functions above.
728 (initialize_constant_pool_replacements): Output initializers by
729 traversing the access tree.
731 2020-01-29 Martin Jambor <mjambor@suse.cz>
733 * tree-sra.c (verify_sra_access_forest): New function.
734 (verify_all_sra_access_forests): Likewise.
735 (create_artificial_child_access): Set parent.
736 (analyze_all_variable_accesses): Call the verifier.
738 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
740 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
741 if called on indirect edge.
742 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
743 speculative call if needed.
745 2020-01-29 Richard Biener <rguenther@suse.de>
747 PR tree-optimization/93428
748 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
749 permutation when the load node is created.
750 (vect_analyze_slp_instance): Re-use it here.
752 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
754 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
756 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
758 PR rtl-optimization/93272
759 * ira-lives.c (process_out_of_region_eh_regs): New function.
760 (process_bb_node_lives): Call it.
762 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
764 * coverage.c (read_counts_file): Make error message lowercase.
766 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
768 * profile-count.c (profile_quality_display_names): Fix ordering.
770 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
773 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
774 hash only when edge is first within the sequence.
775 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
776 (symbol_table::create_edge): Do not set target_prob.
777 (cgraph_edge::remove_caller): Watch for speculative calls when updating
779 (cgraph_edge::make_speculative): Drop target_prob parameter.
780 (cgraph_edge::speculative_call_info): Remove.
781 (cgraph_edge::first_speculative_call_target): New member function.
782 (update_call_stmt_hash_for_removing_direct_edge): New function.
783 (cgraph_edge::resolve_speculation): Rewrite to new API.
784 (cgraph_edge::speculative_call_for_target): New member function.
785 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
786 multiple speculation targets.
787 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
789 (verify_speculative_call): Verify that targets form an interval.
790 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
791 (cgraph_edge::first_speculative_call_target): New member function.
792 (cgraph_edge::next_speculative_call_target): New member function.
793 (cgraph_edge::speculative_call_target_ref): New member function.
794 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
795 (cgraph_edge): Remove target_prob.
796 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
797 Fix handling of speculative calls.
798 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
799 * ipa-fnsummary.c (analyze_function_body): Likewise.
800 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
801 * ipa-profile.c (dump_histogram): Fix formating.
802 (ipa_profile_generate_summary): Watch for overflows.
803 (ipa_profile): Do not require probablity to be 1/2; update to new API.
804 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
805 (update_indirect_edges_after_inlining): Update to new API.
806 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
808 * profile-count.h: (profile_probability::adjusted): New.
809 * tree-inline.c (copy_bb): Update to new speculative call API; fix
811 * value-prof.c (gimple_ic_transform): Rename to ...
812 (dump_ic_profile): ... this one; update dumping.
813 (stream_in_histogram_value): Fix formating.
814 (gimple_value_profile_transformations): Update.
816 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
819 * config/i386/i386.md (*movoi_internal_avx): Remove
820 TARGET_SSE_TYPELESS_STORES check.
821 (*movti_internal): Prefer TARGET_AVX over
822 TARGET_SSE_TYPELESS_STORES.
823 (*movtf_internal): Likewise.
824 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
825 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
826 from TARGET_SSE_TYPELESS_STORES.
828 2020-01-28 David Malcolm <dmalcolm@redhat.com>
830 * diagnostic-core.h (warning_at): Rename overload to...
831 (warning_meta): ...this.
832 (emit_diagnostic_valist): Delete decl of overload taking
834 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
835 (warning_at): Rename overload taking diagnostic_metadata to...
836 (warning_meta): ...this.
838 2020-01-28 Richard Biener <rguenther@suse.de>
840 PR tree-optimization/93439
841 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
842 * tree-cfg.c (move_sese_region_to_fn): ... here.
843 (verify_types_in_gimple_reference): Verify used cliques are
846 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
849 * config/i386/i386-options.c (set_ix86_tune_features): Add an
850 argument of a pointer to struct gcc_options and pass it to
851 parse_mtune_ctrl_str.
852 (ix86_function_specific_restore): Pass opts to
853 set_ix86_tune_features.
854 (ix86_option_override_internal): Likewise.
855 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
856 gcc_options and use it for x_ix86_tune_ctrl_string.
858 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
860 PR rtl-optimization/87763
861 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
862 simplification to handle subregs as well as bare regs.
863 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
865 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
867 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
868 for reduction chains that (now) include a call.
870 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
872 PR tree-optimization/92822
873 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
874 out the don't-care elements of a vector whose significant elements
875 are duplicates, make the don't-care elements duplicates too.
877 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
879 PR tree-optimization/93434
880 * tree-predcom.c (split_data_refs_to_components): Record which
881 components have had aliasing loads removed. Prevent store-store
882 commoning for all such components.
884 2020-01-28 Jakub Jelinek <jakub@redhat.com>
887 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
888 -1 or is_vshift is true, use new_vector with number of elts npatterns
889 rather than new_unary_operation.
891 PR tree-optimization/93454
892 * gimple-fold.c (fold_array_ctor_reference): Perform
893 elt_size.to_uhwi () just once, instead of calling it in every
894 iteration. Punt if that value is above size of the temporary
895 buffer. Decrease third native_encode_expr argument when
896 bufoff + elt_sz is above size of buf.
898 2020-01-27 Joseph Myers <joseph@codesourcery.com>
900 * config/mips/mips.c (mips_declare_object_name)
901 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
903 2020-01-27 Martin Liska <mliska@suse.cz>
905 PR gcov-profile/93403
906 * tree-profile.c (gimple_init_gcov_profiler): Generate
907 both __gcov_indirect_call_profiler_v4 and
908 __gcov_indirect_call_profiler_v4_atomic.
910 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
913 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
915 (@aarch64_split_simd_mov<mode>): Use it.
916 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
917 Leave the vec_extract patterns to handle 2-element vectors.
918 (aarch64_simd_mov_from_<mode>high): Likewise.
919 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
920 (vec_extractv2dfv1df): Likewise.
922 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
924 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
925 jump conditions for *compare_condjump<GPI:mode>.
927 2020-01-27 David Malcolm <dmalcolm@redhat.com>
930 * digraph.cc (test_edge::test_edge): Specify template for base
933 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
935 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
937 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
939 * config/arc/arc-protos.h (gen_mlo): Remove.
941 * config/arc/arc.c (AUX_MULHI): Define.
942 (arc_must_save_reister): Special handling for r58/59.
943 (arc_compute_frame_size): Consider mlo/mhi registers.
944 (arc_save_callee_saves): Emit fp/sp move only when emit_move
946 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
947 mlo/mhi name selection.
948 (arc_restore_callee_saves): Don't early restore blink when ISR.
949 (arc_expand_prologue): Add mlo/mhi saving.
950 (arc_expand_epilogue): Add mlo/mhi restoring.
953 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
954 numbering when MUL64 option is used.
955 (DWARF2_FRAME_REG_OUT): Define.
956 * config/arc/arc.md (arc600_stall): New pattern.
957 (VUNSPEC_ARC_ARC600_STALL): Define.
958 (mulsi64): Use correct mlo/mhi registers.
959 (mulsi_600): Clean it up.
960 * config/arc/predicates.md (mlo_operand): Remove any dependency on
962 (mhi_operand): Likewise.
964 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
965 Petro Karashchenko <petro.karashchenko@ring.com>
967 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
968 attributes if needed.
969 (prepare_move_operands): Generate special unspec instruction for
971 (arc_isuncached_mem_p): Propagate uncached attribute to each
973 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
974 (VUNSPEC_ARC_STDI): Likewise.
975 (ALLI): New mode iterator.
976 (mALLI): New mode attribute.
977 (lddi): New instruction pattern.
979 (stdidi_split): Split instruction for architectures which are not
980 supporting ll64 option.
981 (lddidi_split): Likewise.
983 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
985 PR rtl-optimization/92989
986 * lra-lives.c (process_bb_lives): Update the live-in set before
987 processing additional clobbers.
989 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
991 PR rtl-optimization/93170
992 * cselib.c (cselib_invalidate_regno_val): New function, split out
994 (cselib_invalidate_regno): ...here.
995 (cselib_invalidated_by_call_p): New function.
996 (cselib_process_insn): Iterate over all the hard-register entries in
997 REG_VALUES and invalidate any that cross call-clobbered registers.
999 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1001 * dojump.c (split_comparison): Use HONOR_NANS rather than
1002 HONOR_SNANS when splitting LTGT.
1004 2020-01-27 Martin Liska <mliska@suse.cz>
1007 * opts.c (print_filtered_help): Exclude language-specific
1008 options from --help=common unless enabled in all FEs.
1010 2020-01-27 Martin Liska <mliska@suse.cz>
1012 * opts.c (print_help): Exclude params from
1013 all except --help=param.
1015 2020-01-27 Martin Liska <mliska@suse.cz>
1018 * config/i386/i386-features.c (make_resolver_func):
1019 Align the code with ppc64 target implementation.
1020 Do not generate a unique name for resolver function.
1022 2020-01-27 Richard Biener <rguenther@suse.de>
1024 PR tree-optimization/93397
1025 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1026 converted reduction chain SLP graph adjustment.
1028 2020-01-26 Marek Polacek <polacek@redhat.com>
1031 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1034 2020-01-26 Jason Merrill <jason@redhat.com>
1037 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1040 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1042 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1043 (rx_setmem): Likewise.
1045 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1048 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1049 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1050 drop <di> from constraint of last operand.
1053 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1054 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1055 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1057 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1060 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1063 2020-01-24 Jeff Law <law@redhat.com>
1065 PR tree-optimization/92788
1066 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1069 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1072 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1073 *avx_vperm_broadcast_<mode>,
1074 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1075 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1076 Move before avx2_perm<mode>/avx512f_perm<mode>.
1079 * simplify-rtx.c (simplify_const_unary_operation,
1080 simplify_const_binary_operation): Punt for mode precision above
1081 MAX_BITSIZE_MODE_ANY_INT.
1083 2020-01-24 Andrew Pinski <apinski@marvell.com>
1085 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1088 2020-01-24 Jeff Law <law@redhat.com>
1091 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1092 for REGs. Call output_operand_lossage to get more reasonable
1095 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1097 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1098 gcn_fp_compare_operator.
1099 (vec_cmpu<mode>di): Use gcn_compare_operator.
1100 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1101 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1102 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1103 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1104 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1105 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1106 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1107 gcn_fp_compare_operator.
1108 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1109 gcn_fp_compare_operator.
1110 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1111 gcn_fp_compare_operator.
1112 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1113 gcn_fp_compare_operator.
1115 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1117 * doc/install.texi (Cross-Compiler-Specific Options): Document
1118 `--with-toolexeclibdir' option.
1120 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1122 * target.def (flags_regnum): Also mention effect on delay slot filling.
1123 * doc/tm.texi: Regenerate.
1125 2020-01-23 Jeff Law <law@redhat.com>
1127 PR translation/90162
1128 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1130 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1133 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1136 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1138 PR rtl-optimization/93402
1139 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1142 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1144 * config.in: Regenerated.
1145 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1146 for TARGET_LIBC_GNUSTACK.
1147 * configure: Regenerated.
1148 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1149 found to be 2.31 or greater.
1151 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1153 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1155 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1156 (mips_asm_file_end): New function. Delegate to
1157 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1158 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1160 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1163 * config/i386/i386-modes.def (POImode): New mode.
1164 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1165 * config/i386/i386.md (DPWI): New mode attribute.
1166 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1168 (QPWI): ... this. Use POI instead of OI for TImode.
1169 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1170 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1173 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1176 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1178 (speculation_tracker_rev): New pattern.
1179 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1180 Use speculation_tracker_rev to track the inverse condition.
1182 2020-01-23 Richard Biener <rguenther@suse.de>
1184 PR tree-optimization/93381
1185 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1186 alias-set of the def as argument and record the first one.
1187 (vn_walk_cb_data::first_set): New member.
1188 (vn_reference_lookup_3): Pass the alias-set of the current def
1189 to push_partial_def. Fix alias-set used in the aggregate copy
1191 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1192 * real.c (clear_significand_below): Fix out-of-bound access.
1194 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1197 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1198 New define_insn patterns.
1200 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1202 * doc/sourcebuild.texi (check-function-bodies): Add an
1203 optional target/xfail selector.
1205 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1207 PR rtl-optimization/93124
1208 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1209 bare USE and CLOBBER insns.
1211 2020-01-22 Andrew Pinski <apinski@marvell.com>
1213 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1215 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1218 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1219 diagnostic_manager into "ana" namespace.
1220 * selftest-run-tests.c (selftest::run_tests): Update for move of
1221 selftest::run_analyzer_selftests to
1222 ana::selftest::run_analyzer_selftests.
1224 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1226 * cfgexpand.c (union_stack_vars): Update the size.
1228 2020-01-22 Richard Biener <rguenther@suse.de>
1230 PR tree-optimization/93381
1231 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1232 throughout, handle all conversions the same.
1234 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1237 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1238 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1239 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1240 Call force_reg on high_in2 unconditionally.
1242 2020-01-22 Martin Liska <mliska@suse.cz>
1244 PR tree-optimization/92924
1245 * profile.c (compute_value_histograms): Divide
1248 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1251 * output.h (assemble_name_resolve): Declare.
1252 * varasm.c (assemble_name_resolve): New function.
1253 (assemble_name): Use it.
1254 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1256 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1258 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1259 update_web_docs_git instead of update_web_docs_svn.
1261 2020-01-21 Andrew Pinski <apinski@marvell.com>
1264 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1265 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1266 (*tlsgd_small_<mode>): Likewise.
1267 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1268 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1269 register. Convert that register back to dest using convert_mode.
1271 2020-01-21 Jim Wilson <jimw@sifive.com>
1273 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1276 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1277 Uros Bizjak <ubizjak@gmail.com>
1280 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1282 (legitimize_tls_address): Do GNU2 TLS address computation in
1283 ptr_mode and zero-extend result to Pmode.
1284 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1285 :P with :PTR and Pmode with ptr_mode.
1286 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1287 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1288 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1290 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1293 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1294 the last two operands are CONST_INT_P before using them as such.
1296 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1298 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1299 to get the integer element types.
1301 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1303 * config/aarch64/aarch64-sve-builtins.h
1304 (function_expander::convert_to_pmode): Declare.
1305 * config/aarch64/aarch64-sve-builtins.cc
1306 (function_expander::convert_to_pmode): New function.
1307 (function_expander::get_contiguous_base): Use it.
1308 (function_expander::prepare_gather_address_operands): Likewise.
1309 * config/aarch64/aarch64-sve-builtins-sve2.cc
1310 (svwhilerw_svwhilewr_impl::expand): Likewise.
1312 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1315 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1316 cfun->machine->label_is_assembled.
1317 (aarch64_print_patchable_function_entry): New.
1318 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1319 * config/aarch64/aarch64.h (struct machine_function): New field,
1322 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1325 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1328 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1331 * cgraph.c (cgraph_edge::resolve_speculation,
1332 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1333 call_stmt_site_hash.
1335 2020-01-21 Martin Liska <mliska@suse.cz>
1337 * config/rs6000/rs6000.c (common_mode_defined): Remove
1340 2020-01-21 Richard Biener <rguenther@suse.de>
1342 PR tree-optimization/92328
1343 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1344 type when value-numbering same-sized store by inserting a
1346 (eliminate_dom_walker::eliminate_stmt): When eliminating
1347 a redundant store handle bit-reinterpretation of the same value.
1349 2020-01-21 Andrew Pinski <apinski@marvel.com>
1352 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1354 (prepare_block_for_update): This. Use a worklist instead of
1357 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1359 * gcc/config/arm/arm.c (clear_operation_p):
1360 Initialise last_regno, skip first iteration
1361 based on the first_set value and use ints instead
1362 of the unnecessary HOST_WIDE_INTs.
1364 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1367 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1368 compare_mode other than SFmode or DFmode.
1370 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1373 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1374 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1375 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1377 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1379 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1381 2020-01-20 Andrew Pinski <apinski@marvell.com>
1384 * targhooks.c (default_print_patchable_function_entry): Use
1385 output_asm_insn to emit the nop instruction.
1387 2020-01-20 Fangrui Song <maskray@google.com>
1390 * targhooks.c (default_print_patchable_function_entry): Align to
1393 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1396 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1397 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1398 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1399 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1400 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1401 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1402 Remove the {q} suffix from lea.
1403 (*tls_dynamic_gnu2_call_64): Renamed to ...
1404 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1405 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1406 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1407 Pass Pmode to gen_tls_dynamic_gnu2_64.
1409 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1411 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1413 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1415 * config/aarch64/aarch64-sve-builtins-base.cc
1416 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1418 2020-01-20 Richard Biener <rguenther@suse.de>
1421 * dwarf2out.c (prune_unused_types): Unconditionally mark
1422 called function DIEs.
1424 2020-01-20 Martin Liska <mliska@suse.cz>
1426 PR tree-optimization/93199
1427 * tree-eh.c (struct leh_state): Add
1428 new field outer_non_cleanup.
1429 (cleanup_is_dead_in): Pass leh_state instead
1430 of eh_region. Add a checking that state->outer_non_cleanup
1431 points to outer non-clean up region.
1432 (lower_try_finally): Record outer_non_cleanup
1434 (lower_catch): Likewise.
1435 (lower_eh_filter): Likewise.
1436 (lower_eh_must_not_throw): Likewise.
1437 (lower_cleanup): Likewise.
1439 2020-01-20 Richard Biener <rguenther@suse.de>
1441 PR tree-optimization/93094
1442 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1443 (vect_transform_loop): Likewise.
1444 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1445 loop_vectorized_call to vect_transform_loop.
1446 * tree-vect-loop.c (vect_transform_loop): Pass down
1447 loop_vectorized_call to vect_loop_versioning.
1448 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1449 the earlier discovered loop_vectorized_call.
1451 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1453 * doc/contribute.texi: Update for SVN -> Git transition.
1454 * doc/install.texi: Likewise.
1456 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1459 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1460 speculative targets.
1461 (verify_speculative_call): New function
1462 (cgraph_node::verify_node): Use it.
1463 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1466 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1469 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1470 (cgraph_edge::make_direct): Remove all indirect targets.
1471 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1472 (cgraph_node::verify_node): Verify that only one call_stmt or
1473 lto_stmt_uid is set.
1474 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1476 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1477 (lto_output_ref): Simplify streaming of stmt.
1478 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1480 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1482 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1483 Mark parameter unused.
1485 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1487 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1489 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1491 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1493 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1495 * Makefile.in: Add coroutine-passes.o.
1496 * builtin-types.def (BT_CONST_SIZE): New.
1497 (BT_FN_BOOL_PTR): New.
1498 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1499 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1500 * coroutine-builtins.def: New file.
1501 * coroutine-passes.cc: New file.
1502 * function.h (struct GTY function): Add a bit to indicate that the
1503 function is a coroutine component.
1504 * internal-fn.c (expand_CO_FRAME): New.
1505 (expand_CO_YIELD): New.
1506 (expand_CO_SUSPN): New.
1507 (expand_CO_ACTOR): New.
1508 * internal-fn.def (CO_ACTOR): New.
1512 * passes.def: Add pass_coroutine_lower_builtins,
1513 pass_coroutine_early_expand_ifns.
1514 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1515 (make_pass_coroutine_early_expand_ifns): New.
1516 * doc/invoke.texi: Document the fcoroutines command line
1519 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1521 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1524 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1525 after checking the argument is a REG. Don't use REGNO (reg)
1526 again to set last_regno, reuse regno variable instead.
1528 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1530 * doc/analyzer.texi (Limitations): Add note about NaN.
1532 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1533 Sudakshina Das <sudi.das@arm.com>
1535 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1536 and valid immediate.
1537 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1538 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1539 * config/arm/constraints.md (Pg): New.
1540 * config/arm/predicates.md (long_shift_imm): New.
1541 (arm_reg_or_long_shift_imm): Likewise.
1542 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1543 (thumb2_lsll): Likewise.
1546 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1547 Sudakshina Das <sudi.das@arm.com>
1549 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1550 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1551 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1552 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1553 * config/arm/thumb2.md (thumb2_asrl): New.
1554 (thumb2_lsll): Likewise.
1556 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1558 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1561 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1563 * gdbinit.in (help-gcc-hooks): New command.
1564 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1565 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1568 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1570 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1571 correct target macro.
1573 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1575 * config/aarch64/aarch64-protos.h
1576 (aarch64_sve_ld1ro_operand_p): New.
1577 * config/aarch64/aarch64-sve-builtins-base.cc
1578 (class load_replicate): New.
1579 (class svld1ro_impl): New.
1580 (class svld1rq_impl): Change to inherit from load_replicate.
1581 (svld1ro): New sve intrinsic function base.
1582 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1583 New DEF_SVE_FUNCTION.
1584 * config/aarch64/aarch64-sve-builtins-base.h
1585 (svld1ro): New decl.
1586 * config/aarch64/aarch64-sve-builtins.cc
1587 (function_expander::add_mem_operand): Modify assert to allow
1589 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1591 * config/aarch64/aarch64.c
1592 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1593 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1594 (aarch64_sve_ld1ro_operand_p): New.
1595 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1596 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1597 * config/aarch64/predicates.md
1598 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1600 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1602 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1603 Introduce this ACLE specified predefined macro.
1604 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1605 (fp): Disabling this disables f64mm.
1606 (simd): Disabling this disables f64mm.
1607 (fp16): Disabling this disables f64mm.
1608 (sve): Disabling this disables f64mm.
1609 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1610 (AARCH64_ISA_F64MM): New.
1611 (TARGET_F64MM): New.
1612 * doc/invoke.texi (f64mm): Document new option.
1614 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1616 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1617 (neoversen1_tunings): Likewise.
1619 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1622 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1623 Add assert to ensure prolog has been emitted.
1624 (aarch64_split_atomic_op): Likewise.
1625 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1626 Use epilogue_completed rather than reload_completed.
1627 (aarch64_atomic_exchange<mode>): Likewise.
1628 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1629 (atomic_nand<mode>): Likewise.
1630 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1631 (atomic_fetch_nand<mode>): Likewise.
1632 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1633 (atomic_nand_fetch<mode>): Likewise.
1635 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1638 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1640 (REVERSE_CONDITION): Delete.
1641 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1642 (CCFP_CCFPE): Likewise.
1643 (e): New mode attribute.
1644 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1645 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1646 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1647 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1648 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1649 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1650 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1651 name of generator from gen_ccmpdi to gen_ccmpccdi.
1652 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1653 the previous comparison but aren't able to, use the new ccmp_rev
1656 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1658 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1659 than testing directly for INTEGER_CST.
1660 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1662 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1664 PR tree-optimization/93292
1665 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1666 get_vectype_for_scalar_type returns NULL.
1668 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1670 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1671 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1673 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1675 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1677 * params.opt: (max-predicted-iterations): Set bounds.
1678 * predict.c (real_almost_one, real_br_prob_base,
1679 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1680 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1681 probabilities; do not truncate to reg_br_prob_bases.
1682 (estimate_loops_at_level): Pass max_cyclic_prob.
1683 (estimate_loops): Compute max_cyclic_prob.
1684 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1686 * profile-count.c (profile_probability::to_sreal): New.
1687 * profile-count.h (class sreal): Move up in file.
1688 (profile_probability::to_sreal): Declare.
1690 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1693 (arm_invalid_conversion): New function for target hook.
1694 (arm_invalid_unary_op): New function for target hook.
1695 (arm_invalid_binary_op): New function for target hook.
1697 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1699 * config.gcc: Add arm_bf16.h.
1700 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1701 (arm_simd_builtin_std_type): Add BFmode.
1702 (arm_init_simd_builtin_types): Define element types for vector types.
1703 (arm_init_bf16_types): New function.
1704 (arm_init_builtins): Add arm_init_bf16_types function call.
1705 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1706 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1707 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1708 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1709 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1710 (arm_mangle_type): Add __bf16.
1711 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1712 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1713 arm_bf16_ptr_type_node.
1714 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1715 define_split between ARM registers.
1716 * config/arm/arm_bf16.h: New file.
1717 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1718 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1720 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1721 * config/arm/vfp.md: Add BFmode to movhf patterns.
1723 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1724 Andre Vieira <andre.simoesdiasvieira@arm.com>
1726 * config/arm/arm-cpus.in (mve, mve_float): New features.
1727 (dsp, mve, mve.fp): New options.
1728 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1729 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1730 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1732 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1733 Thomas Preud'homme <thomas.preudhomme@arm.com>
1735 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1737 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1738 error for using -mcmse when targeting Armv8.1-M Mainline.
1740 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1741 Thomas Preud'homme <thomas.preudhomme@arm.com>
1743 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1744 address in r4 when targeting Armv8.1-M Mainline.
1745 (nonsecure_call_value_internal): Likewise.
1746 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1747 a register match_operand again. Emit BLXNS when targeting
1749 (nonsecure_call_value_reg_thumb2): Likewise.
1751 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1752 Thomas Preud'homme <thomas.preudhomme@arm.com>
1754 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1755 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1756 variable as true when floating-point ABI is not hard. Replace
1757 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1758 Generate VLSTM and VLLDM instruction respectively before and
1759 after a function call to cmse_nonsecure_call function.
1760 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1761 (VUNSPEC_VLLDM): Likewise.
1762 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1763 (lazy_load_multiple_insn): Likewise.
1765 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1766 Thomas Preud'homme <thomas.preudhomme@arm.com>
1768 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1769 (arm_emit_vfp_multi_reg_pop): Likewise.
1770 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1771 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1772 restore callee-saved VFP registers.
1774 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1775 Thomas Preud'homme <thomas.preudhomme@arm.com>
1777 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1778 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1779 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1780 callee-saved GPRs as well as clear ip register before doing a nonsecure
1781 call then restore callee-saved GPRs after it when targeting
1783 (arm_reorg): Adapt to function rename.
1785 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1786 Thomas Preud'homme <thomas.preudhomme@arm.com>
1788 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1789 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1790 clear_vfp_multiple pattern based on a new vfp parameter.
1791 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1792 targeting Armv8.1-M Mainline.
1793 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1794 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1795 whether VFP registers are available before looking call_used_regs for a
1797 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1798 of prototype of clear_operation_p.
1799 (clear_vfp_multiple_operation): New predicate.
1800 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1801 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1803 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1804 Thomas Preud'homme <thomas.preudhomme@arm.com>
1806 * config/arm/arm-protos.h (clear_operation_p): Declare.
1807 * config/arm/arm.c (clear_operation_p): New function.
1808 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1809 targeting Armv8.1-M Mainline or successor.
1810 (output_return_instruction): Only output APSR register clearing if
1811 Armv8.1-M Mainline instructions not available.
1812 (thumb_exit): Likewise.
1813 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1814 * config/arm/thumb2.md (clear_apsr): New define_insn.
1815 (clear_multiple): Likewise.
1816 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1818 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1819 Thomas Preud'homme <thomas.preudhomme@arm.com>
1821 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1822 (use_return_insn): Also return false for Armv8.1-M Mainline.
1823 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1824 Mainline instructions are available.
1825 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1826 when targeting Armv8.1-M Mainline Security Extensions.
1827 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1828 Mainline entry function.
1829 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1830 targeting Armv8.1-M Mainline or successor.
1831 (arm_expand_epilogue): Fix indentation of caller-saved register
1832 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1834 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1835 (FP_SYSREGS): Likewise.
1836 (enum vfp_sysregs_encoding): Define enum.
1837 (fp_sysreg_names): Declare.
1838 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1839 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1840 (pop_fpsysreg_insn): Likewise.
1842 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1843 Thomas Preud'homme <thomas.preudhomme@arm.com>
1845 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1846 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1847 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1848 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1849 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1850 (ARMv8_1m_main): New feature group.
1851 (armv8.1-m.main): New architecture.
1852 * config/arm/arm-tables.opt: Regenerate.
1853 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1854 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1855 (arm_options_perform_arch_sanity_checks): Error out when targeting
1856 Armv8.1-M Mainline Security Extensions.
1857 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1859 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1861 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1862 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1863 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1864 aarch64_bfdot_laneq): New.
1865 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1866 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1867 vbfdotq_laneq_f32): New.
1868 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1869 VBFMLA_W, VBF): New.
1870 (isquadop): Add V4BF, V8BF.
1872 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1874 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1875 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1876 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1877 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1878 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1879 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1880 usdot_laneq, sudot_lane,sudot_laneq): New.
1881 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1882 (aarch64_<sur>dot_lane): New.
1883 * config/aarch64/arm_neon.h (vusdot_s32): New.
1885 (vusdot_lane_s32): New.
1886 (vsudot_lane_s32): New.
1887 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1888 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1890 2020-01-16 Martin Liska <mliska@suse.cz>
1892 * value-prof.c (dump_histogram_value): Fix
1893 obvious spacing issue.
1895 2020-01-16 Andrew Pinski <apinski@marvell.com>
1897 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1898 !storage_order_barrier_p.
1900 2020-01-16 Andrew Pinski <apinski@marvell.com>
1902 * sched-int.h (_dep): Add unused bit-field field for the padding.
1903 * sched-deps.c (init_dep_1): Init unused field.
1905 2020-01-16 Andrew Pinski <apinski@marvell.com>
1907 * optabs.h (create_expand_operand): Initialize target field also.
1909 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
1911 PR tree-optimization/92429
1912 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1913 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1915 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1918 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
1920 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1921 aarch64_sve_int_mode to each mode.
1923 2020-01-15 David Malcolm <dmalcolm@redhat.com>
1925 * doc/analyzer.texi (Overview): Add note about
1926 -fdump-ipa-analyzer.
1928 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
1930 PR tree-optimization/93231
1931 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
1932 input_type is unsigned. Use tree_to_shwi for shift constant.
1933 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
1934 (simplify_count_trailing_zeroes): Add test to handle known non-zero
1935 inputs more efficiently.
1937 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
1939 * config/i386/i386.md (*movsf_internal): Do not require
1940 SSE2 ISA for alternatives 14 and 15.
1942 2020-01-15 Richard Biener <rguenther@suse.de>
1945 * tree-eh.c (sink_clobbers): If we already visited the destination
1946 block do not defer insertion.
1947 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
1948 the purpose of defered insertion.
1950 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1952 * BASE-VER: Bump to 10.0.1.
1954 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1956 PR tree-optimization/93247
1957 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
1958 type of the stmt that we're going to vectorize.
1960 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1962 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
1963 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
1966 2020-01-15 Martin Liska <mliska@suse.cz>
1968 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
1969 2 calls of streamer_read_hwi in a function call.
1971 2020-01-15 Richard Biener <rguenther@suse.de>
1973 * alias.c (record_alias_subset): Avoid redundant work when
1974 subset is already recorded.
1976 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1978 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
1979 the analyzer options provide CWE identifiers.
1981 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1983 * tree-diagnostic-path.cc (path_summary::event_range::print):
1984 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
1985 using get_pure_location.
1987 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1989 PR tree-optimization/93262
1990 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
1991 perform head trimming only if the last argument is constant,
1992 either all ones, or larger or equal to head trim, in the latter
1993 case decrease the last argument by head_trim.
1995 PR tree-optimization/93249
1996 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
1997 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
1998 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
1999 perform head trim unless we can prove there are no '\0' chars
2000 from the source among the first head_trim chars.
2002 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2004 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2006 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2009 * config/i386/sse.md
2010 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2011 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2012 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2013 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2014 just a single alternative instead of two, make operands 1 and 2
2017 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2020 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2023 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2025 * Makefile.in (lang_opt_files): Add analyzer.opt.
2026 (ANALYZER_OBJS): New.
2027 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2028 tristate.o and ANALYZER_OBJS.
2029 (TEXI_GCCINT_FILES): Add analyzer.texi.
2030 * common.opt (-fanalyzer): New driver option.
2031 * config.in: Regenerate.
2032 * configure: Regenerate.
2033 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2034 (gccdepdir): Also create depdir for "analyzer" subdir.
2035 * digraph.cc: New file.
2036 * digraph.h: New file.
2037 * doc/analyzer.texi: New file.
2038 * doc/gccint.texi ("Static Analyzer") New menu item.
2039 (analyzer.texi): Include it.
2040 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2041 ("Warning Options"): Add static analysis warnings to the list.
2042 (-Wno-analyzer-double-fclose): New option.
2043 (-Wno-analyzer-double-free): New option.
2044 (-Wno-analyzer-exposure-through-output-file): New option.
2045 (-Wno-analyzer-file-leak): New option.
2046 (-Wno-analyzer-free-of-non-heap): New option.
2047 (-Wno-analyzer-malloc-leak): New option.
2048 (-Wno-analyzer-possible-null-argument): New option.
2049 (-Wno-analyzer-possible-null-dereference): New option.
2050 (-Wno-analyzer-null-argument): New option.
2051 (-Wno-analyzer-null-dereference): New option.
2052 (-Wno-analyzer-stale-setjmp-buffer): New option.
2053 (-Wno-analyzer-tainted-array-index): New option.
2054 (-Wno-analyzer-use-after-free): New option.
2055 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2056 (-Wno-analyzer-use-of-uninitialized-value): New option.
2057 (-Wanalyzer-too-complex): New option.
2058 (-fanalyzer-call-summaries): New warning.
2059 (-fanalyzer-checker=): New warning.
2060 (-fanalyzer-fine-grained): New warning.
2061 (-fno-analyzer-state-merge): New warning.
2062 (-fno-analyzer-state-purge): New warning.
2063 (-fanalyzer-transitivity): New warning.
2064 (-fanalyzer-verbose-edges): New warning.
2065 (-fanalyzer-verbose-state-changes): New warning.
2066 (-fanalyzer-verbosity=): New warning.
2067 (-fdump-analyzer): New warning.
2068 (-fdump-analyzer-callgraph): New warning.
2069 (-fdump-analyzer-exploded-graph): New warning.
2070 (-fdump-analyzer-exploded-nodes): New warning.
2071 (-fdump-analyzer-exploded-nodes-2): New warning.
2072 (-fdump-analyzer-exploded-nodes-3): New warning.
2073 (-fdump-analyzer-supergraph): New warning.
2074 * doc/sourcebuild.texi (dg-require-dot): New.
2075 (dg-check-dot): New.
2076 * gdbinit.in (break-on-saved-diagnostic): New command.
2077 * graphviz.cc: New file.
2078 * graphviz.h: New file.
2079 * ordered-hash-map-tests.cc: New file.
2080 * ordered-hash-map.h: New file.
2081 * passes.def (pass_analyzer): Add before
2082 pass_ipa_whole_program_visibility.
2083 * selftest-run-tests.c (selftest::run_tests): Call
2084 selftest::ordered_hash_map_tests_cc_tests.
2085 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2087 * shortest-paths.h: New file.
2088 * timevar.def (TV_ANALYZER): New timevar.
2089 (TV_ANALYZER_SUPERGRAPH): Likewise.
2090 (TV_ANALYZER_STATE_PURGE): Likewise.
2091 (TV_ANALYZER_PLAN): Likewise.
2092 (TV_ANALYZER_SCC): Likewise.
2093 (TV_ANALYZER_WORKLIST): Likewise.
2094 (TV_ANALYZER_DUMP): Likewise.
2095 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2096 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2097 * tree-pass.h (make_pass_analyzer): New decl.
2098 * tristate.cc: New file.
2099 * tristate.h: New file.
2101 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2104 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2105 alternatives 9 and 10.
2107 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2109 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2110 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2111 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2112 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2113 (selftest::hash_map_tests_c_tests): Call it.
2114 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2115 New static constant, using the value of = H::empty_zero_p.
2116 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2117 from default_hash_traits <Value>.
2118 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2120 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2121 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2122 calls to mark_empty with !Descriptor::empty_zero_p.
2123 (hash_table::empty_slow): Conditionalize the memset call with a
2124 check that Descriptor::empty_zero_p; otherwise, loop through the
2125 entries calling mark_empty on them.
2126 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2127 (pointer_hash::empty_zero_p): Likewise.
2128 (pair_hash::empty_zero_p): Likewise.
2129 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2131 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2132 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2133 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2134 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2135 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2136 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2137 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2138 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2140 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2143 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2145 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2146 fix typo on return value.
2148 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2151 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2153 (cgraph_edge::make_speculative): Add param for setting speculative_id
2155 (cgraph_edge::speculative_call_info): Update comments and find reference
2156 by speculative_id for multiple indirect targets.
2157 (cgraph_edge::resolve_speculation): Decrease the speculations
2158 for indirect edge, drop it's speculative if not direct target
2159 left. Update comments.
2160 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2161 (cgraph_node::dump): Print num_speculative_call_targets.
2162 (cgraph_node::verify_node): Don't report error if speculative
2163 edge not include statement.
2164 (cgraph_edge::num_speculative_call_targets_p): New function.
2165 * cgraph.h (int common_target_id): Remove.
2166 (int common_target_probability): Remove.
2167 (num_speculative_call_targets): New variable.
2168 (make_speculative): Add param for setting speculative_id.
2169 (cgraph_edge::num_speculative_call_targets_p): New declare.
2170 (target_prob): New variable.
2171 (speculative_id): New variable.
2172 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2173 call summaries for multiple speculative call targets.
2174 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2175 * ipa-profile.c (struct speculative_call_target): New struct.
2176 (class speculative_call_summary): New class.
2177 (class speculative_call_summaries): New class.
2178 (call_sums): New variable.
2179 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2180 (ipa_profile_write_edge_summary): New function.
2181 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2182 (ipa_profile_dump_all_summaries): New function.
2183 (ipa_profile_read_edge_summary): New function.
2184 (ipa_profile_read_summary_section): New function.
2185 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2186 (ipa_profile): Generate num_speculative_call_targets from
2188 * ipa-ref.h (speculative_id): New variable.
2189 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2190 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2191 common_target_probability. Stream out speculative_id and
2192 num_speculative_call_targets.
2193 (input_edge): Likewise.
2194 * predict.c (dump_prediction): Remove edges count assert to be
2196 * symtab.c (symtab_node::create_reference): Init speculative_id.
2197 (symtab_node::clone_references): Clone speculative_id.
2198 (symtab_node::clone_referring): Clone speculative_id.
2199 (symtab_node::clone_reference): Clone speculative_id.
2200 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2201 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2202 if indirect call contains multiple speculative targets.
2203 * value-prof.h (check_ic_target): Remove.
2204 * value-prof.c (gimple_value_profile_transformations):
2205 Use void function gimple_ic_transform.
2206 * value-prof.c (gimple_ic_transform): Handle topn case.
2207 Fix comment typos. Change it to a void function.
2209 2020-01-13 Andrew Pinski <apinski@marvell.com>
2211 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2212 (octeontx2t98): New define.
2213 (octeontx2t96): New define.
2214 (octeontx2t93): New define.
2215 (octeontx2f95): New define.
2216 (octeontx2f95n): New define.
2217 (octeontx2f95mm): New define.
2218 * config/aarch64/aarch64-tune.md: Regenerate.
2219 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2221 2020-01-13 Jason Merrill <jason@redhat.com>
2223 PR c++/33799 - destroy return value if local cleanup throws.
2224 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2226 2020-01-13 Martin Liska <mliska@suse.cz>
2228 * ipa-cp.c (get_max_overall_size): Use newly
2229 renamed param param_ipa_cp_unit_growth.
2230 * params.opt: Remove legacy param name.
2232 2020-01-13 Martin Sebor <msebor@redhat.com>
2234 PR tree-optimization/93213
2235 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2236 stores to be eliminated.
2238 2020-01-13 Martin Liska <mliska@suse.cz>
2240 * opts.c (print_help): Do not print CL_PARAM
2241 and CL_WARNING for CL_OPTIMIZATION.
2243 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2246 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2247 depending on optimization settings.
2249 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2251 PR tree-optimization/90838
2252 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2253 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2254 argument rather than to initialize temporary for targets that
2255 don't use the mode argument at all. Initialize ctzval to avoid
2258 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2260 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2261 * tree-core.h: Document it.
2262 * gimplify.c (gimplify_omp_workshare): Set it.
2263 * omp-low.c (lower_omp_target): Use it.
2264 * tree-pretty-print.c (dump_omp_clause): Print it.
2266 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2267 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2269 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2271 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2272 * common.opt (fdiagnostics-path-format=): New option.
2273 (diagnostic_path_format): New enum.
2274 (fdiagnostics-show-path-depths): New option.
2275 * coretypes.h (diagnostic_event_id_t): New forward decl.
2276 * diagnostic-color.c (color_dict): Add "path".
2277 * diagnostic-event-id.h: New file.
2278 * diagnostic-format-json.cc (json_from_expanded_location): Make
2280 (json_end_diagnostic): Call context->make_json_for_path if it
2281 exists and the diagnostic has a path.
2282 (diagnostic_output_format_init): Clear context->print_path.
2283 * diagnostic-path.h: New file.
2284 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2285 when printing a run of events in a diagnostic_path so that they
2286 all get the same color.
2287 (layout::m_diagnostic_path_p): New field.
2288 (layout::layout): Initialize it.
2289 (layout::print_any_labels): Don't colorize the label text for an
2290 event in a diagnostic_path.
2291 (gcc_rich_location::add_location_if_nearby): Add
2292 "restrict_to_current_line_spans" and "label" params. Pass the
2293 former to layout.maybe_add_location_range; pass the latter
2294 when calling add_range.
2295 * diagnostic.c: Include "diagnostic-path.h".
2296 (diagnostic_initialize): Initialize context->path_format and
2297 context->show_path_depths.
2298 (diagnostic_show_any_path): New function.
2299 (diagnostic_path::interprocedural_p): New function.
2300 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2301 (simple_diagnostic_path::num_events): New function.
2302 (simple_diagnostic_path::get_event): New function.
2303 (simple_diagnostic_path::add_event): New function.
2304 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2305 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2306 (debug): New overload taking a diagnostic_path *.
2307 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2308 * diagnostic.h (enum diagnostic_path_format): New enum.
2309 (json::value): New forward decl.
2310 (diagnostic_context::path_format): New field.
2311 (diagnostic_context::show_path_depths): New field.
2312 (diagnostic_context::print_path): New callback field.
2313 (diagnostic_context::make_json_for_path): New callback field.
2314 (diagnostic_show_any_path): New decl.
2315 (json_from_expanded_location): New decl.
2316 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2317 (-fdiagnostics-show-path-depths): New option.
2318 (-fdiagnostics-color): Add "path" to description of default
2319 GCC_COLORS; describe it.
2320 (-fdiagnostics-format=json): Document how diagnostic paths are
2321 represented in the JSON output format.
2322 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2323 Add optional params "restrict_to_current_line_spans" and "label".
2324 * opts.c (common_handle_option): Handle
2325 OPT_fdiagnostics_path_format_ and
2326 OPT_fdiagnostics_show_path_depths.
2327 * pretty-print.c: Include "diagnostic-event-id.h".
2328 (pp_format): Implement "%@" format code for printing
2329 diagnostic_event_id_t *.
2330 (selftest::test_pp_format): Add tests for "%@".
2331 * selftest-run-tests.c (selftest::run_tests): Call
2332 selftest::tree_diagnostic_path_cc_tests.
2333 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2334 * toplev.c (general_init): Initialize global_dc->path_format and
2335 global_dc->show_path_depths.
2336 * tree-diagnostic-path.cc: New file.
2337 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2338 non-static. Drop "diagnostic" param in favor of storing the
2339 original value of "where" and re-using it.
2340 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2341 maybe_unwind_expanded_macro_loc.
2342 (tree_diagnostics_defaults): Initialize context->print_path and
2343 context->make_json_for_path.
2344 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2346 (default_tree_make_json_for_path): New decl.
2347 (maybe_unwind_expanded_macro_loc): New decl.
2349 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2351 PR tree-optimization/93210
2352 * fold-const.h (native_encode_initializer,
2353 can_native_interpret_type_p): Declare.
2354 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2356 (native_encode_initializer): New function, moved from dwarf2out.c.
2357 Adjust to native_encode_expr compatible arguments, including dry-run
2358 and partial extraction modes. Don't handle STRING_CST.
2359 (can_native_interpret_type_p): No longer static.
2360 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2361 offset / BITS_PER_UNIT fits into int and don't call it if
2362 can_native_interpret_type_p fails. If suboff is NULL and for
2363 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2364 native_encode_initializer.
2365 (fold_const_aggregate_ref_1): Formatting fix.
2366 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2367 (tree_add_const_value_attribute): Adjust caller.
2369 PR tree-optimization/90838
2370 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2371 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2372 CTZ_DEFINED_VALUE_AT_ZERO.
2374 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2377 * lra-constraints.c (match_reload): Permit input operands have the
2378 same mode as output while other input operands have a different
2381 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2383 PR tree-optimization/90838
2384 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2385 (check_ctz_string): Likewise.
2386 (optimize_count_trailing_zeroes): Likewise.
2387 (simplify_count_trailing_zeroes): Likewise.
2388 (pass_forwprop::execute): Try ctz simplification.
2389 * match.pd: Add matching for ctz idioms.
2391 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2393 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2395 (aarch64_invalid_unary_op): New function for target hook.
2396 (aarch64_invalid_binary_op): New function for target hook.
2398 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2400 * config.gcc: Add arm_bf16.h.
2401 * config/aarch64/aarch64-builtins.c
2402 (aarch64_simd_builtin_std_type): Add BFmode.
2403 (aarch64_init_simd_builtin_types): Define element types for vector
2405 (aarch64_init_bf16_types): New function.
2406 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2407 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2409 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2410 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2412 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2413 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2414 * config/aarch64/aarch64.c
2415 (aarch64_classify_vector_mode): Add support for BF types.
2416 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2417 (aarch64_vq_mode): Add support for BF types.
2418 (aarch64_simd_container_mode): Add support for BF types.
2419 (aarch64_mangle_type): Add support for BF scalar type.
2420 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2421 * config/aarch64/arm_bf16.h: New file.
2422 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2423 * config/aarch64/iterators.md: Add BF types to mode attributes.
2424 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2426 2020-01-10 Jason Merrill <jason@redhat.com>
2428 PR c++/93173 - incorrect tree sharing.
2429 * gimplify.c (copy_if_shared): No longer static.
2430 * gimplify.h: Declare it.
2432 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2434 * doc/invoke.texi (-msve-vector-bits=): Document that
2435 -msve-vector-bits=128 now generates VL-specific code for
2436 little-endian targets.
2437 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2438 build_vector_type_for_mode to construct the data vector types.
2439 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2440 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2441 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2442 for 128-bit vectors.
2444 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2446 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2449 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2451 * config/aarch64/aarch64-builtins.c
2452 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2453 rather than checking the number of elements and the element mode.
2455 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2457 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2458 get_related_vectype_for_scalar_type rather than build_vector_type
2459 to create the index type for a conditional reduction.
2461 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2463 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2464 for any type of gather or scatter, including strided accesses.
2466 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2468 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2471 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2473 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2475 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2476 parameter and its use to reset DR_OFFSET's.
2477 (vect_transform_loop): Remove orig_drs_init argument.
2478 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2479 member of dr_vec_info rather than the offset of the associated
2480 data_reference's innermost_loop_behavior.
2481 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2482 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2483 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2484 get_dr_vinfo_offset.
2485 (vectorizable_store): Likewise.
2486 (vectorizable_load): Likewise.
2488 2020-01-10 Richard Biener <rguenther@suse.de>
2490 * gimple-ssa-store-merging
2491 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2493 2020-01-10 Martin Liska <mliska@suse.cz>
2496 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2497 encapsulation that was there before r280040.
2499 2020-01-10 Richard Biener <rguenther@suse.de>
2502 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2503 sequences to avoid walking them again for secondary opportunities.
2504 (pass_lower_eh_dispatch::execute): Instead actually insert
2507 2020-01-10 Richard Biener <rguenther@suse.de>
2510 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2511 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2512 avoid quadraticness.
2514 2020-01-10 Martin Jambor <mjambor@suse.cz>
2516 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2517 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2518 to get param_ipa_sra_max_replacements.
2519 (param_splitting_across_edge): Pass the caller to
2520 pull_accesses_from_callee.
2522 2020-01-10 Martin Jambor <mjambor@suse.cz>
2524 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2525 * ipa-cp.c (max_new_size): Removed.
2526 (orig_overall_size): New variable.
2527 (get_max_overall_size): New function.
2528 (estimate_local_effects): Use it. Adjust dump.
2529 (decide_about_value): Likewise.
2530 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2531 orig_overall_size. Adjust dump.
2532 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2534 2020-01-10 Martin Jambor <mjambor@suse.cz>
2536 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2537 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2538 instead of param_ipa_max_agg_items.
2539 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2540 optimization info for the callee.
2542 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2544 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2545 markers if debug_inline_points is false.
2547 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2549 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2551 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2552 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2553 aarch64-sve-builtins-sve2.h.
2554 (aarch64-sve-builtins-sve2.o): New rule.
2555 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2556 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2557 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2558 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2559 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2560 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2562 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2563 instructions that are handled here.
2564 (@cond_asrd<mode>): Generalize to...
2565 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2566 (*cond_asrd<mode>_2): Generalize to...
2567 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2568 (*cond_asrd<mode>_z): Generalize to...
2569 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2570 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2571 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2572 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2573 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2575 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2576 (@aarch64_scatter_stnt<mode>): Likewise.
2577 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2578 (@aarch64_mul_lane_<mode>): Likewise.
2579 (@aarch64_sve_suqadd<mode>_const): Likewise.
2580 (*<sur>h<addsub><mode>): Generalize to...
2581 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2583 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2584 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2585 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2586 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2587 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2588 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2589 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2590 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2591 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2592 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2593 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2594 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2595 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2596 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2597 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2598 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2599 (@aarch64_sve2_xar<mode>): Likewise.
2600 (@aarch64_sve2_bcax<mode>): Likewise.
2601 (*aarch64_sve2_eor3<mode>): Rename to...
2602 (@aarch64_sve2_eor3<mode>): ...this.
2603 (@aarch64_sve2_bsl<mode>): New expander.
2604 (@aarch64_sve2_nbsl<mode>): Likewise.
2605 (@aarch64_sve2_bsl1n<mode>): Likewise.
2606 (@aarch64_sve2_bsl2n<mode>): Likewise.
2607 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2608 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2609 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2610 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2611 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2612 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2613 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2614 (<su>mull<bt><Vwide>): Generalize to...
2615 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2617 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2618 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2619 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2620 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2621 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2622 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2623 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2624 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2625 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2626 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2627 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2628 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2629 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2630 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2631 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2632 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2633 (<SHRNB:r>shrnb<mode>): Generalize to...
2634 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2636 (<SHRNT:r>shrnt<mode>): Generalize to...
2637 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2639 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2640 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2641 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2642 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2643 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2644 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2645 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2646 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2647 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2648 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2649 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2650 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2651 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2652 (@aarch64_sve2_cvtnt<mode>): Likewise.
2653 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2654 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2655 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2656 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2657 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2658 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2659 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2660 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2661 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2662 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2663 (@aarch64_sve2_pmul<mode>): Likewise.
2664 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2665 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2666 (@aarch64_sve2_tbl2<mode>): Likewise.
2667 (@aarch64_sve2_tbx<mode>): Likewise.
2668 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2669 (@aarch64_sve2_histcnt<mode>): Likewise.
2670 (@aarch64_sve2_histseg<mode>): Likewise.
2671 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2672 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2673 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2674 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2675 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2676 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2677 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2678 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2679 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2680 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2681 (SVE2_PMULL_PAIR_I): New mode iterators.
2682 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2683 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2684 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2685 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2686 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2687 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2688 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2689 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2690 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2691 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2692 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2693 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2694 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2695 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2696 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2697 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2698 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2699 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2700 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2701 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2702 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2703 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2704 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2705 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2706 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2707 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2708 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2709 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2710 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2711 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2712 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2714 (VNARROW, Ventype): New mode attributes.
2715 (Vewtype): Handle VNx2DI. Fix typo in comment.
2716 (VDOUBLE): New mode attribute.
2717 (sve_lane_con): Handle VNx8HI.
2718 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2719 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2720 (sve_int_op, sve_int_op_rev): Handle the above codes.
2721 (sve_pred_int_rhs2_operand): Likewise.
2722 (MULLBT, SHRNB, SHRNT): Delete.
2723 (SVE_INT_SHIFT_IMM): New int iterator.
2724 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2725 and UNSPEC_WHILEHS for TARGET_SVE2.
2726 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2727 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2728 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2729 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2730 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2731 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2732 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2733 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2734 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2735 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2736 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2737 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2738 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2739 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2740 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2741 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2742 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2743 (optab): Handle the new unspecs.
2744 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2746 (lr): Handle the new unspecs.
2748 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2749 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2750 (sve_int_qsub_op): New int attributes.
2751 (sve_fp_op, rot): Handle the new unspecs.
2752 * config/aarch64/aarch64-sve-builtins.h
2753 (function_resolver::require_matching_pointer_type): Declare.
2754 (function_resolver::resolve_unary): Add an optional boolean argument.
2755 (function_resolver::finish_opt_n_resolution): Add an optional
2756 type_suffix_index argument.
2757 (gimple_folder::redirect_call): Declare.
2758 (gimple_expander::prepare_gather_address_operands): Add an optional
2760 * config/aarch64/aarch64-sve-builtins.cc: Include
2761 aarch64-sve-builtins-sve2.h.
2762 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2763 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2764 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2765 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2766 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2767 (TYPES_s_unsigned): Likewise.
2768 (TYPES_s_integer): Use TYPES_s_unsigned.
2769 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2770 (TYPES_sd_integer): Use them.
2771 (TYPES_d_unsigned): New macro.
2772 (TYPES_d_integer): Use it.
2773 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2774 (TYPES_cvt_narrow): Likewise.
2775 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2776 (preds_mx): New variable.
2777 (function_builder::add_overloaded_function): Allow the new feature
2778 set to be more restrictive than the original one.
2779 (function_resolver::infer_pointer_type): Remove qualifiers from
2780 the pointer type before printing it.
2781 (function_resolver::require_matching_pointer_type): New function.
2782 (function_resolver::resolve_sv_displacement): Handle functions
2783 that don't support 32-bit vector indices or svint32_t vector offsets.
2784 (function_resolver::finish_opt_n_resolution): Take the inferred type
2785 as a separate argument.
2786 (function_resolver::resolve_unary): Optionally treat all forms in
2787 the same way as normal merging functions.
2788 (gimple_folder::redirect_call): New function.
2789 (function_expander::prepare_gather_address_operands): Add an argument
2790 that says whether scaled forms are available. If they aren't,
2791 handle scaling of vector indices and don't add the extension and
2793 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2794 fall back to using cond_* instead.
2795 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2796 Split out the member variables into...
2797 (rtx_code_function_base): ...this new base class.
2798 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2799 (unspec_based_function): Split out the member variables into...
2800 (unspec_based_function_base): ...this new base class.
2801 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2802 (unspec_based_function_exact_insn): New class.
2803 (unspec_based_add_function, unspec_based_add_lane_function)
2804 (unspec_based_lane_function, unspec_based_pred_function)
2805 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2806 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2807 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2809 (unspec_based_fused_function): New class.
2810 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2811 (unspec_based_fused_lane_function): New class.
2812 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2814 (CODE_FOR_MODE1): New macro.
2815 (fixed_insn_function): New class.
2816 (while_comparison): Likewise.
2817 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2818 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2819 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2820 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2821 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2822 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2823 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2824 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2825 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2826 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2827 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2828 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2829 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2830 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2831 Also add an initial argument for unary_convert_narrowt, regardless
2832 of the predication type.
2833 (build_32_64): Allow loads and stores to specify MODE_none.
2834 (build_sv_index64, build_sv_uint_offset): New functions.
2835 (long_type_suffix): New function.
2836 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2837 (binary_imm_long_base, load_gather_sv_base): Likewise.
2838 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2839 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2840 (unary_narrowb_base, unary_narrowt_base): Likewise.
2841 (binary_long_lane_def, binary_long_lane): New shape.
2842 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2843 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2844 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2845 (binary_to_uint_def, binary_to_uint): Likewise.
2846 (binary_wide_def, binary_wide): Likewise.
2847 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2848 (compare_def, compare): Likewise.
2849 (compare_ptr_def, compare_ptr): Likewise.
2850 (load_ext_gather_index_restricted_def,
2851 load_ext_gather_index_restricted): Likewise.
2852 (load_ext_gather_offset_restricted_def,
2853 load_ext_gather_offset_restricted): Likewise.
2854 (load_gather_sv_def): Inherit from load_gather_sv_base.
2855 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2856 (shift_left_imm_def, shift_left_imm): Likewise.
2857 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2858 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2859 (store_scatter_index_restricted_def,
2860 store_scatter_index_restricted): Likewise.
2861 (store_scatter_offset_restricted_def,
2862 store_scatter_offset_restricted): Likewise.
2863 (tbl_tuple_def, tbl_tuple): Likewise.
2864 (ternary_long_lane_def, ternary_long_lane): Likewise.
2865 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2866 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2867 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2868 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2869 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2870 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2871 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2872 (ternary_uint_def, ternary_uint): Likewise.
2873 (unary_convert): Fix typo in comment.
2874 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2875 (unary_long_def, unary_long): Likewise.
2876 (unary_narrowb_def, unary_narrowb): Likewise.
2877 (unary_narrowt_def, unary_narrowt): Likewise.
2878 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2879 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2880 (unary_to_int_def, unary_to_int): Likewise.
2881 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2882 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2883 (svasrd_impl): Delete.
2884 (svcadd_impl::expand): Handle integer operations too.
2885 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2886 new functions to derive the unspec numbers.
2887 (svmla_svmls_lane_impl): Replace with...
2888 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2889 integer operations too.
2890 (svwhile_impl): Rename to...
2891 (svwhilelx_impl): ...this and inherit from while_comparison.
2892 (svasrd): Use unspec_based_function.
2893 (svmla_lane): Use svmla_lane_impl.
2894 (svmls_lane): Use svmls_lane_impl.
2895 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2896 (svwhilele, svwhilelt): Use svwhilelx_impl.
2897 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2898 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2899 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2900 * config/aarch64/aarch64-sve-builtins.def: Include
2901 aarch64-sve-builtins-sve2.def.
2903 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2905 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2906 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2907 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2908 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
2909 immediates as well as vector ones.
2910 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2911 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2912 (aarch64_sve_qsub_immediate): Update calls accordingly.
2914 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2916 * config/aarch64/aarch64-sve2.md: Add banner comments.
2917 (<su>mulh<r>s<mode>3): Move further up file.
2918 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2919 (*aarch64_sve2_sra<mode>): Move further down file.
2920 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2922 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2924 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2926 (while_optab_cmp): Handle them.
2927 * config/aarch64/aarch64-sve.md
2928 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
2929 and add a "@" marker.
2930 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
2931 instead of gen_aarch64_sve2_while_ptest.
2932 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
2934 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2936 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
2937 (UNSPEC_WHILELE): ...this.
2938 (UNSPEC_WHILE_LO): Rename to...
2939 (UNSPEC_WHILELO): ...this.
2940 (UNSPEC_WHILE_LS): Rename to...
2941 (UNSPEC_WHILELS): ...this.
2942 (UNSPEC_WHILE_LT): Rename to...
2943 (UNSPEC_WHILELT): ...this.
2944 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
2945 (cmp_op, while_optab_cmp): Likewise.
2946 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
2947 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
2948 (svwhilelt): Likewise.
2950 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2952 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
2953 (unary_to_uint): Define.
2954 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
2955 (unary_count): Rename to...
2956 (unary_to_uint_def, unary_to_uint): ...this.
2957 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
2959 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2961 * config/aarch64/aarch64-sve-builtins-functions.h
2962 (code_for_mode_function): New class.
2963 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
2964 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
2965 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
2966 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
2967 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
2969 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2971 * config/aarch64/iterators.md (addsub): New code attribute.
2972 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
2974 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
2975 in the asm string and attributes. Fix indentation.
2976 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
2978 (@aarch64_sve_<optab><mode>): ...this.
2979 * config/aarch64/aarch64-sve-builtins.h
2980 (function_expander::expand_signed_unpred_op): Delete.
2981 * config/aarch64/aarch64-sve-builtins.cc
2982 (function_expander::expand_signed_unpred_op): Likewise.
2983 (function_expander::map_to_rtx_codes): If the optab isn't defined,
2984 try using code_for_aarch64_sve instead.
2985 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
2986 (svqsub_impl): Likewise.
2987 (svqadd, svqsub): Use rtx_code_function instead.
2989 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2991 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
2992 (HADDSUB, sur, addsub): Remove them.
2994 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2996 * tree-nrv.c (pass_return_slot::execute): Handle all internal
2997 functions the same way, rather than singling out those that
2998 aren't mapped directly to optabs.
3000 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3002 * target.def (compatible_vector_types_p): New target hook.
3003 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3004 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3005 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3006 * doc/tm.texi: Regenerate.
3007 * gimple-expr.c: Include target.h.
3008 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3009 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3011 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3012 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3013 Use the original predicate if it already has a suitable type.
3015 2020-01-09 Martin Jambor <mjambor@suse.cz>
3017 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3018 resolve_speculation and redirect_call_stmt_to_callee static. Change
3019 return type of set_call_stmt to cgraph_edge *.
3020 * auto-profile.c (afdo_indirect_call): Adjust call to
3021 redirect_call_stmt_to_callee.
3022 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3023 make the this pointer explicit, adjust self-recursive calls and the
3024 call top make_direct. Return the resulting edge.
3025 (cgraph_edge::remove): Make this pointer explicit.
3026 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3027 (cgraph_edge::make_direct): Likewise, adjust call to
3028 resolve_speculation.
3029 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3030 call to set_call_stmt.
3031 (cgraph_update_edges_for_call_stmt_node): Update call to
3032 set_call_stmt and remove.
3033 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3034 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3035 (cgraph_node::create_edge_including_clones): Moved "first" definition
3036 of edge to the block where it was used. Adjusted calls to
3038 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3039 cgraph_edge::remove.
3040 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3041 make_direct and redirect_call_stmt_to_callee.
3042 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3043 resolve_speculation and make_direct.
3044 * ipa-inline-transform.c (inline_transform): Adjust call to
3045 redirect_call_stmt_to_callee.
3046 (check_speculations_1):: Adjust call to resolve_speculation.
3047 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3048 resolve-speculation.
3049 (inline_small_functions): Adjust call to resolve_speculation.
3050 (ipa_inline): Likewise.
3051 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3053 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3054 safe with regards to edge removal, adjust calls to
3055 redirect_call_stmt_to_callee.
3056 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3057 and redirect_call_stmt_to_callee.
3058 * multiple_target.c (create_dispatcher_calls): Adjust call to
3059 redirect_call_stmt_to_callee
3060 (redirect_to_specific_clone): Likewise.
3061 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3062 Adjust calls to cgraph_edge::remove.
3063 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3064 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3065 (expand_call_inline): Adjust call to cgraph_edge::remove.
3067 2020-01-09 Martin Liska <mliska@suse.cz>
3069 * params.opt: Set Optimization for
3070 param_max_speculative_devirt_maydefs.
3072 2020-01-09 Martin Sebor <msebor@redhat.com>
3076 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3078 2020-01-09 Martin Liska <mliska@suse.cz>
3080 * auto-profile.c (auto_profile): Use opt_for_fn
3082 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3083 (propagate_vals_across_arith_jfunc): Likewise.
3084 (hint_time_bonus): Likewise.
3085 (incorporate_penalties): Likewise.
3086 (good_cloning_opportunity_p): Likewise.
3087 (perform_estimation_of_a_value): Likewise.
3088 (estimate_local_effects): Likewise.
3089 (ipcp_propagate_stage): Likewise.
3090 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3091 (set_switch_stmt_execution_predicate): Likewise.
3092 (analyze_function_body): Likewise.
3093 * ipa-inline-analysis.c (offline_size): Likewise.
3094 * ipa-inline.c (early_inliner): Likewise.
3095 * ipa-prop.c (ipa_analyze_node): Likewise.
3096 (ipcp_transform_function): Likewise.
3097 * ipa-sra.c (process_scan_results): Likewise.
3098 (ipa_sra_summarize_function): Likewise.
3099 * params.opt: Rename ipcp-unit-growth to
3100 ipa-cp-unit-growth. Add Optimization for various
3101 IPA-related parameters.
3103 2020-01-09 Richard Biener <rguenther@suse.de>
3106 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3108 2020-01-09 Richard Biener <rguenther@suse.de>
3110 PR tree-optimization/93040
3111 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3113 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3115 * common/config/avr/avr-common.c (avr_option_optimization_table)
3116 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3118 2020-01-09 Martin Liska <mliska@suse.cz>
3120 * cgraphclones.c (symbol_table::materialize_all_clones):
3121 Use cgraph_node::dump_name.
3123 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3126 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3127 output_operand_lossage instead of gcc_unreachable.
3128 * doc/md.texi (riscv f constraint): Fix typo.
3131 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3132 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3133 CONST_SCALAR_INT_P instead of CONST_INT_P.
3134 (*subv<mode>4_1): Rename to ...
3135 (subv<mode>4_1): ... this.
3136 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3137 define_insn_and_split patterns.
3138 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3141 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3143 * vec.c (class selftest::count_dtor): New class.
3144 (selftest::test_auto_delete_vec): New test.
3145 (selftest::vec_c_tests): Call it.
3146 * vec.h (class auto_delete_vec): New class template.
3147 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3149 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3151 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3153 2020-01-08 Jim Wilson <jimw@sifive.com>
3155 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3156 use of TLS_MODEL_LOCAL_EXEC when not pic.
3158 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3160 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3163 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3166 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3167 *stack_protect_set_3 peephole2): Also check that the second
3168 insns source is general_operand.
3171 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3172 predicate for output operand instead of register_operand.
3173 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3174 memory destination and non-memory operands[2].
3176 2020-01-08 Martin Liska <mliska@suse.cz>
3178 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3179 ::dump_asm_name instead of (::name or ::asm_name).
3180 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3181 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3182 (analyze_functions): Likewise.
3183 (expand_all_functions): Likewise.
3184 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3185 (propagate_bits_across_jump_function): Likewise.
3186 (dump_profile_updates): Likewise.
3187 (ipcp_store_bits_results): Likewise.
3188 (ipcp_store_vr_results): Likewise.
3189 * ipa-devirt.c (dump_targets): Likewise.
3190 * ipa-fnsummary.c (analyze_function_body): Likewise.
3191 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3192 (process_hsa_functions): Likewise.
3193 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3194 (set_alias_uids): Likewise.
3195 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3196 * ipa-inline.c (recursive_inlining): Likewise.
3197 (inline_to_all_callers_1): Likewise.
3198 (ipa_inline): Likewise.
3199 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3200 (ipa_propagate_frequency): Likewise.
3201 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3202 (remove_described_reference): Likewise.
3203 * ipa-pure-const.c (worse_state): Likewise.
3204 (check_retval_uses): Likewise.
3205 (analyze_function): Likewise.
3206 (propagate_pure_const): Likewise.
3207 (propagate_nothrow): Likewise.
3208 (dump_malloc_lattice): Likewise.
3209 (propagate_malloc): Likewise.
3210 (pass_local_pure_const::execute): Likewise.
3211 * ipa-visibility.c (optimize_weakref): Likewise.
3212 (function_and_variable_visibility): Likewise.
3213 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3214 (ipa_discover_variable_flags): Likewise.
3215 * lto-streamer-out.c (output_function): Likewise.
3216 (output_constructor): Likewise.
3217 * tree-inline.c (copy_bb): Likewise.
3218 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3219 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3221 2020-01-08 Richard Biener <rguenther@suse.de>
3224 * tree-eh.c (sink_clobbers): Update virtual operands for
3225 the first and last stmt only. Add a dry-run capability.
3226 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3227 after CFG manipulations and in RPO order to catch all
3228 secondary opportunities reliably.
3230 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3233 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3235 2019-01-08 Richard Biener <rguenther@suse.de>
3238 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3239 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3240 virtual operand, also updating SSA use.
3241 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3242 Update stmt after resetting virtual operand.
3243 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3244 * gimple-iterator.c (gsi_remove): When not removing the stmt
3245 permanently do not delink immediate uses or mark the stmt modified.
3247 2020-01-08 Martin Liska <mliska@suse.cz>
3249 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3250 (ipa_call_context::estimate_size_and_time): Likewise.
3251 (inline_analyze_function): Likewise.
3253 2020-01-08 Martin Liska <mliska@suse.cz>
3255 * cgraph.c (cgraph_node::dump): Use systematically
3258 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3260 Add -nodevicespecs option for avr.
3263 * config/avr/avr.opt (-nodevicespecs): New driver option.
3264 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3265 "-specs=device-specs/..." if that option is not set.
3266 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3268 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3270 Implement 64-bit double functions for avr.
3273 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3274 --with-double-comparison.
3275 * doc/install.texi: Document them.
3276 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3277 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3278 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3279 * doc/invoke.texi (AVR Built-in Macros): Document them.
3280 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3281 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3282 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3284 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3287 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3288 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3289 when only building rm-profile multilibs.
3291 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3294 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3295 lattice for a value to check.
3296 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3297 finite propagation in self-recursive scc.
3299 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3301 * ipa-inline.c (caller_growth_limits): Restore the AND.
3303 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3305 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3306 (VEC_ALLREG_ALT): New iterator.
3307 (VEC_ALLREG_INT_MODE): New iterator.
3308 (VCMP_MODE): New iterator.
3309 (VCMP_MODE_INT): New iterator.
3310 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3311 (vec_cmp<u>v64qidi): New define_expand.
3312 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3313 (vec_cmpu<mode>di_exec): New define_expand.
3314 (vec_cmp<u>v64qidi_exec): New define_expand.
3315 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3316 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3317 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3318 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3319 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3320 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3321 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3322 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3323 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3324 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3326 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3327 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3329 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3331 * config/gcn/constraints.md (DA): Update description and match.
3333 (Db): New constraint.
3334 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3336 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3337 Implement 'Db' mixed immediate type.
3338 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3339 (addcv64si3_dup<exec_vcc>): Delete.
3340 (subcv64si3<exec_vcc>): Rework constraints.
3341 (addv64di3): Rework constraints.
3342 (addv64di3_exec): Rework constraints.
3343 (subv64di3): Rework constraints.
3344 (addv64di3_dup): Delete.
3345 (addv64di3_dup_exec): Delete.
3346 (addv64di3_zext): Rework constraints.
3347 (addv64di3_zext_exec): Rework constraints.
3348 (addv64di3_zext_dup): Rework constraints.
3349 (addv64di3_zext_dup_exec): Rework constraints.
3350 (addv64di3_zext_dup2): Rework constraints.
3351 (addv64di3_zext_dup2_exec): Rework constraints.
3352 (addv64di3_sext_dup2): Rework constraints.
3353 (addv64di3_sext_dup2_exec): Rework constraints.
3355 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3357 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3358 existing target checks.
3360 2020-01-07 Richard Biener <rguenther@suse.de>
3362 * doc/install.texi: Bump minimal supported MPC version.
3364 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3366 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3367 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3368 * langhooks.c: Include stor-layout.h.
3369 (lhd_simulate_enum_decl): New function.
3370 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3371 handle_arm_sve_h for the LTO frontend.
3372 (register_vector_type): Cope with null returns from pushdecl.
3374 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3376 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3377 (aarch64_sve::nvectors_if_data_type): Replace with...
3378 (aarch64_sve::builtin_type_p): ...this.
3379 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3380 (find_vector_type): Delete.
3381 (add_sve_type_attribute): New function.
3382 (lookup_sve_type_attribute): Likewise.
3383 (register_builtin_types): Add an "SVE type" attribute to each type.
3384 (register_tuple_type): Likewise.
3385 (svbool_type_p, nvectors_if_data_type): Delete.
3386 (mangle_builtin_type): Use lookup_sve_type_attribute.
3387 (builtin_type_p): Likewise. Add an overload that returns the
3388 number of constituent vector and predicate registers.
3389 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3390 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3391 instead of aarch64_sve_argument_p.
3392 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3393 (aarch64_pass_by_reference): Likewise.
3394 (aarch64_function_value_1): Likewise.
3395 (aarch64_return_in_memory): Likewise.
3396 (aarch64_layout_arg): Likewise.
3398 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3400 PR tree-optimization/93156
3401 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3402 least significant bit is always clear.
3404 PR tree-optimization/93118
3405 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3406 simplifier with two intermediate conversions.
3408 2020-01-07 Martin Liska <mliska@suse.cz>
3410 * params.opt: Add Optimization for various parameters.
3412 2020-01-07 Martin Liska <mliska@suse.cz>
3415 * doc/extend.texi: Explain cloning for target_clone
3418 2020-01-07 Martin Liska <mliska@suse.cz>
3420 PR tree-optimization/92860
3421 * common.opt: Make in Optimization option
3422 as it is affected by -O0, which is an Optimization
3424 * tree-inline.c (tree_inlinable_function_p):
3425 Use opt_for_fn for warn_inline.
3426 (expand_call_inline): Likewise.
3428 2020-01-07 Martin Liska <mliska@suse.cz>
3430 PR tree-optimization/92860
3431 * common.opt: Make flag_ree as optimization
3434 2020-01-07 Martin Liska <mliska@suse.cz>
3436 PR optimization/92860
3437 * params.opt: Mark param_min_crossjump_insns with Optimization
3440 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3442 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3443 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3445 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3447 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3448 helper function to return the valid addressing formats for a given
3449 hard register and mode.
3450 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3452 * config/rs6000/constraints.md (Q constraint): Update
3454 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3457 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3458 Use 'Q' for doing vector extract from memory.
3459 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3461 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3462 doing vector extract from memory.
3463 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3464 extract from memory.
3466 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3467 for the offset being 34-bits when -mcpu=future is used.
3469 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3471 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3472 instead of cmpib_comparison_operator in cmpib patterns.
3473 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3474 of cmpib_comparison_operator. Revise comment.
3476 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3478 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3479 in an IFN_DIV_POW2 node to be equal.
3481 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3483 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3484 (vect_check_scalar_mask): ...this.
3485 (vectorizable_store, vectorizable_load): Update call accordingly.
3486 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3487 argument in calls to conditional internal functions.
3489 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3491 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3492 '0' matching inputs.
3493 (subv64di3_exec): Likewise.
3495 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3497 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3498 * doc/md.texi (movstr): Likewise.
3500 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3502 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3505 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3507 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3509 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3510 to a temporary file and use move-if-change to update the real
3511 file where necessary.
3513 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3515 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3516 rather than Upa for CPY /M.
3518 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3520 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3523 2020-01-06 Martin Liska <mliska@suse.cz>
3525 PR tree-optimization/92860
3526 * params.opt: Mark param_max_combine_insns with Optimization
3529 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3532 * config/i386/i386.md (SWIDWI): New mode iterator.
3533 (DWI, dwi): Add TImode variants.
3534 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3535 <general_hilo_operand> instead of <general_operand>. Use
3536 CONST_SCALAR_INT_P instead of CONST_INT_P.
3537 (*addv<mode>4_1): Rename to ...
3538 (addv<mode>4_1): ... this.
3539 (QWI): New mode attribute.
3540 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3541 define_insn_and_split patterns.
3542 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3544 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3545 <general_hilo_operand> instead of <general_operand>.
3546 (*addcarry<mode>_1): New define_insn.
3547 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3549 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3551 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3552 Use "call" instead of "set".
3554 2020-01-03 Martin Jambor <mjambor@suse.cz>
3557 * ipa-cp.c (print_all_lattices): Skip functions without info.
3559 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3562 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3563 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3564 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3565 for 'e' simd clones.
3568 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3570 (mprefer-vector-width=): Add Save.
3571 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3572 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3573 (ix86_debug_options, ix86_function_specific_print): Adjust
3574 ix86_target_string callers.
3575 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3576 (ix86_valid_target_attribute_tree): Likewise.
3577 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3578 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3579 ix86_target_string caller.
3582 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3583 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3584 instead of gen_int_shift_amount + convert_modes.
3586 PR rtl-optimization/93088
3587 * loop-iv.c (find_single_def_src): Punt after looking through
3588 128 reg copies for regs with single definitions. Move definitions
3591 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3593 * config/arm/arm-c.c (arm_cpu_builtins): Define
3594 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3595 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3596 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3597 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3598 * config/arm/arm-tables.opt: Regenerated.
3599 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3600 arm_arch_i8mm and arm_arch_bf16 when enabled.
3601 * config/arm/arm.h (TARGET_I8MM): New macro.
3602 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3603 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3604 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3605 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3606 (v8_6_a_simd_variants): New.
3607 (v8_*_a_simd_variants): Add i8mm and bf16.
3608 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3610 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3613 * predict.c (compute_function_frequency): Don't call
3614 warn_function_cold on functions that already have cold attribute.
3616 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3619 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3620 COMDAT group function labels in .data.rel.ro.local section.
3621 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3624 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3625 comparison_operator in B and S integer comparisons. Likewise, use
3626 ordered_comparison_operator instead of cmpib_comparison_operator in
3628 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3630 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3632 Update copyright years.
3634 * gcc.c (process_command): Update copyright notice dates.
3635 * gcov-dump.c (print_version): Ditto.
3636 * gcov.c (print_version): Ditto.
3637 * gcov-tool.c (print_version): Ditto.
3638 * gengtype.c (create_file): Ditto.
3639 * doc/cpp.texi: Bump @copying's copyright year.
3640 * doc/cppinternals.texi: Ditto.
3641 * doc/gcc.texi: Ditto.
3642 * doc/gccint.texi: Ditto.
3643 * doc/gcov.texi: Ditto.
3644 * doc/install.texi: Ditto.
3645 * doc/invoke.texi: Ditto.
3647 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3649 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3652 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3654 PR tree-optimization/93098
3655 * match.pd (popcount): For shift amounts, use integer_onep
3656 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3657 tests. Make sure that precision is power of two larger than or equal
3658 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3659 instead of ULL suffixed constants. Formatting fixes.
3661 Copyright (C) 2020 Free Software Foundation, Inc.
3663 Copying and distribution of this file, with or without modification,
3664 are permitted in any medium without royalty provided the copyright
3665 notice and this notice are preserved.