1 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
3 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
4 (simd_ummla, simd_usmmla): Likewise.
5 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
6 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
9 2020-02-07 Richard Biener <rguenther@suse.de>
12 * tree-inline.c (fold_marked_statements): Do a PRE walk,
13 skipping unreachable regions.
14 (optimize_inline_calls): Skip folding stmts when we didn't
17 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
20 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
21 Don't return aggregates with only SFmode and DFmode in SSE
23 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
25 2020-02-07 Jakub Jelinek <jakub@redhat.com>
28 * config/rs6000/rs6000-logue.c
29 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
30 if it fails, move rs into end_addr and retry. Add
31 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
32 the insn pattern doesn't describe well what exactly happens to
36 * config/i386/predicates.md (avx_identity_operand): Remove.
37 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
38 (avx_<castmode><avxsizesuffix>_<castmode>,
39 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
40 a VEC_CONCAT of the operand and UNSPEC_CAST.
41 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
42 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
46 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
47 recog_data.insn if distance_non_agu_define changed it.
49 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
52 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
53 we only had X-FORM (reg+reg) addressing for vectors. Also before
54 ISA 3.0, we only had X-FORM addressing for scalars in the
55 traditional Altivec registers.
57 2020-02-06 <zhongyunde@huawei.com>
58 Vladimir Makarov <vmakarov@redhat.com>
60 PR rtl-optimization/93561
61 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
64 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
66 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
69 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
71 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
72 where the low and the high 32 bits are equal to each other specially,
73 with an rldimi instruction.
75 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
77 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
79 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
81 * config/arm/arm-tables.opt: Regenerate.
83 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
86 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
87 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
88 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
90 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
92 PR rtl-optimization/87763
93 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
95 2020-02-06 Delia Burduv <delia.burduv@arm.com>
97 * config/aarch64/aarch64-simd-builtins.def
98 (bfmlaq): New built-in function.
99 (bfmlalb): New built-in function.
100 (bfmlalt): New built-in function.
101 (bfmlalb_lane): New built-in function.
102 (bfmlalt_lane): New built-in function.
103 * config/aarch64/aarch64-simd.md
104 (aarch64_bfmmlaqv4sf): New pattern.
105 (aarch64_bfmlal<bt>v4sf): New pattern.
106 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
107 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
108 (vbfmlalbq_f32): New intrinsic.
109 (vbfmlaltq_f32): New intrinsic.
110 (vbfmlalbq_lane_f32): New intrinsic.
111 (vbfmlaltq_lane_f32): New intrinsic.
112 (vbfmlalbq_laneq_f32): New intrinsic.
113 (vbfmlaltq_laneq_f32): New intrinsic.
114 * config/aarch64/iterators.md (BF_MLA): New int iterator.
115 (bt): New int attribute.
117 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
119 * config/i386/i386.md (*pushtf): Emit "#" instead of
120 calling gcc_unreachable in insn output.
123 (*pushsf_rex64): Ditto for alternatives other than 1.
124 (*pushsf): Ditto for alternatives other than 1.
126 2020-02-06 Martin Liska <mliska@suse.cz>
128 PR gcov-profile/91971
129 PR gcov-profile/93466
130 * coverage.c (coverage_init): Revert mangling of
131 path into filename. It can lead to huge filename length.
132 Creation of subfolders seem more natural.
134 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
137 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
138 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
139 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
141 2020-02-06 Jakub Jelinek <jakub@redhat.com>
144 * config/i386/predicates.md (avx_identity_operand): New predicate.
145 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
146 define_insn_and_split.
149 * omp-low.c (use_pointer_for_field): For nested constructs, also
150 look for map clauses on target construct.
151 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
152 taskreg_nesting_level.
155 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
156 shared clause, call omp_notice_variable on outer context if any.
158 2020-02-05 Jason Merrill <jason@redhat.com>
161 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
162 non-zero address even if weak and not yet defined.
164 2020-02-05 Martin Sebor <msebor@redhat.com>
166 PR tree-optimization/92765
167 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
168 * tree-ssa-strlen.c (compute_string_length): Remove.
169 (determine_min_objsize): Remove.
170 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
171 Avoid using type size as the upper bound on string length.
172 (handle_builtin_string_cmp): Add an argument. Adjust.
173 (strlen_check_and_optimize_call): Pass additional argument to
174 handle_builtin_string_cmp.
176 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
178 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
179 (*pushdi2_rex64 peephole2): Unconditionally split after
181 (*ashl<mode>3_doubleword): Ditto.
182 (*<shift_insn><mode>3_doubleword): Ditto.
184 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
187 * config/rs6000/rs6000.c (get_vector_offset): Fix
189 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
191 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
193 2020-02-05 David Malcolm <dmalcolm@redhat.com>
196 (Special Functions for Debugging the Analyzer): Update description
197 of __analyzer_dump_exploded_nodes.
199 2020-02-05 Jakub Jelinek <jakub@redhat.com>
202 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
203 include sets and not clobbers in the vzeroupper pattern.
204 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
205 the parallel has 17 (64-bit) or 9 (32-bit) elts.
206 (*avx_vzeroupper_1): New define_insn_and_split.
209 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
210 don't run when !optimize.
211 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
214 2020-02-05 Richard Biener <rguenther@suse.de>
217 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
218 checks before matching calls.
220 2020-02-05 Jakub Jelinek <jakub@redhat.com>
222 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
223 function comment typo.
226 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
227 simd_clone_create failed when i == 0, adjust clone->nargs by
230 2020-02-05 Martin Liska <mliska@suse.cz>
233 * doc/invoke.texi: Document that one should
234 not combine ASLR and -fpch.
236 2020-02-04 Richard Biener <rguenther@suse.de>
238 PR tree-optimization/93538
239 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
241 2020-02-04 Richard Biener <rguenther@suse.de>
243 PR tree-optimization/91123
244 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
245 (vn_walk_cb_data::last_vuse): New member.
246 (vn_walk_cb_data::saved_operands): Likewsie.
247 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
248 (vn_walk_cb_data::push_partial_def): Use finish.
249 (vn_reference_lookup_2): Update last_vuse and use finish if
250 we've saved operands.
251 (vn_reference_lookup_3): Use finish and update calls to
252 push_partial_defs everywhere. When translating through
253 memcpy or aggregate copies save off operands and alias-set.
254 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
255 operation for redundant store removal.
257 2020-02-04 Richard Biener <rguenther@suse.de>
259 PR tree-optimization/92819
260 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
261 generating more stmts than before.
263 2020-02-04 Martin Liska <mliska@suse.cz>
265 * config/arm/arm.c (arm_gen_far_branch): Move the function
266 outside of selftests.
268 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
270 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
271 function to adjust PC-relative vector addresses.
272 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
273 handle vectors with PC-relative addresses.
275 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
277 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
279 (hard_reg_and_mode_to_addr_mask): Delete.
280 (rs6000_adjust_vec_address): If the original vector address
281 was REG+REG or REG+OFFSET and the element is not zero, do the add
282 of the elements in the original address before adding the offset
283 for the vector element. Use address_to_insn_form to validate the
284 address using the register being loaded, rather than guessing
285 whether the address is a DS-FORM or DQ-FORM address.
287 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
289 * config/rs6000/rs6000.c (get_vector_offset): New helper function
290 to calculate the offset in memory from the start of a vector of a
291 particular element. Add code to keep the element number in
292 bounds if the element number is variable.
293 (rs6000_adjust_vec_address): Move calculation of offset of the
294 vector element to get_vector_offset.
295 (rs6000_split_vec_extract_var): Do not do the initial AND of
296 element here, move the code to get_vector_offset.
298 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
300 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
303 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
305 * config/rs6000/constraints.md: Improve documentation.
307 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
310 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
311 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
313 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
315 * config.gcc: Remove "carrizo" support.
316 * config/gcn/gcn-opts.h (processor_type): Likewise.
317 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
318 * config/gcn/gcn.opt (gpu_type): Likewise.
319 * config/gcn/t-omp-device: Likewise.
321 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
324 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
325 * config/arm/arm.c (arm_gen_far_branch): New function
327 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
329 2020-02-03 Julian Brown <julian@codesourcery.com>
330 Tobias Burnus <tobias@codesourcery.com>
332 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
334 2020-02-03 Jakub Jelinek <jakub@redhat.com>
337 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
338 valid RTL to sum up the lowest and second lowest bytes of the popcnt
341 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
343 PR rtl-optimization/91333
344 * ira-color.c (struct allocno_color_data): Add member
346 (init_allocno_threads): Set the member up.
347 (bucket_allocno_compare_func): Add compare hard reg
350 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
352 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
354 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
355 * config.in: Regenerated.
356 * configure: Regenerated.
357 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
358 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
359 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
361 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
363 * configure: Regenerate.
365 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
367 PR rtl-optimization/91333
368 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
369 reg preferences comparison up.
371 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
373 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
374 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
375 aarch64-sve-builtins-base.h.
376 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
377 aarch64-sve-builtins-base.cc.
378 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
379 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
381 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
382 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
383 (svcvtnt): New functions.
384 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
385 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
386 (svcvtnt): New functions.
387 (svcvt): Add a form that converts f32 to bf16.
388 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
389 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
391 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
392 Treat B as bfloat16_t.
393 (ternary_bfloat_lane_base): New class.
394 (ternary_bfloat_def): Likewise.
395 (ternary_bfloat): New shape.
396 (ternary_bfloat_lane_def): New class.
397 (ternary_bfloat_lane): New shape.
398 (ternary_bfloat_lanex2_def): New class.
399 (ternary_bfloat_lanex2): New shape.
400 (ternary_bfloat_opt_n_def): New class.
401 (ternary_bfloat_opt_n): New shape.
402 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
403 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
404 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
405 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
406 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
407 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
408 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
409 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
410 the pattern off the narrow mode instead of the wider one.
411 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
412 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
413 (sve_fp_op): Handle them.
414 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
415 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
417 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
419 * config/aarch64/arm_sve.h: Include arm_bf16.h.
420 * config/aarch64/aarch64-modes.def (BF): Move definition before
421 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
422 (SVE_MODES): Handle BF modes.
423 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
425 (aarch64_full_sve_mode): Likewise.
426 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
428 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
429 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
430 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
431 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
433 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
435 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
436 (TYPES_all_data): Add bf16.
437 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
438 (register_tuple_type): Increase buffer size.
439 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
440 (bf16): New type suffix.
441 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
442 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
443 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
444 Change type from all_data to all_arith.
445 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
448 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
449 Matthew Malcomson <matthew.malcomson@arm.com>
450 Richard Sandiford <richard.sandiford@arm.com>
452 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
453 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
454 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
455 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
456 __ARM_FEATURE_MATMUL_FP64.
457 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
458 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
459 be disabled at the same time.
460 (f32mm): New extension.
461 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
462 (AARCH64_FL_F64MM): Bump to the next bit up.
463 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
464 (TARGET_SVE_F64MM): New macros.
465 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
466 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
467 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
468 (UNSPEC_ZIP2Q): New unspeccs.
469 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
470 (optab, sur, perm_insn): Handle the new unspecs.
471 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
472 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
473 TARGET_SVE_F64MM instead of separate tests.
474 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
475 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
476 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
477 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
478 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
479 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
480 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
481 (TYPES_s_signed): New macro.
482 (TYPES_s_integer): Use it.
483 (TYPES_d_float): New macro.
484 (TYPES_d_data): Use it.
485 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
486 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
487 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
488 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
490 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
492 (ternary_resize2_lane_base): Likewise.
493 (ternary_resize2_base): New class.
494 (ternary_qq_lane_base): Likewise.
495 (ternary_intq_uintq_lane_def): Likewise.
496 (ternary_intq_uintq_lane): New shape.
497 (ternary_intq_uintq_opt_n_def): New class
498 (ternary_intq_uintq_opt_n): New shape.
499 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
500 (ternary_uintq_intq_def): New class.
501 (ternary_uintq_intq): New shape.
502 (ternary_uintq_intq_lane_def): New class.
503 (ternary_uintq_intq_lane): New shape.
504 (ternary_uintq_intq_opt_n_def): New class.
505 (ternary_uintq_intq_opt_n): New shape.
506 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
507 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
508 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
509 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
511 (svdotprod_lane_impl): ...this new class.
512 (svmmla_impl, svusdot_impl): New classes.
513 (svdot_lane): Update to use svdotprod_lane_impl.
514 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
515 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
517 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
518 function, with no types defined.
519 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
520 AARCH64_FL_I8MM functions.
521 (svmmla): New AARCH64_FL_F32MM function.
522 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
523 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
524 AARCH64_FL_F64MM function.
525 (REQUIRED_EXTENSIONS):
527 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
529 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
532 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
534 * config/i386/i386.md (*movoi_internal_avx): Do not check for
535 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
536 (*movti_internal): Do not check for
537 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
538 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
539 just after check for TARGET_AVX.
540 (*movdf_internal): Ditto.
541 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
542 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
543 * config/i386/sse.md (mov<mode>_internal): Only check
544 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
545 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
546 (<sse>_andnot<mode>3<mask_name>): Move check for
547 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
548 (<code><mode>3<mask_name>): Ditto.
549 (*andnot<mode>3): Ditto.
551 (*<code><mode>3): Ditto.
553 (*andnot<VI:mode>3): Remove
554 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
555 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
556 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
557 (sse4_1_blendv<ssemodesuffix>): Ditto.
558 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
559 Explain that tune applies to 128bit instructions only.
561 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
563 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
564 to definition of hsa_kernel_description. Parse assembly to find SGPR
565 and VGPR count of kernel and store in hsa_kernel_description.
567 2020-01-31 Tamar Christina <tamar.christina@arm.com>
569 PR rtl-optimization/91838
570 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
571 to truncate if allowed or reject combination.
573 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
575 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
576 (find_inv_vars_cb): Likewise.
578 2020-01-31 David Malcolm <dmalcolm@redhat.com>
580 * calls.c (special_function_p): Split out the check for DECL_NAME
581 being non-NULL and fndecl being extern at file scope into a
582 new maybe_special_function_p and call it. Drop check for fndecl
583 being non-NULL that was after a usage of DECL_NAME (fndecl).
584 * tree.h (maybe_special_function_p): New inline function.
586 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
588 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
589 (mask_gather_load<mode>): ... here, and zero-initialize the
591 (maskload<mode>di): Zero-initialize the destination.
594 2020-01-30 David Malcolm <dmalcolm@redhat.com>
597 * doc/analyzer.texi (Limitations): Note that constraints on
598 floating-point values are currently ignored.
600 2020-01-30 Jakub Jelinek <jakub@redhat.com>
603 * symtab.c (symtab_node::noninterposable_alias): If localalias
604 already exists, but is not usable, append numbers after it until
605 a unique name is found. Formatting fix.
608 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
611 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
613 * config/gcn/gcn.c (print_operand): Handle LTGT.
614 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
616 2020-01-30 Richard Biener <rguenther@suse.de>
618 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
619 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
621 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
623 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
624 without a DECL in .data.rel.ro.local.
626 2020-01-30 Jakub Jelinek <jakub@redhat.com>
629 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
634 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
635 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
636 any_extend code iterator instead of always zero_extend.
637 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
638 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
639 Use any_extend code iterator instead of always zero_extend.
640 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
641 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
642 Use any_extend code iterator instead of always zero_extend.
643 (*sse2_pmovmskb_ext): New define_insn.
644 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
647 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
648 (*popcountsi2_zext_falsedep): New define_insn.
650 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
652 * config.in: Regenerated.
653 * configure: Regenerated.
655 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
658 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
659 LLVM's assembler changed the default in version 9.
661 2020-01-24 Jeff Law <law@redhat.com>
663 PR tree-optimization/89689
664 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
666 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
670 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
672 PR rtl-optimization/87763
673 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
674 simplification to handle subregs as well as bare regs.
675 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
677 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
680 * ira.c (ira): Revert use of simplified LRA algorithm.
682 2020-01-29 Martin Jambor <mjambor@suse.cz>
684 PR tree-optimization/92706
685 * tree-sra.c (struct access): Fields first_link, last_link,
686 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
687 next_rhs_queued and grp_rhs_queued respectively, new fields
688 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
689 (struct assign_link): Field next renamed to next_rhs, new field
690 next_lhs. Updated comment.
691 (work_queue_head): Renamed to rhs_work_queue_head.
692 (lhs_work_queue_head): New variable.
693 (add_link_to_lhs): New function.
694 (relink_to_new_repr): Also relink LHS lists.
695 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
696 (add_access_to_lhs_work_queue): New function.
697 (pop_access_from_work_queue): Renamed to
698 pop_access_from_rhs_work_queue.
699 (pop_access_from_lhs_work_queue): New function.
700 (build_accesses_from_assign): Also add links to LHS lists and to LHS
702 (child_would_conflict_in_lacc): Renamed to
703 child_would_conflict_in_acc. Adjusted parameter names.
704 (create_artificial_child_access): New parameter set_grp_read, use it.
705 (subtree_mark_written_and_enqueue): Renamed to
706 subtree_mark_written_and_rhs_enqueue.
707 (propagate_subaccesses_across_link): Renamed to
708 propagate_subaccesses_from_rhs.
709 (propagate_subaccesses_from_lhs): New function.
710 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
713 2020-01-29 Martin Jambor <mjambor@suse.cz>
715 PR tree-optimization/92706
716 * tree-sra.c (struct access): Adjust comment of
717 grp_total_scalarization.
718 (find_access_in_subtree): Look for single children spanning an entire
720 (scalarizable_type_p): Allow register accesses, adjust callers.
721 (completely_scalarize): Remove function.
722 (scalarize_elem): Likewise.
723 (create_total_scalarization_access): Likewise.
724 (sort_and_splice_var_accesses): Do not track total scalarization
726 (analyze_access_subtree): New parameter totally, adjust to new meaning
727 of grp_total_scalarization.
728 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
729 (can_totally_scalarize_forest_p): New function.
730 (create_total_scalarization_access): Likewise.
731 (create_total_access_and_reshape): Likewise.
732 (total_should_skip_creating_access): Likewise.
733 (totally_scalarize_subtree): Likewise.
734 (analyze_all_variable_accesses): Perform total scalarization after
735 subaccess propagation using the new functions above.
736 (initialize_constant_pool_replacements): Output initializers by
737 traversing the access tree.
739 2020-01-29 Martin Jambor <mjambor@suse.cz>
741 * tree-sra.c (verify_sra_access_forest): New function.
742 (verify_all_sra_access_forests): Likewise.
743 (create_artificial_child_access): Set parent.
744 (analyze_all_variable_accesses): Call the verifier.
746 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
748 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
749 if called on indirect edge.
750 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
751 speculative call if needed.
753 2020-01-29 Richard Biener <rguenther@suse.de>
755 PR tree-optimization/93428
756 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
757 permutation when the load node is created.
758 (vect_analyze_slp_instance): Re-use it here.
760 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
762 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
764 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
766 PR rtl-optimization/93272
767 * ira-lives.c (process_out_of_region_eh_regs): New function.
768 (process_bb_node_lives): Call it.
770 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
772 * coverage.c (read_counts_file): Make error message lowercase.
774 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
776 * profile-count.c (profile_quality_display_names): Fix ordering.
778 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
781 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
782 hash only when edge is first within the sequence.
783 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
784 (symbol_table::create_edge): Do not set target_prob.
785 (cgraph_edge::remove_caller): Watch for speculative calls when updating
787 (cgraph_edge::make_speculative): Drop target_prob parameter.
788 (cgraph_edge::speculative_call_info): Remove.
789 (cgraph_edge::first_speculative_call_target): New member function.
790 (update_call_stmt_hash_for_removing_direct_edge): New function.
791 (cgraph_edge::resolve_speculation): Rewrite to new API.
792 (cgraph_edge::speculative_call_for_target): New member function.
793 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
794 multiple speculation targets.
795 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
797 (verify_speculative_call): Verify that targets form an interval.
798 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
799 (cgraph_edge::first_speculative_call_target): New member function.
800 (cgraph_edge::next_speculative_call_target): New member function.
801 (cgraph_edge::speculative_call_target_ref): New member function.
802 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
803 (cgraph_edge): Remove target_prob.
804 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
805 Fix handling of speculative calls.
806 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
807 * ipa-fnsummary.c (analyze_function_body): Likewise.
808 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
809 * ipa-profile.c (dump_histogram): Fix formating.
810 (ipa_profile_generate_summary): Watch for overflows.
811 (ipa_profile): Do not require probablity to be 1/2; update to new API.
812 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
813 (update_indirect_edges_after_inlining): Update to new API.
814 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
816 * profile-count.h: (profile_probability::adjusted): New.
817 * tree-inline.c (copy_bb): Update to new speculative call API; fix
819 * value-prof.c (gimple_ic_transform): Rename to ...
820 (dump_ic_profile): ... this one; update dumping.
821 (stream_in_histogram_value): Fix formating.
822 (gimple_value_profile_transformations): Update.
824 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
827 * config/i386/i386.md (*movoi_internal_avx): Remove
828 TARGET_SSE_TYPELESS_STORES check.
829 (*movti_internal): Prefer TARGET_AVX over
830 TARGET_SSE_TYPELESS_STORES.
831 (*movtf_internal): Likewise.
832 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
833 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
834 from TARGET_SSE_TYPELESS_STORES.
836 2020-01-28 David Malcolm <dmalcolm@redhat.com>
838 * diagnostic-core.h (warning_at): Rename overload to...
839 (warning_meta): ...this.
840 (emit_diagnostic_valist): Delete decl of overload taking
842 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
843 (warning_at): Rename overload taking diagnostic_metadata to...
844 (warning_meta): ...this.
846 2020-01-28 Richard Biener <rguenther@suse.de>
848 PR tree-optimization/93439
849 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
850 * tree-cfg.c (move_sese_region_to_fn): ... here.
851 (verify_types_in_gimple_reference): Verify used cliques are
854 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
857 * config/i386/i386-options.c (set_ix86_tune_features): Add an
858 argument of a pointer to struct gcc_options and pass it to
859 parse_mtune_ctrl_str.
860 (ix86_function_specific_restore): Pass opts to
861 set_ix86_tune_features.
862 (ix86_option_override_internal): Likewise.
863 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
864 gcc_options and use it for x_ix86_tune_ctrl_string.
866 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
868 PR rtl-optimization/87763
869 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
870 simplification to handle subregs as well as bare regs.
871 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
873 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
875 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
876 for reduction chains that (now) include a call.
878 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
880 PR tree-optimization/92822
881 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
882 out the don't-care elements of a vector whose significant elements
883 are duplicates, make the don't-care elements duplicates too.
885 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
887 PR tree-optimization/93434
888 * tree-predcom.c (split_data_refs_to_components): Record which
889 components have had aliasing loads removed. Prevent store-store
890 commoning for all such components.
892 2020-01-28 Jakub Jelinek <jakub@redhat.com>
895 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
896 -1 or is_vshift is true, use new_vector with number of elts npatterns
897 rather than new_unary_operation.
899 PR tree-optimization/93454
900 * gimple-fold.c (fold_array_ctor_reference): Perform
901 elt_size.to_uhwi () just once, instead of calling it in every
902 iteration. Punt if that value is above size of the temporary
903 buffer. Decrease third native_encode_expr argument when
904 bufoff + elt_sz is above size of buf.
906 2020-01-27 Joseph Myers <joseph@codesourcery.com>
908 * config/mips/mips.c (mips_declare_object_name)
909 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
911 2020-01-27 Martin Liska <mliska@suse.cz>
913 PR gcov-profile/93403
914 * tree-profile.c (gimple_init_gcov_profiler): Generate
915 both __gcov_indirect_call_profiler_v4 and
916 __gcov_indirect_call_profiler_v4_atomic.
918 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
921 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
923 (@aarch64_split_simd_mov<mode>): Use it.
924 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
925 Leave the vec_extract patterns to handle 2-element vectors.
926 (aarch64_simd_mov_from_<mode>high): Likewise.
927 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
928 (vec_extractv2dfv1df): Likewise.
930 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
932 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
933 jump conditions for *compare_condjump<GPI:mode>.
935 2020-01-27 David Malcolm <dmalcolm@redhat.com>
938 * digraph.cc (test_edge::test_edge): Specify template for base
941 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
943 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
945 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
947 * config/arc/arc-protos.h (gen_mlo): Remove.
949 * config/arc/arc.c (AUX_MULHI): Define.
950 (arc_must_save_reister): Special handling for r58/59.
951 (arc_compute_frame_size): Consider mlo/mhi registers.
952 (arc_save_callee_saves): Emit fp/sp move only when emit_move
954 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
955 mlo/mhi name selection.
956 (arc_restore_callee_saves): Don't early restore blink when ISR.
957 (arc_expand_prologue): Add mlo/mhi saving.
958 (arc_expand_epilogue): Add mlo/mhi restoring.
961 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
962 numbering when MUL64 option is used.
963 (DWARF2_FRAME_REG_OUT): Define.
964 * config/arc/arc.md (arc600_stall): New pattern.
965 (VUNSPEC_ARC_ARC600_STALL): Define.
966 (mulsi64): Use correct mlo/mhi registers.
967 (mulsi_600): Clean it up.
968 * config/arc/predicates.md (mlo_operand): Remove any dependency on
970 (mhi_operand): Likewise.
972 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
973 Petro Karashchenko <petro.karashchenko@ring.com>
975 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
976 attributes if needed.
977 (prepare_move_operands): Generate special unspec instruction for
979 (arc_isuncached_mem_p): Propagate uncached attribute to each
981 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
982 (VUNSPEC_ARC_STDI): Likewise.
983 (ALLI): New mode iterator.
984 (mALLI): New mode attribute.
985 (lddi): New instruction pattern.
987 (stdidi_split): Split instruction for architectures which are not
988 supporting ll64 option.
989 (lddidi_split): Likewise.
991 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
993 PR rtl-optimization/92989
994 * lra-lives.c (process_bb_lives): Update the live-in set before
995 processing additional clobbers.
997 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
999 PR rtl-optimization/93170
1000 * cselib.c (cselib_invalidate_regno_val): New function, split out
1002 (cselib_invalidate_regno): ...here.
1003 (cselib_invalidated_by_call_p): New function.
1004 (cselib_process_insn): Iterate over all the hard-register entries in
1005 REG_VALUES and invalidate any that cross call-clobbered registers.
1007 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1009 * dojump.c (split_comparison): Use HONOR_NANS rather than
1010 HONOR_SNANS when splitting LTGT.
1012 2020-01-27 Martin Liska <mliska@suse.cz>
1015 * opts.c (print_filtered_help): Exclude language-specific
1016 options from --help=common unless enabled in all FEs.
1018 2020-01-27 Martin Liska <mliska@suse.cz>
1020 * opts.c (print_help): Exclude params from
1021 all except --help=param.
1023 2020-01-27 Martin Liska <mliska@suse.cz>
1026 * config/i386/i386-features.c (make_resolver_func):
1027 Align the code with ppc64 target implementation.
1028 Do not generate a unique name for resolver function.
1030 2020-01-27 Richard Biener <rguenther@suse.de>
1032 PR tree-optimization/93397
1033 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1034 converted reduction chain SLP graph adjustment.
1036 2020-01-26 Marek Polacek <polacek@redhat.com>
1039 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1042 2020-01-26 Jason Merrill <jason@redhat.com>
1045 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1048 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1050 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1051 (rx_setmem): Likewise.
1053 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1056 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1057 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1058 drop <di> from constraint of last operand.
1061 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1062 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1063 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1065 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1068 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1071 2020-01-24 Jeff Law <law@redhat.com>
1073 PR tree-optimization/92788
1074 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1077 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1080 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1081 *avx_vperm_broadcast_<mode>,
1082 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1083 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1084 Move before avx2_perm<mode>/avx512f_perm<mode>.
1087 * simplify-rtx.c (simplify_const_unary_operation,
1088 simplify_const_binary_operation): Punt for mode precision above
1089 MAX_BITSIZE_MODE_ANY_INT.
1091 2020-01-24 Andrew Pinski <apinski@marvell.com>
1093 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1096 2020-01-24 Jeff Law <law@redhat.com>
1099 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1100 for REGs. Call output_operand_lossage to get more reasonable
1103 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1105 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1106 gcn_fp_compare_operator.
1107 (vec_cmpu<mode>di): Use gcn_compare_operator.
1108 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1109 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1110 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1111 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1112 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1113 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1114 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1115 gcn_fp_compare_operator.
1116 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1117 gcn_fp_compare_operator.
1118 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1119 gcn_fp_compare_operator.
1120 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1121 gcn_fp_compare_operator.
1123 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1125 * doc/install.texi (Cross-Compiler-Specific Options): Document
1126 `--with-toolexeclibdir' option.
1128 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1130 * target.def (flags_regnum): Also mention effect on delay slot filling.
1131 * doc/tm.texi: Regenerate.
1133 2020-01-23 Jeff Law <law@redhat.com>
1135 PR translation/90162
1136 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1138 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1141 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1144 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1146 PR rtl-optimization/93402
1147 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1150 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1152 * config.in: Regenerated.
1153 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1154 for TARGET_LIBC_GNUSTACK.
1155 * configure: Regenerated.
1156 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1157 found to be 2.31 or greater.
1159 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1161 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1163 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1164 (mips_asm_file_end): New function. Delegate to
1165 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1166 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1168 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1171 * config/i386/i386-modes.def (POImode): New mode.
1172 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1173 * config/i386/i386.md (DPWI): New mode attribute.
1174 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1176 (QPWI): ... this. Use POI instead of OI for TImode.
1177 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1178 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1181 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1184 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1186 (speculation_tracker_rev): New pattern.
1187 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1188 Use speculation_tracker_rev to track the inverse condition.
1190 2020-01-23 Richard Biener <rguenther@suse.de>
1192 PR tree-optimization/93381
1193 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1194 alias-set of the def as argument and record the first one.
1195 (vn_walk_cb_data::first_set): New member.
1196 (vn_reference_lookup_3): Pass the alias-set of the current def
1197 to push_partial_def. Fix alias-set used in the aggregate copy
1199 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1200 * real.c (clear_significand_below): Fix out-of-bound access.
1202 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1205 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1206 New define_insn patterns.
1208 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1210 * doc/sourcebuild.texi (check-function-bodies): Add an
1211 optional target/xfail selector.
1213 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1215 PR rtl-optimization/93124
1216 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1217 bare USE and CLOBBER insns.
1219 2020-01-22 Andrew Pinski <apinski@marvell.com>
1221 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1223 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1226 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1227 diagnostic_manager into "ana" namespace.
1228 * selftest-run-tests.c (selftest::run_tests): Update for move of
1229 selftest::run_analyzer_selftests to
1230 ana::selftest::run_analyzer_selftests.
1232 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1234 * cfgexpand.c (union_stack_vars): Update the size.
1236 2020-01-22 Richard Biener <rguenther@suse.de>
1238 PR tree-optimization/93381
1239 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1240 throughout, handle all conversions the same.
1242 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1245 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1246 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1247 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1248 Call force_reg on high_in2 unconditionally.
1250 2020-01-22 Martin Liska <mliska@suse.cz>
1252 PR tree-optimization/92924
1253 * profile.c (compute_value_histograms): Divide
1256 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1259 * output.h (assemble_name_resolve): Declare.
1260 * varasm.c (assemble_name_resolve): New function.
1261 (assemble_name): Use it.
1262 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1264 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1266 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1267 update_web_docs_git instead of update_web_docs_svn.
1269 2020-01-21 Andrew Pinski <apinski@marvell.com>
1272 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1273 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1274 (*tlsgd_small_<mode>): Likewise.
1275 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1276 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1277 register. Convert that register back to dest using convert_mode.
1279 2020-01-21 Jim Wilson <jimw@sifive.com>
1281 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1284 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1285 Uros Bizjak <ubizjak@gmail.com>
1288 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1290 (legitimize_tls_address): Do GNU2 TLS address computation in
1291 ptr_mode and zero-extend result to Pmode.
1292 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1293 :P with :PTR and Pmode with ptr_mode.
1294 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1295 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1296 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1298 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1301 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1302 the last two operands are CONST_INT_P before using them as such.
1304 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1306 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1307 to get the integer element types.
1309 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1311 * config/aarch64/aarch64-sve-builtins.h
1312 (function_expander::convert_to_pmode): Declare.
1313 * config/aarch64/aarch64-sve-builtins.cc
1314 (function_expander::convert_to_pmode): New function.
1315 (function_expander::get_contiguous_base): Use it.
1316 (function_expander::prepare_gather_address_operands): Likewise.
1317 * config/aarch64/aarch64-sve-builtins-sve2.cc
1318 (svwhilerw_svwhilewr_impl::expand): Likewise.
1320 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1323 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1324 cfun->machine->label_is_assembled.
1325 (aarch64_print_patchable_function_entry): New.
1326 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1327 * config/aarch64/aarch64.h (struct machine_function): New field,
1330 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1333 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1336 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1339 * cgraph.c (cgraph_edge::resolve_speculation,
1340 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1341 call_stmt_site_hash.
1343 2020-01-21 Martin Liska <mliska@suse.cz>
1345 * config/rs6000/rs6000.c (common_mode_defined): Remove
1348 2020-01-21 Richard Biener <rguenther@suse.de>
1350 PR tree-optimization/92328
1351 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1352 type when value-numbering same-sized store by inserting a
1354 (eliminate_dom_walker::eliminate_stmt): When eliminating
1355 a redundant store handle bit-reinterpretation of the same value.
1357 2020-01-21 Andrew Pinski <apinski@marvel.com>
1360 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1362 (prepare_block_for_update): This. Use a worklist instead of
1365 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1367 * gcc/config/arm/arm.c (clear_operation_p):
1368 Initialise last_regno, skip first iteration
1369 based on the first_set value and use ints instead
1370 of the unnecessary HOST_WIDE_INTs.
1372 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1375 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1376 compare_mode other than SFmode or DFmode.
1378 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1381 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1382 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1383 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1385 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1387 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1389 2020-01-20 Andrew Pinski <apinski@marvell.com>
1392 * targhooks.c (default_print_patchable_function_entry): Use
1393 output_asm_insn to emit the nop instruction.
1395 2020-01-20 Fangrui Song <maskray@google.com>
1398 * targhooks.c (default_print_patchable_function_entry): Align to
1401 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1404 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1405 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1406 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1407 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1408 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1409 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1410 Remove the {q} suffix from lea.
1411 (*tls_dynamic_gnu2_call_64): Renamed to ...
1412 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1413 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1414 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1415 Pass Pmode to gen_tls_dynamic_gnu2_64.
1417 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1419 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1421 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1423 * config/aarch64/aarch64-sve-builtins-base.cc
1424 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1426 2020-01-20 Richard Biener <rguenther@suse.de>
1429 * dwarf2out.c (prune_unused_types): Unconditionally mark
1430 called function DIEs.
1432 2020-01-20 Martin Liska <mliska@suse.cz>
1434 PR tree-optimization/93199
1435 * tree-eh.c (struct leh_state): Add
1436 new field outer_non_cleanup.
1437 (cleanup_is_dead_in): Pass leh_state instead
1438 of eh_region. Add a checking that state->outer_non_cleanup
1439 points to outer non-clean up region.
1440 (lower_try_finally): Record outer_non_cleanup
1442 (lower_catch): Likewise.
1443 (lower_eh_filter): Likewise.
1444 (lower_eh_must_not_throw): Likewise.
1445 (lower_cleanup): Likewise.
1447 2020-01-20 Richard Biener <rguenther@suse.de>
1449 PR tree-optimization/93094
1450 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1451 (vect_transform_loop): Likewise.
1452 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1453 loop_vectorized_call to vect_transform_loop.
1454 * tree-vect-loop.c (vect_transform_loop): Pass down
1455 loop_vectorized_call to vect_loop_versioning.
1456 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1457 the earlier discovered loop_vectorized_call.
1459 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1461 * doc/contribute.texi: Update for SVN -> Git transition.
1462 * doc/install.texi: Likewise.
1464 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1467 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1468 speculative targets.
1469 (verify_speculative_call): New function
1470 (cgraph_node::verify_node): Use it.
1471 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1474 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1477 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1478 (cgraph_edge::make_direct): Remove all indirect targets.
1479 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1480 (cgraph_node::verify_node): Verify that only one call_stmt or
1481 lto_stmt_uid is set.
1482 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1484 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1485 (lto_output_ref): Simplify streaming of stmt.
1486 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1488 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1490 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1491 Mark parameter unused.
1493 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1495 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1497 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1499 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
1501 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
1503 * Makefile.in: Add coroutine-passes.o.
1504 * builtin-types.def (BT_CONST_SIZE): New.
1505 (BT_FN_BOOL_PTR): New.
1506 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
1507 * builtins.def (DEF_COROUTINE_BUILTIN): New.
1508 * coroutine-builtins.def: New file.
1509 * coroutine-passes.cc: New file.
1510 * function.h (struct GTY function): Add a bit to indicate that the
1511 function is a coroutine component.
1512 * internal-fn.c (expand_CO_FRAME): New.
1513 (expand_CO_YIELD): New.
1514 (expand_CO_SUSPN): New.
1515 (expand_CO_ACTOR): New.
1516 * internal-fn.def (CO_ACTOR): New.
1520 * passes.def: Add pass_coroutine_lower_builtins,
1521 pass_coroutine_early_expand_ifns.
1522 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
1523 (make_pass_coroutine_early_expand_ifns): New.
1524 * doc/invoke.texi: Document the fcoroutines command line
1527 2020-01-18 Jakub Jelinek <jakub@redhat.com>
1529 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
1532 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
1533 after checking the argument is a REG. Don't use REGNO (reg)
1534 again to set last_regno, reuse regno variable instead.
1536 2020-01-17 David Malcolm <dmalcolm@redhat.com>
1538 * doc/analyzer.texi (Limitations): Add note about NaN.
1540 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1541 Sudakshina Das <sudi.das@arm.com>
1543 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
1544 and valid immediate.
1545 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
1546 (lshrdi3): Generate thumb2_lsrl for valid immediates.
1547 * config/arm/constraints.md (Pg): New.
1548 * config/arm/predicates.md (long_shift_imm): New.
1549 (arm_reg_or_long_shift_imm): Likewise.
1550 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
1551 (thumb2_lsll): Likewise.
1554 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1555 Sudakshina Das <sudi.das@arm.com>
1557 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
1558 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
1559 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
1560 register pairs for doubleword quantities for ARMv8.1M-Mainline.
1561 * config/arm/thumb2.md (thumb2_asrl): New.
1562 (thumb2_lsll): Likewise.
1564 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1566 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
1569 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
1571 * gdbinit.in (help-gcc-hooks): New command.
1572 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
1573 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
1576 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1578 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
1579 correct target macro.
1581 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1583 * config/aarch64/aarch64-protos.h
1584 (aarch64_sve_ld1ro_operand_p): New.
1585 * config/aarch64/aarch64-sve-builtins-base.cc
1586 (class load_replicate): New.
1587 (class svld1ro_impl): New.
1588 (class svld1rq_impl): Change to inherit from load_replicate.
1589 (svld1ro): New sve intrinsic function base.
1590 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
1591 New DEF_SVE_FUNCTION.
1592 * config/aarch64/aarch64-sve-builtins-base.h
1593 (svld1ro): New decl.
1594 * config/aarch64/aarch64-sve-builtins.cc
1595 (function_expander::add_mem_operand): Modify assert to allow
1597 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
1599 * config/aarch64/aarch64.c
1600 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
1601 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
1602 (aarch64_sve_ld1ro_operand_p): New.
1603 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
1604 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
1605 * config/aarch64/predicates.md
1606 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
1608 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
1610 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
1611 Introduce this ACLE specified predefined macro.
1612 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
1613 (fp): Disabling this disables f64mm.
1614 (simd): Disabling this disables f64mm.
1615 (fp16): Disabling this disables f64mm.
1616 (sve): Disabling this disables f64mm.
1617 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
1618 (AARCH64_ISA_F64MM): New.
1619 (TARGET_F64MM): New.
1620 * doc/invoke.texi (f64mm): Document new option.
1622 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1624 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
1625 (neoversen1_tunings): Likewise.
1627 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
1630 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
1631 Add assert to ensure prolog has been emitted.
1632 (aarch64_split_atomic_op): Likewise.
1633 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
1634 Use epilogue_completed rather than reload_completed.
1635 (aarch64_atomic_exchange<mode>): Likewise.
1636 (aarch64_atomic_<atomic_optab><mode>): Likewise.
1637 (atomic_nand<mode>): Likewise.
1638 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
1639 (atomic_fetch_nand<mode>): Likewise.
1640 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
1641 (atomic_nand_fetch<mode>): Likewise.
1643 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1646 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
1648 (REVERSE_CONDITION): Delete.
1649 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
1650 (CCFP_CCFPE): Likewise.
1651 (e): New mode attribute.
1652 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
1653 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
1654 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
1655 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
1656 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
1657 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
1658 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
1659 name of generator from gen_ccmpdi to gen_ccmpccdi.
1660 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
1661 the previous comparison but aren't able to, use the new ccmp_rev
1664 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
1666 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
1667 than testing directly for INTEGER_CST.
1668 (gimplify_target_expr, gimplify_omp_depend): Likewise.
1670 2020-01-17 Jakub Jelinek <jakub@redhat.com>
1672 PR tree-optimization/93292
1673 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
1674 get_vectype_for_scalar_type returns NULL.
1676 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1678 * params.opt (-param=max-predicted-iterations): Increase range from 0.
1679 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
1681 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
1683 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
1685 * params.opt: (max-predicted-iterations): Set bounds.
1686 * predict.c (real_almost_one, real_br_prob_base,
1687 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
1688 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
1689 probabilities; do not truncate to reg_br_prob_bases.
1690 (estimate_loops_at_level): Pass max_cyclic_prob.
1691 (estimate_loops): Compute max_cyclic_prob.
1692 (estimate_bb_frequencies): Do not initialize real_*; update calculation
1694 * profile-count.c (profile_probability::to_sreal): New.
1695 * profile-count.h (class sreal): Move up in file.
1696 (profile_probability::to_sreal): Declare.
1698 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1701 (arm_invalid_conversion): New function for target hook.
1702 (arm_invalid_unary_op): New function for target hook.
1703 (arm_invalid_binary_op): New function for target hook.
1705 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1707 * config.gcc: Add arm_bf16.h.
1708 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
1709 (arm_simd_builtin_std_type): Add BFmode.
1710 (arm_init_simd_builtin_types): Define element types for vector types.
1711 (arm_init_bf16_types): New function.
1712 (arm_init_builtins): Add arm_init_bf16_types function call.
1713 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
1714 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
1715 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
1716 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
1717 (arm_vector_mode_supported_p): Add V4BF, V8BF.
1718 (arm_mangle_type): Add __bf16.
1719 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
1720 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
1721 arm_bf16_ptr_type_node.
1722 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
1723 define_split between ARM registers.
1724 * config/arm/arm_bf16.h: New file.
1725 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
1726 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
1728 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
1729 * config/arm/vfp.md: Add BFmode to movhf patterns.
1731 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
1732 Andre Vieira <andre.simoesdiasvieira@arm.com>
1734 * config/arm/arm-cpus.in (mve, mve_float): New features.
1735 (dsp, mve, mve.fp): New options.
1736 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
1737 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
1738 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
1740 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1741 Thomas Preud'homme <thomas.preudhomme@arm.com>
1743 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
1745 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
1746 error for using -mcmse when targeting Armv8.1-M Mainline.
1748 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1749 Thomas Preud'homme <thomas.preudhomme@arm.com>
1751 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
1752 address in r4 when targeting Armv8.1-M Mainline.
1753 (nonsecure_call_value_internal): Likewise.
1754 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
1755 a register match_operand again. Emit BLXNS when targeting
1757 (nonsecure_call_value_reg_thumb2): Likewise.
1759 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1760 Thomas Preud'homme <thomas.preudhomme@arm.com>
1762 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
1763 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
1764 variable as true when floating-point ABI is not hard. Replace
1765 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
1766 Generate VLSTM and VLLDM instruction respectively before and
1767 after a function call to cmse_nonsecure_call function.
1768 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
1769 (VUNSPEC_VLLDM): Likewise.
1770 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
1771 (lazy_load_multiple_insn): Likewise.
1773 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1774 Thomas Preud'homme <thomas.preudhomme@arm.com>
1776 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
1777 (arm_emit_vfp_multi_reg_pop): Likewise.
1778 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
1779 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
1780 restore callee-saved VFP registers.
1782 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1783 Thomas Preud'homme <thomas.preudhomme@arm.com>
1785 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
1786 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
1787 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
1788 callee-saved GPRs as well as clear ip register before doing a nonsecure
1789 call then restore callee-saved GPRs after it when targeting
1791 (arm_reorg): Adapt to function rename.
1793 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1794 Thomas Preud'homme <thomas.preudhomme@arm.com>
1796 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
1797 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
1798 clear_vfp_multiple pattern based on a new vfp parameter.
1799 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
1800 targeting Armv8.1-M Mainline.
1801 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
1802 unconditionally when targeting Armv8.1-M Mainline architecture. Check
1803 whether VFP registers are available before looking call_used_regs for a
1805 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
1806 of prototype of clear_operation_p.
1807 (clear_vfp_multiple_operation): New predicate.
1808 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
1809 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
1811 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1812 Thomas Preud'homme <thomas.preudhomme@arm.com>
1814 * config/arm/arm-protos.h (clear_operation_p): Declare.
1815 * config/arm/arm.c (clear_operation_p): New function.
1816 (cmse_clear_registers): Generate clear_multiple instruction pattern if
1817 targeting Armv8.1-M Mainline or successor.
1818 (output_return_instruction): Only output APSR register clearing if
1819 Armv8.1-M Mainline instructions not available.
1820 (thumb_exit): Likewise.
1821 * config/arm/predicates.md (clear_multiple_operation): New predicate.
1822 * config/arm/thumb2.md (clear_apsr): New define_insn.
1823 (clear_multiple): Likewise.
1824 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
1826 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1827 Thomas Preud'homme <thomas.preudhomme@arm.com>
1829 * config/arm/arm.c (fp_sysreg_names): Declare and define.
1830 (use_return_insn): Also return false for Armv8.1-M Mainline.
1831 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
1832 Mainline instructions are available.
1833 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
1834 when targeting Armv8.1-M Mainline Security Extensions.
1835 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
1836 Mainline entry function.
1837 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
1838 targeting Armv8.1-M Mainline or successor.
1839 (arm_expand_epilogue): Fix indentation of caller-saved register
1840 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
1842 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
1843 (FP_SYSREGS): Likewise.
1844 (enum vfp_sysregs_encoding): Define enum.
1845 (fp_sysreg_names): Declare.
1846 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
1847 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
1848 (pop_fpsysreg_insn): Likewise.
1850 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1851 Thomas Preud'homme <thomas.preudhomme@arm.com>
1853 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
1854 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
1855 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
1856 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
1857 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
1858 (ARMv8_1m_main): New feature group.
1859 (armv8.1-m.main): New architecture.
1860 * config/arm/arm-tables.opt: Regenerate.
1861 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
1862 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
1863 (arm_options_perform_arch_sanity_checks): Error out when targeting
1864 Armv8.1-M Mainline Security Extensions.
1865 * config/arm/arm.h (arm_arch8_1m_main): Declare.
1867 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1869 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
1870 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
1871 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
1872 aarch64_bfdot_laneq): New.
1873 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
1874 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
1875 vbfdotq_laneq_f32): New.
1876 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
1877 VBFMLA_W, VBF): New.
1878 (isquadop): Add V4BF, V8BF.
1880 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1882 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
1883 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
1884 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
1885 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
1886 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
1887 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
1888 usdot_laneq, sudot_lane,sudot_laneq): New.
1889 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
1890 (aarch64_<sur>dot_lane): New.
1891 * config/aarch64/arm_neon.h (vusdot_s32): New.
1893 (vusdot_lane_s32): New.
1894 (vsudot_lane_s32): New.
1895 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
1896 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
1898 2020-01-16 Martin Liska <mliska@suse.cz>
1900 * value-prof.c (dump_histogram_value): Fix
1901 obvious spacing issue.
1903 2020-01-16 Andrew Pinski <apinski@marvell.com>
1905 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
1906 !storage_order_barrier_p.
1908 2020-01-16 Andrew Pinski <apinski@marvell.com>
1910 * sched-int.h (_dep): Add unused bit-field field for the padding.
1911 * sched-deps.c (init_dep_1): Init unused field.
1913 2020-01-16 Andrew Pinski <apinski@marvell.com>
1915 * optabs.h (create_expand_operand): Initialize target field also.
1917 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
1919 PR tree-optimization/92429
1920 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
1921 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
1923 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
1926 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
1928 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
1929 aarch64_sve_int_mode to each mode.
1931 2020-01-15 David Malcolm <dmalcolm@redhat.com>
1933 * doc/analyzer.texi (Overview): Add note about
1934 -fdump-ipa-analyzer.
1936 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
1938 PR tree-optimization/93231
1939 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
1940 input_type is unsigned. Use tree_to_shwi for shift constant.
1941 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
1942 (simplify_count_trailing_zeroes): Add test to handle known non-zero
1943 inputs more efficiently.
1945 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
1947 * config/i386/i386.md (*movsf_internal): Do not require
1948 SSE2 ISA for alternatives 14 and 15.
1950 2020-01-15 Richard Biener <rguenther@suse.de>
1953 * tree-eh.c (sink_clobbers): If we already visited the destination
1954 block do not defer insertion.
1955 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
1956 the purpose of defered insertion.
1958 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1960 * BASE-VER: Bump to 10.0.1.
1962 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1964 PR tree-optimization/93247
1965 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
1966 type of the stmt that we're going to vectorize.
1968 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
1970 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
1971 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
1974 2020-01-15 Martin Liska <mliska@suse.cz>
1976 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
1977 2 calls of streamer_read_hwi in a function call.
1979 2020-01-15 Richard Biener <rguenther@suse.de>
1981 * alias.c (record_alias_subset): Avoid redundant work when
1982 subset is already recorded.
1984 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1986 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
1987 the analyzer options provide CWE identifiers.
1989 2020-01-14 David Malcolm <dmalcolm@redhat.com>
1991 * tree-diagnostic-path.cc (path_summary::event_range::print):
1992 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
1993 using get_pure_location.
1995 2020-01-15 Jakub Jelinek <jakub@redhat.com>
1997 PR tree-optimization/93262
1998 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
1999 perform head trimming only if the last argument is constant,
2000 either all ones, or larger or equal to head trim, in the latter
2001 case decrease the last argument by head_trim.
2003 PR tree-optimization/93249
2004 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2005 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2006 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2007 perform head trim unless we can prove there are no '\0' chars
2008 from the source among the first head_trim chars.
2010 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2012 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2014 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2017 * config/i386/sse.md
2018 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2019 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2020 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2021 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2022 just a single alternative instead of two, make operands 1 and 2
2025 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2028 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2031 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2033 * Makefile.in (lang_opt_files): Add analyzer.opt.
2034 (ANALYZER_OBJS): New.
2035 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2036 tristate.o and ANALYZER_OBJS.
2037 (TEXI_GCCINT_FILES): Add analyzer.texi.
2038 * common.opt (-fanalyzer): New driver option.
2039 * config.in: Regenerate.
2040 * configure: Regenerate.
2041 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2042 (gccdepdir): Also create depdir for "analyzer" subdir.
2043 * digraph.cc: New file.
2044 * digraph.h: New file.
2045 * doc/analyzer.texi: New file.
2046 * doc/gccint.texi ("Static Analyzer") New menu item.
2047 (analyzer.texi): Include it.
2048 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2049 ("Warning Options"): Add static analysis warnings to the list.
2050 (-Wno-analyzer-double-fclose): New option.
2051 (-Wno-analyzer-double-free): New option.
2052 (-Wno-analyzer-exposure-through-output-file): New option.
2053 (-Wno-analyzer-file-leak): New option.
2054 (-Wno-analyzer-free-of-non-heap): New option.
2055 (-Wno-analyzer-malloc-leak): New option.
2056 (-Wno-analyzer-possible-null-argument): New option.
2057 (-Wno-analyzer-possible-null-dereference): New option.
2058 (-Wno-analyzer-null-argument): New option.
2059 (-Wno-analyzer-null-dereference): New option.
2060 (-Wno-analyzer-stale-setjmp-buffer): New option.
2061 (-Wno-analyzer-tainted-array-index): New option.
2062 (-Wno-analyzer-use-after-free): New option.
2063 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2064 (-Wno-analyzer-use-of-uninitialized-value): New option.
2065 (-Wanalyzer-too-complex): New option.
2066 (-fanalyzer-call-summaries): New warning.
2067 (-fanalyzer-checker=): New warning.
2068 (-fanalyzer-fine-grained): New warning.
2069 (-fno-analyzer-state-merge): New warning.
2070 (-fno-analyzer-state-purge): New warning.
2071 (-fanalyzer-transitivity): New warning.
2072 (-fanalyzer-verbose-edges): New warning.
2073 (-fanalyzer-verbose-state-changes): New warning.
2074 (-fanalyzer-verbosity=): New warning.
2075 (-fdump-analyzer): New warning.
2076 (-fdump-analyzer-callgraph): New warning.
2077 (-fdump-analyzer-exploded-graph): New warning.
2078 (-fdump-analyzer-exploded-nodes): New warning.
2079 (-fdump-analyzer-exploded-nodes-2): New warning.
2080 (-fdump-analyzer-exploded-nodes-3): New warning.
2081 (-fdump-analyzer-supergraph): New warning.
2082 * doc/sourcebuild.texi (dg-require-dot): New.
2083 (dg-check-dot): New.
2084 * gdbinit.in (break-on-saved-diagnostic): New command.
2085 * graphviz.cc: New file.
2086 * graphviz.h: New file.
2087 * ordered-hash-map-tests.cc: New file.
2088 * ordered-hash-map.h: New file.
2089 * passes.def (pass_analyzer): Add before
2090 pass_ipa_whole_program_visibility.
2091 * selftest-run-tests.c (selftest::run_tests): Call
2092 selftest::ordered_hash_map_tests_cc_tests.
2093 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2095 * shortest-paths.h: New file.
2096 * timevar.def (TV_ANALYZER): New timevar.
2097 (TV_ANALYZER_SUPERGRAPH): Likewise.
2098 (TV_ANALYZER_STATE_PURGE): Likewise.
2099 (TV_ANALYZER_PLAN): Likewise.
2100 (TV_ANALYZER_SCC): Likewise.
2101 (TV_ANALYZER_WORKLIST): Likewise.
2102 (TV_ANALYZER_DUMP): Likewise.
2103 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2104 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2105 * tree-pass.h (make_pass_analyzer): New decl.
2106 * tristate.cc: New file.
2107 * tristate.h: New file.
2109 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2112 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2113 alternatives 9 and 10.
2115 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2117 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2118 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2119 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2120 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2121 (selftest::hash_map_tests_c_tests): Call it.
2122 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2123 New static constant, using the value of = H::empty_zero_p.
2124 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2125 from default_hash_traits <Value>.
2126 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2128 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2129 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2130 calls to mark_empty with !Descriptor::empty_zero_p.
2131 (hash_table::empty_slow): Conditionalize the memset call with a
2132 check that Descriptor::empty_zero_p; otherwise, loop through the
2133 entries calling mark_empty on them.
2134 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2135 (pointer_hash::empty_zero_p): Likewise.
2136 (pair_hash::empty_zero_p): Likewise.
2137 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2139 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2140 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2141 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2142 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2143 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2144 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2145 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2146 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2148 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2151 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2153 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2154 fix typo on return value.
2156 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2159 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2161 (cgraph_edge::make_speculative): Add param for setting speculative_id
2163 (cgraph_edge::speculative_call_info): Update comments and find reference
2164 by speculative_id for multiple indirect targets.
2165 (cgraph_edge::resolve_speculation): Decrease the speculations
2166 for indirect edge, drop it's speculative if not direct target
2167 left. Update comments.
2168 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2169 (cgraph_node::dump): Print num_speculative_call_targets.
2170 (cgraph_node::verify_node): Don't report error if speculative
2171 edge not include statement.
2172 (cgraph_edge::num_speculative_call_targets_p): New function.
2173 * cgraph.h (int common_target_id): Remove.
2174 (int common_target_probability): Remove.
2175 (num_speculative_call_targets): New variable.
2176 (make_speculative): Add param for setting speculative_id.
2177 (cgraph_edge::num_speculative_call_targets_p): New declare.
2178 (target_prob): New variable.
2179 (speculative_id): New variable.
2180 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2181 call summaries for multiple speculative call targets.
2182 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2183 * ipa-profile.c (struct speculative_call_target): New struct.
2184 (class speculative_call_summary): New class.
2185 (class speculative_call_summaries): New class.
2186 (call_sums): New variable.
2187 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2188 (ipa_profile_write_edge_summary): New function.
2189 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2190 (ipa_profile_dump_all_summaries): New function.
2191 (ipa_profile_read_edge_summary): New function.
2192 (ipa_profile_read_summary_section): New function.
2193 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2194 (ipa_profile): Generate num_speculative_call_targets from
2196 * ipa-ref.h (speculative_id): New variable.
2197 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2198 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2199 common_target_probability. Stream out speculative_id and
2200 num_speculative_call_targets.
2201 (input_edge): Likewise.
2202 * predict.c (dump_prediction): Remove edges count assert to be
2204 * symtab.c (symtab_node::create_reference): Init speculative_id.
2205 (symtab_node::clone_references): Clone speculative_id.
2206 (symtab_node::clone_referring): Clone speculative_id.
2207 (symtab_node::clone_reference): Clone speculative_id.
2208 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2209 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2210 if indirect call contains multiple speculative targets.
2211 * value-prof.h (check_ic_target): Remove.
2212 * value-prof.c (gimple_value_profile_transformations):
2213 Use void function gimple_ic_transform.
2214 * value-prof.c (gimple_ic_transform): Handle topn case.
2215 Fix comment typos. Change it to a void function.
2217 2020-01-13 Andrew Pinski <apinski@marvell.com>
2219 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2220 (octeontx2t98): New define.
2221 (octeontx2t96): New define.
2222 (octeontx2t93): New define.
2223 (octeontx2f95): New define.
2224 (octeontx2f95n): New define.
2225 (octeontx2f95mm): New define.
2226 * config/aarch64/aarch64-tune.md: Regenerate.
2227 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2229 2020-01-13 Jason Merrill <jason@redhat.com>
2231 PR c++/33799 - destroy return value if local cleanup throws.
2232 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2234 2020-01-13 Martin Liska <mliska@suse.cz>
2236 * ipa-cp.c (get_max_overall_size): Use newly
2237 renamed param param_ipa_cp_unit_growth.
2238 * params.opt: Remove legacy param name.
2240 2020-01-13 Martin Sebor <msebor@redhat.com>
2242 PR tree-optimization/93213
2243 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2244 stores to be eliminated.
2246 2020-01-13 Martin Liska <mliska@suse.cz>
2248 * opts.c (print_help): Do not print CL_PARAM
2249 and CL_WARNING for CL_OPTIMIZATION.
2251 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2254 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2255 depending on optimization settings.
2257 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2259 PR tree-optimization/90838
2260 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2261 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2262 argument rather than to initialize temporary for targets that
2263 don't use the mode argument at all. Initialize ctzval to avoid
2266 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2268 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2269 * tree-core.h: Document it.
2270 * gimplify.c (gimplify_omp_workshare): Set it.
2271 * omp-low.c (lower_omp_target): Use it.
2272 * tree-pretty-print.c (dump_omp_clause): Print it.
2274 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2275 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2277 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2279 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2280 * common.opt (fdiagnostics-path-format=): New option.
2281 (diagnostic_path_format): New enum.
2282 (fdiagnostics-show-path-depths): New option.
2283 * coretypes.h (diagnostic_event_id_t): New forward decl.
2284 * diagnostic-color.c (color_dict): Add "path".
2285 * diagnostic-event-id.h: New file.
2286 * diagnostic-format-json.cc (json_from_expanded_location): Make
2288 (json_end_diagnostic): Call context->make_json_for_path if it
2289 exists and the diagnostic has a path.
2290 (diagnostic_output_format_init): Clear context->print_path.
2291 * diagnostic-path.h: New file.
2292 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2293 when printing a run of events in a diagnostic_path so that they
2294 all get the same color.
2295 (layout::m_diagnostic_path_p): New field.
2296 (layout::layout): Initialize it.
2297 (layout::print_any_labels): Don't colorize the label text for an
2298 event in a diagnostic_path.
2299 (gcc_rich_location::add_location_if_nearby): Add
2300 "restrict_to_current_line_spans" and "label" params. Pass the
2301 former to layout.maybe_add_location_range; pass the latter
2302 when calling add_range.
2303 * diagnostic.c: Include "diagnostic-path.h".
2304 (diagnostic_initialize): Initialize context->path_format and
2305 context->show_path_depths.
2306 (diagnostic_show_any_path): New function.
2307 (diagnostic_path::interprocedural_p): New function.
2308 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2309 (simple_diagnostic_path::num_events): New function.
2310 (simple_diagnostic_path::get_event): New function.
2311 (simple_diagnostic_path::add_event): New function.
2312 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2313 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2314 (debug): New overload taking a diagnostic_path *.
2315 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2316 * diagnostic.h (enum diagnostic_path_format): New enum.
2317 (json::value): New forward decl.
2318 (diagnostic_context::path_format): New field.
2319 (diagnostic_context::show_path_depths): New field.
2320 (diagnostic_context::print_path): New callback field.
2321 (diagnostic_context::make_json_for_path): New callback field.
2322 (diagnostic_show_any_path): New decl.
2323 (json_from_expanded_location): New decl.
2324 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2325 (-fdiagnostics-show-path-depths): New option.
2326 (-fdiagnostics-color): Add "path" to description of default
2327 GCC_COLORS; describe it.
2328 (-fdiagnostics-format=json): Document how diagnostic paths are
2329 represented in the JSON output format.
2330 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2331 Add optional params "restrict_to_current_line_spans" and "label".
2332 * opts.c (common_handle_option): Handle
2333 OPT_fdiagnostics_path_format_ and
2334 OPT_fdiagnostics_show_path_depths.
2335 * pretty-print.c: Include "diagnostic-event-id.h".
2336 (pp_format): Implement "%@" format code for printing
2337 diagnostic_event_id_t *.
2338 (selftest::test_pp_format): Add tests for "%@".
2339 * selftest-run-tests.c (selftest::run_tests): Call
2340 selftest::tree_diagnostic_path_cc_tests.
2341 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2342 * toplev.c (general_init): Initialize global_dc->path_format and
2343 global_dc->show_path_depths.
2344 * tree-diagnostic-path.cc: New file.
2345 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2346 non-static. Drop "diagnostic" param in favor of storing the
2347 original value of "where" and re-using it.
2348 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2349 maybe_unwind_expanded_macro_loc.
2350 (tree_diagnostics_defaults): Initialize context->print_path and
2351 context->make_json_for_path.
2352 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2354 (default_tree_make_json_for_path): New decl.
2355 (maybe_unwind_expanded_macro_loc): New decl.
2357 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2359 PR tree-optimization/93210
2360 * fold-const.h (native_encode_initializer,
2361 can_native_interpret_type_p): Declare.
2362 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2364 (native_encode_initializer): New function, moved from dwarf2out.c.
2365 Adjust to native_encode_expr compatible arguments, including dry-run
2366 and partial extraction modes. Don't handle STRING_CST.
2367 (can_native_interpret_type_p): No longer static.
2368 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2369 offset / BITS_PER_UNIT fits into int and don't call it if
2370 can_native_interpret_type_p fails. If suboff is NULL and for
2371 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2372 native_encode_initializer.
2373 (fold_const_aggregate_ref_1): Formatting fix.
2374 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2375 (tree_add_const_value_attribute): Adjust caller.
2377 PR tree-optimization/90838
2378 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2379 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2380 CTZ_DEFINED_VALUE_AT_ZERO.
2382 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2385 * lra-constraints.c (match_reload): Permit input operands have the
2386 same mode as output while other input operands have a different
2389 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2391 PR tree-optimization/90838
2392 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2393 (check_ctz_string): Likewise.
2394 (optimize_count_trailing_zeroes): Likewise.
2395 (simplify_count_trailing_zeroes): Likewise.
2396 (pass_forwprop::execute): Try ctz simplification.
2397 * match.pd: Add matching for ctz idioms.
2399 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2401 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2403 (aarch64_invalid_unary_op): New function for target hook.
2404 (aarch64_invalid_binary_op): New function for target hook.
2406 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2408 * config.gcc: Add arm_bf16.h.
2409 * config/aarch64/aarch64-builtins.c
2410 (aarch64_simd_builtin_std_type): Add BFmode.
2411 (aarch64_init_simd_builtin_types): Define element types for vector
2413 (aarch64_init_bf16_types): New function.
2414 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2415 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2417 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2418 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2420 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2421 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2422 * config/aarch64/aarch64.c
2423 (aarch64_classify_vector_mode): Add support for BF types.
2424 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2425 (aarch64_vq_mode): Add support for BF types.
2426 (aarch64_simd_container_mode): Add support for BF types.
2427 (aarch64_mangle_type): Add support for BF scalar type.
2428 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2429 * config/aarch64/arm_bf16.h: New file.
2430 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2431 * config/aarch64/iterators.md: Add BF types to mode attributes.
2432 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2434 2020-01-10 Jason Merrill <jason@redhat.com>
2436 PR c++/93173 - incorrect tree sharing.
2437 * gimplify.c (copy_if_shared): No longer static.
2438 * gimplify.h: Declare it.
2440 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2442 * doc/invoke.texi (-msve-vector-bits=): Document that
2443 -msve-vector-bits=128 now generates VL-specific code for
2444 little-endian targets.
2445 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2446 build_vector_type_for_mode to construct the data vector types.
2447 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2448 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2449 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2450 for 128-bit vectors.
2452 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2454 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2457 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2459 * config/aarch64/aarch64-builtins.c
2460 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2461 rather than checking the number of elements and the element mode.
2463 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2465 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2466 get_related_vectype_for_scalar_type rather than build_vector_type
2467 to create the index type for a conditional reduction.
2469 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2471 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2472 for any type of gather or scatter, including strided accesses.
2474 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2476 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2479 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2481 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2483 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2484 parameter and its use to reset DR_OFFSET's.
2485 (vect_transform_loop): Remove orig_drs_init argument.
2486 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2487 member of dr_vec_info rather than the offset of the associated
2488 data_reference's innermost_loop_behavior.
2489 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2490 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2491 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2492 get_dr_vinfo_offset.
2493 (vectorizable_store): Likewise.
2494 (vectorizable_load): Likewise.
2496 2020-01-10 Richard Biener <rguenther@suse.de>
2498 * gimple-ssa-store-merging
2499 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
2501 2020-01-10 Martin Liska <mliska@suse.cz>
2504 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
2505 encapsulation that was there before r280040.
2507 2020-01-10 Richard Biener <rguenther@suse.de>
2510 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
2511 sequences to avoid walking them again for secondary opportunities.
2512 (pass_lower_eh_dispatch::execute): Instead actually insert
2515 2020-01-10 Richard Biener <rguenther@suse.de>
2518 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
2519 (cleanup_all_empty_eh): Walk landing pads in reverse order to
2520 avoid quadraticness.
2522 2020-01-10 Martin Jambor <mjambor@suse.cz>
2524 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
2525 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
2526 to get param_ipa_sra_max_replacements.
2527 (param_splitting_across_edge): Pass the caller to
2528 pull_accesses_from_callee.
2530 2020-01-10 Martin Jambor <mjambor@suse.cz>
2532 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
2533 * ipa-cp.c (max_new_size): Removed.
2534 (orig_overall_size): New variable.
2535 (get_max_overall_size): New function.
2536 (estimate_local_effects): Use it. Adjust dump.
2537 (decide_about_value): Likewise.
2538 (ipcp_propagate_stage): Do not calculate max_new_size, just store
2539 orig_overall_size. Adjust dump.
2540 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
2542 2020-01-10 Martin Jambor <mjambor@suse.cz>
2544 * params.opt (param_ipa_max_agg_items): Mark as Optimization
2545 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
2546 instead of param_ipa_max_agg_items.
2547 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
2548 optimization info for the callee.
2550 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
2552 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
2553 markers if debug_inline_points is false.
2555 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2557 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
2559 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
2560 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
2561 aarch64-sve-builtins-sve2.h.
2562 (aarch64-sve-builtins-sve2.o): New rule.
2563 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
2564 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
2565 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
2566 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
2567 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
2568 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
2570 * config/aarch64/aarch64-sve.md: Update comments with SVE2
2571 instructions that are handled here.
2572 (@cond_asrd<mode>): Generalize to...
2573 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
2574 (*cond_asrd<mode>_2): Generalize to...
2575 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
2576 (*cond_asrd<mode>_z): Generalize to...
2577 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
2578 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
2579 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
2580 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
2581 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
2583 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2584 (@aarch64_scatter_stnt<mode>): Likewise.
2585 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
2586 (@aarch64_mul_lane_<mode>): Likewise.
2587 (@aarch64_sve_suqadd<mode>_const): Likewise.
2588 (*<sur>h<addsub><mode>): Generalize to...
2589 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
2591 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
2592 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
2593 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
2594 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
2595 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
2596 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
2597 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
2598 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
2599 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
2600 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
2601 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
2602 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
2603 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
2604 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
2605 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
2606 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
2607 (@aarch64_sve2_xar<mode>): Likewise.
2608 (@aarch64_sve2_bcax<mode>): Likewise.
2609 (*aarch64_sve2_eor3<mode>): Rename to...
2610 (@aarch64_sve2_eor3<mode>): ...this.
2611 (@aarch64_sve2_bsl<mode>): New expander.
2612 (@aarch64_sve2_nbsl<mode>): Likewise.
2613 (@aarch64_sve2_bsl1n<mode>): Likewise.
2614 (@aarch64_sve2_bsl2n<mode>): Likewise.
2615 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
2616 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
2617 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
2618 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
2619 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
2620 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
2621 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
2622 (<su>mull<bt><Vwide>): Generalize to...
2623 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
2625 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
2626 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
2627 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
2628 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2629 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
2630 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2631 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
2632 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2633 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
2634 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
2635 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
2636 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
2637 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
2638 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
2639 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
2640 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
2641 (<SHRNB:r>shrnb<mode>): Generalize to...
2642 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
2644 (<SHRNT:r>shrnt<mode>): Generalize to...
2645 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
2647 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
2648 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
2649 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
2650 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
2651 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
2652 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
2653 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
2654 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
2655 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
2656 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
2657 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
2658 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
2659 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
2660 (@aarch64_sve2_cvtnt<mode>): Likewise.
2661 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
2662 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
2663 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
2664 (@aarch64_sve2_cvtxnt<mode>): Likewise.
2665 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
2666 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
2667 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
2668 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
2669 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
2670 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
2671 (@aarch64_sve2_pmul<mode>): Likewise.
2672 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
2673 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
2674 (@aarch64_sve2_tbl2<mode>): Likewise.
2675 (@aarch64_sve2_tbx<mode>): Likewise.
2676 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
2677 (@aarch64_sve2_histcnt<mode>): Likewise.
2678 (@aarch64_sve2_histseg<mode>): Likewise.
2679 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
2680 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
2681 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
2682 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
2683 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
2684 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
2685 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
2686 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
2687 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
2688 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
2689 (SVE2_PMULL_PAIR_I): New mode iterators.
2690 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
2691 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
2692 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
2693 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
2694 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
2695 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
2696 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
2697 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
2698 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
2699 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
2700 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
2701 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
2702 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
2703 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
2704 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
2705 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
2706 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
2707 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
2708 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
2709 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
2710 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
2711 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
2712 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
2713 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
2714 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
2715 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
2716 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
2717 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
2718 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
2719 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
2720 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
2722 (VNARROW, Ventype): New mode attributes.
2723 (Vewtype): Handle VNx2DI. Fix typo in comment.
2724 (VDOUBLE): New mode attribute.
2725 (sve_lane_con): Handle VNx8HI.
2726 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
2727 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
2728 (sve_int_op, sve_int_op_rev): Handle the above codes.
2729 (sve_pred_int_rhs2_operand): Likewise.
2730 (MULLBT, SHRNB, SHRNT): Delete.
2731 (SVE_INT_SHIFT_IMM): New int iterator.
2732 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
2733 and UNSPEC_WHILEHS for TARGET_SVE2.
2734 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
2735 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
2736 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
2737 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
2738 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
2739 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
2740 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
2741 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
2742 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
2743 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
2744 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
2745 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
2746 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
2747 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
2748 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
2749 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
2750 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
2751 (optab): Handle the new unspecs.
2752 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
2754 (lr): Handle the new unspecs.
2756 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
2757 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
2758 (sve_int_qsub_op): New int attributes.
2759 (sve_fp_op, rot): Handle the new unspecs.
2760 * config/aarch64/aarch64-sve-builtins.h
2761 (function_resolver::require_matching_pointer_type): Declare.
2762 (function_resolver::resolve_unary): Add an optional boolean argument.
2763 (function_resolver::finish_opt_n_resolution): Add an optional
2764 type_suffix_index argument.
2765 (gimple_folder::redirect_call): Declare.
2766 (gimple_expander::prepare_gather_address_operands): Add an optional
2768 * config/aarch64/aarch64-sve-builtins.cc: Include
2769 aarch64-sve-builtins-sve2.h.
2770 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
2771 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
2772 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
2773 (TYPES_hsd_integer): Use TYPES_hsd_signed.
2774 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
2775 (TYPES_s_unsigned): Likewise.
2776 (TYPES_s_integer): Use TYPES_s_unsigned.
2777 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
2778 (TYPES_sd_integer): Use them.
2779 (TYPES_d_unsigned): New macro.
2780 (TYPES_d_integer): Use it.
2781 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
2782 (TYPES_cvt_narrow): Likewise.
2783 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
2784 (preds_mx): New variable.
2785 (function_builder::add_overloaded_function): Allow the new feature
2786 set to be more restrictive than the original one.
2787 (function_resolver::infer_pointer_type): Remove qualifiers from
2788 the pointer type before printing it.
2789 (function_resolver::require_matching_pointer_type): New function.
2790 (function_resolver::resolve_sv_displacement): Handle functions
2791 that don't support 32-bit vector indices or svint32_t vector offsets.
2792 (function_resolver::finish_opt_n_resolution): Take the inferred type
2793 as a separate argument.
2794 (function_resolver::resolve_unary): Optionally treat all forms in
2795 the same way as normal merging functions.
2796 (gimple_folder::redirect_call): New function.
2797 (function_expander::prepare_gather_address_operands): Add an argument
2798 that says whether scaled forms are available. If they aren't,
2799 handle scaling of vector indices and don't add the extension and
2801 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
2802 fall back to using cond_* instead.
2803 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
2804 Split out the member variables into...
2805 (rtx_code_function_base): ...this new base class.
2806 (rtx_code_function_rotated): Inherit rtx_code_function_base.
2807 (unspec_based_function): Split out the member variables into...
2808 (unspec_based_function_base): ...this new base class.
2809 (unspec_based_function_rotated): Inherit unspec_based_function_base.
2810 (unspec_based_function_exact_insn): New class.
2811 (unspec_based_add_function, unspec_based_add_lane_function)
2812 (unspec_based_lane_function, unspec_based_pred_function)
2813 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
2814 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
2815 (unspec_based_sub_function, unspec_based_sub_lane_function): New
2817 (unspec_based_fused_function): New class.
2818 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
2819 (unspec_based_fused_lane_function): New class.
2820 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
2822 (CODE_FOR_MODE1): New macro.
2823 (fixed_insn_function): New class.
2824 (while_comparison): Likewise.
2825 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
2826 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
2827 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
2828 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
2829 (load_gather_sv_restricted, shift_left_imm_long): Declare.
2830 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
2831 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
2832 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
2833 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
2834 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
2835 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
2836 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
2837 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
2838 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
2839 Also add an initial argument for unary_convert_narrowt, regardless
2840 of the predication type.
2841 (build_32_64): Allow loads and stores to specify MODE_none.
2842 (build_sv_index64, build_sv_uint_offset): New functions.
2843 (long_type_suffix): New function.
2844 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
2845 (binary_imm_long_base, load_gather_sv_base): Likewise.
2846 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
2847 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
2848 (unary_narrowb_base, unary_narrowt_base): Likewise.
2849 (binary_long_lane_def, binary_long_lane): New shape.
2850 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
2851 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
2852 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
2853 (binary_to_uint_def, binary_to_uint): Likewise.
2854 (binary_wide_def, binary_wide): Likewise.
2855 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
2856 (compare_def, compare): Likewise.
2857 (compare_ptr_def, compare_ptr): Likewise.
2858 (load_ext_gather_index_restricted_def,
2859 load_ext_gather_index_restricted): Likewise.
2860 (load_ext_gather_offset_restricted_def,
2861 load_ext_gather_offset_restricted): Likewise.
2862 (load_gather_sv_def): Inherit from load_gather_sv_base.
2863 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
2864 (shift_left_imm_def, shift_left_imm): Likewise.
2865 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
2866 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
2867 (store_scatter_index_restricted_def,
2868 store_scatter_index_restricted): Likewise.
2869 (store_scatter_offset_restricted_def,
2870 store_scatter_offset_restricted): Likewise.
2871 (tbl_tuple_def, tbl_tuple): Likewise.
2872 (ternary_long_lane_def, ternary_long_lane): Likewise.
2873 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
2874 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
2875 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
2876 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
2877 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
2878 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
2879 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
2880 (ternary_uint_def, ternary_uint): Likewise.
2881 (unary_convert): Fix typo in comment.
2882 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
2883 (unary_long_def, unary_long): Likewise.
2884 (unary_narrowb_def, unary_narrowb): Likewise.
2885 (unary_narrowt_def, unary_narrowt): Likewise.
2886 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
2887 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
2888 (unary_to_int_def, unary_to_int): Likewise.
2889 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
2890 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
2891 (svasrd_impl): Delete.
2892 (svcadd_impl::expand): Handle integer operations too.
2893 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
2894 new functions to derive the unspec numbers.
2895 (svmla_svmls_lane_impl): Replace with...
2896 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
2897 integer operations too.
2898 (svwhile_impl): Rename to...
2899 (svwhilelx_impl): ...this and inherit from while_comparison.
2900 (svasrd): Use unspec_based_function.
2901 (svmla_lane): Use svmla_lane_impl.
2902 (svmls_lane): Use svmls_lane_impl.
2903 (svrecpe, svrsqrte): Handle unsigned integer operations too.
2904 (svwhilele, svwhilelt): Use svwhilelx_impl.
2905 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
2906 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
2907 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
2908 * config/aarch64/aarch64-sve-builtins.def: Include
2909 aarch64-sve-builtins-sve2.def.
2911 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2913 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
2914 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
2915 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
2916 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
2917 immediates as well as vector ones.
2918 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
2919 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
2920 (aarch64_sve_qsub_immediate): Update calls accordingly.
2922 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2924 * config/aarch64/aarch64-sve2.md: Add banner comments.
2925 (<su>mulh<r>s<mode>3): Move further up file.
2926 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
2927 (*aarch64_sve2_sra<mode>): Move further down file.
2928 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
2930 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2932 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
2934 (while_optab_cmp): Handle them.
2935 * config/aarch64/aarch64-sve.md
2936 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
2937 and add a "@" marker.
2938 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
2939 instead of gen_aarch64_sve2_while_ptest.
2940 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
2942 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2944 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
2945 (UNSPEC_WHILELE): ...this.
2946 (UNSPEC_WHILE_LO): Rename to...
2947 (UNSPEC_WHILELO): ...this.
2948 (UNSPEC_WHILE_LS): Rename to...
2949 (UNSPEC_WHILELS): ...this.
2950 (UNSPEC_WHILE_LT): Rename to...
2951 (UNSPEC_WHILELT): ...this.
2952 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
2953 (cmp_op, while_optab_cmp): Likewise.
2954 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
2955 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
2956 (svwhilelt): Likewise.
2958 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2960 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
2961 (unary_to_uint): Define.
2962 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
2963 (unary_count): Rename to...
2964 (unary_to_uint_def, unary_to_uint): ...this.
2965 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
2967 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2969 * config/aarch64/aarch64-sve-builtins-functions.h
2970 (code_for_mode_function): New class.
2971 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
2972 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
2973 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
2974 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
2975 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
2977 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2979 * config/aarch64/iterators.md (addsub): New code attribute.
2980 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
2982 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
2983 in the asm string and attributes. Fix indentation.
2984 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
2986 (@aarch64_sve_<optab><mode>): ...this.
2987 * config/aarch64/aarch64-sve-builtins.h
2988 (function_expander::expand_signed_unpred_op): Delete.
2989 * config/aarch64/aarch64-sve-builtins.cc
2990 (function_expander::expand_signed_unpred_op): Likewise.
2991 (function_expander::map_to_rtx_codes): If the optab isn't defined,
2992 try using code_for_aarch64_sve instead.
2993 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
2994 (svqsub_impl): Likewise.
2995 (svqadd, svqsub): Use rtx_code_function instead.
2997 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
2999 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3000 (HADDSUB, sur, addsub): Remove them.
3002 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3004 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3005 functions the same way, rather than singling out those that
3006 aren't mapped directly to optabs.
3008 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3010 * target.def (compatible_vector_types_p): New target hook.
3011 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3012 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3013 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3014 * doc/tm.texi: Regenerate.
3015 * gimple-expr.c: Include target.h.
3016 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3017 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3019 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3020 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3021 Use the original predicate if it already has a suitable type.
3023 2020-01-09 Martin Jambor <mjambor@suse.cz>
3025 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3026 resolve_speculation and redirect_call_stmt_to_callee static. Change
3027 return type of set_call_stmt to cgraph_edge *.
3028 * auto-profile.c (afdo_indirect_call): Adjust call to
3029 redirect_call_stmt_to_callee.
3030 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3031 make the this pointer explicit, adjust self-recursive calls and the
3032 call top make_direct. Return the resulting edge.
3033 (cgraph_edge::remove): Make this pointer explicit.
3034 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3035 (cgraph_edge::make_direct): Likewise, adjust call to
3036 resolve_speculation.
3037 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3038 call to set_call_stmt.
3039 (cgraph_update_edges_for_call_stmt_node): Update call to
3040 set_call_stmt and remove.
3041 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3042 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3043 (cgraph_node::create_edge_including_clones): Moved "first" definition
3044 of edge to the block where it was used. Adjusted calls to
3046 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3047 cgraph_edge::remove.
3048 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3049 make_direct and redirect_call_stmt_to_callee.
3050 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3051 resolve_speculation and make_direct.
3052 * ipa-inline-transform.c (inline_transform): Adjust call to
3053 redirect_call_stmt_to_callee.
3054 (check_speculations_1):: Adjust call to resolve_speculation.
3055 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3056 resolve-speculation.
3057 (inline_small_functions): Adjust call to resolve_speculation.
3058 (ipa_inline): Likewise.
3059 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3061 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3062 safe with regards to edge removal, adjust calls to
3063 redirect_call_stmt_to_callee.
3064 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3065 and redirect_call_stmt_to_callee.
3066 * multiple_target.c (create_dispatcher_calls): Adjust call to
3067 redirect_call_stmt_to_callee
3068 (redirect_to_specific_clone): Likewise.
3069 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3070 Adjust calls to cgraph_edge::remove.
3071 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3072 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3073 (expand_call_inline): Adjust call to cgraph_edge::remove.
3075 2020-01-09 Martin Liska <mliska@suse.cz>
3077 * params.opt: Set Optimization for
3078 param_max_speculative_devirt_maydefs.
3080 2020-01-09 Martin Sebor <msebor@redhat.com>
3084 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3086 2020-01-09 Martin Liska <mliska@suse.cz>
3088 * auto-profile.c (auto_profile): Use opt_for_fn
3090 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3091 (propagate_vals_across_arith_jfunc): Likewise.
3092 (hint_time_bonus): Likewise.
3093 (incorporate_penalties): Likewise.
3094 (good_cloning_opportunity_p): Likewise.
3095 (perform_estimation_of_a_value): Likewise.
3096 (estimate_local_effects): Likewise.
3097 (ipcp_propagate_stage): Likewise.
3098 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3099 (set_switch_stmt_execution_predicate): Likewise.
3100 (analyze_function_body): Likewise.
3101 * ipa-inline-analysis.c (offline_size): Likewise.
3102 * ipa-inline.c (early_inliner): Likewise.
3103 * ipa-prop.c (ipa_analyze_node): Likewise.
3104 (ipcp_transform_function): Likewise.
3105 * ipa-sra.c (process_scan_results): Likewise.
3106 (ipa_sra_summarize_function): Likewise.
3107 * params.opt: Rename ipcp-unit-growth to
3108 ipa-cp-unit-growth. Add Optimization for various
3109 IPA-related parameters.
3111 2020-01-09 Richard Biener <rguenther@suse.de>
3114 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3116 2020-01-09 Richard Biener <rguenther@suse.de>
3118 PR tree-optimization/93040
3119 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3121 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3123 * common/config/avr/avr-common.c (avr_option_optimization_table)
3124 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3126 2020-01-09 Martin Liska <mliska@suse.cz>
3128 * cgraphclones.c (symbol_table::materialize_all_clones):
3129 Use cgraph_node::dump_name.
3131 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3134 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3135 output_operand_lossage instead of gcc_unreachable.
3136 * doc/md.texi (riscv f constraint): Fix typo.
3139 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3140 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3141 CONST_SCALAR_INT_P instead of CONST_INT_P.
3142 (*subv<mode>4_1): Rename to ...
3143 (subv<mode>4_1): ... this.
3144 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3145 define_insn_and_split patterns.
3146 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3149 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3151 * vec.c (class selftest::count_dtor): New class.
3152 (selftest::test_auto_delete_vec): New test.
3153 (selftest::vec_c_tests): Call it.
3154 * vec.h (class auto_delete_vec): New class template.
3155 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3157 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3159 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3161 2020-01-08 Jim Wilson <jimw@sifive.com>
3163 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3164 use of TLS_MODEL_LOCAL_EXEC when not pic.
3166 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3168 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3171 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3174 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3175 *stack_protect_set_3 peephole2): Also check that the second
3176 insns source is general_operand.
3179 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3180 predicate for output operand instead of register_operand.
3181 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3182 memory destination and non-memory operands[2].
3184 2020-01-08 Martin Liska <mliska@suse.cz>
3186 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3187 ::dump_asm_name instead of (::name or ::asm_name).
3188 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3189 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3190 (analyze_functions): Likewise.
3191 (expand_all_functions): Likewise.
3192 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3193 (propagate_bits_across_jump_function): Likewise.
3194 (dump_profile_updates): Likewise.
3195 (ipcp_store_bits_results): Likewise.
3196 (ipcp_store_vr_results): Likewise.
3197 * ipa-devirt.c (dump_targets): Likewise.
3198 * ipa-fnsummary.c (analyze_function_body): Likewise.
3199 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3200 (process_hsa_functions): Likewise.
3201 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3202 (set_alias_uids): Likewise.
3203 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3204 * ipa-inline.c (recursive_inlining): Likewise.
3205 (inline_to_all_callers_1): Likewise.
3206 (ipa_inline): Likewise.
3207 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3208 (ipa_propagate_frequency): Likewise.
3209 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3210 (remove_described_reference): Likewise.
3211 * ipa-pure-const.c (worse_state): Likewise.
3212 (check_retval_uses): Likewise.
3213 (analyze_function): Likewise.
3214 (propagate_pure_const): Likewise.
3215 (propagate_nothrow): Likewise.
3216 (dump_malloc_lattice): Likewise.
3217 (propagate_malloc): Likewise.
3218 (pass_local_pure_const::execute): Likewise.
3219 * ipa-visibility.c (optimize_weakref): Likewise.
3220 (function_and_variable_visibility): Likewise.
3221 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3222 (ipa_discover_variable_flags): Likewise.
3223 * lto-streamer-out.c (output_function): Likewise.
3224 (output_constructor): Likewise.
3225 * tree-inline.c (copy_bb): Likewise.
3226 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3227 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3229 2020-01-08 Richard Biener <rguenther@suse.de>
3232 * tree-eh.c (sink_clobbers): Update virtual operands for
3233 the first and last stmt only. Add a dry-run capability.
3234 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3235 after CFG manipulations and in RPO order to catch all
3236 secondary opportunities reliably.
3238 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3241 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3243 2019-01-08 Richard Biener <rguenther@suse.de>
3246 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3247 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3248 virtual operand, also updating SSA use.
3249 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3250 Update stmt after resetting virtual operand.
3251 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3252 * gimple-iterator.c (gsi_remove): When not removing the stmt
3253 permanently do not delink immediate uses or mark the stmt modified.
3255 2020-01-08 Martin Liska <mliska@suse.cz>
3257 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3258 (ipa_call_context::estimate_size_and_time): Likewise.
3259 (inline_analyze_function): Likewise.
3261 2020-01-08 Martin Liska <mliska@suse.cz>
3263 * cgraph.c (cgraph_node::dump): Use systematically
3266 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3268 Add -nodevicespecs option for avr.
3271 * config/avr/avr.opt (-nodevicespecs): New driver option.
3272 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3273 "-specs=device-specs/..." if that option is not set.
3274 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3276 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3278 Implement 64-bit double functions for avr.
3281 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3282 --with-double-comparison.
3283 * doc/install.texi: Document them.
3284 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3285 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3286 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3287 * doc/invoke.texi (AVR Built-in Macros): Document them.
3288 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3289 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3290 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3292 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3295 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3296 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3297 when only building rm-profile multilibs.
3299 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3302 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3303 lattice for a value to check.
3304 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3305 finite propagation in self-recursive scc.
3307 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3309 * ipa-inline.c (caller_growth_limits): Restore the AND.
3311 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3313 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3314 (VEC_ALLREG_ALT): New iterator.
3315 (VEC_ALLREG_INT_MODE): New iterator.
3316 (VCMP_MODE): New iterator.
3317 (VCMP_MODE_INT): New iterator.
3318 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3319 (vec_cmp<u>v64qidi): New define_expand.
3320 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3321 (vec_cmpu<mode>di_exec): New define_expand.
3322 (vec_cmp<u>v64qidi_exec): New define_expand.
3323 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3324 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3325 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3326 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3327 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3328 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3329 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3330 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3331 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3332 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3334 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3335 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3337 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3339 * config/gcn/constraints.md (DA): Update description and match.
3341 (Db): New constraint.
3342 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3344 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3345 Implement 'Db' mixed immediate type.
3346 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3347 (addcv64si3_dup<exec_vcc>): Delete.
3348 (subcv64si3<exec_vcc>): Rework constraints.
3349 (addv64di3): Rework constraints.
3350 (addv64di3_exec): Rework constraints.
3351 (subv64di3): Rework constraints.
3352 (addv64di3_dup): Delete.
3353 (addv64di3_dup_exec): Delete.
3354 (addv64di3_zext): Rework constraints.
3355 (addv64di3_zext_exec): Rework constraints.
3356 (addv64di3_zext_dup): Rework constraints.
3357 (addv64di3_zext_dup_exec): Rework constraints.
3358 (addv64di3_zext_dup2): Rework constraints.
3359 (addv64di3_zext_dup2_exec): Rework constraints.
3360 (addv64di3_sext_dup2): Rework constraints.
3361 (addv64di3_sext_dup2_exec): Rework constraints.
3363 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3365 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3366 existing target checks.
3368 2020-01-07 Richard Biener <rguenther@suse.de>
3370 * doc/install.texi: Bump minimal supported MPC version.
3372 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3374 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3375 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3376 * langhooks.c: Include stor-layout.h.
3377 (lhd_simulate_enum_decl): New function.
3378 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3379 handle_arm_sve_h for the LTO frontend.
3380 (register_vector_type): Cope with null returns from pushdecl.
3382 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3384 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3385 (aarch64_sve::nvectors_if_data_type): Replace with...
3386 (aarch64_sve::builtin_type_p): ...this.
3387 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3388 (find_vector_type): Delete.
3389 (add_sve_type_attribute): New function.
3390 (lookup_sve_type_attribute): Likewise.
3391 (register_builtin_types): Add an "SVE type" attribute to each type.
3392 (register_tuple_type): Likewise.
3393 (svbool_type_p, nvectors_if_data_type): Delete.
3394 (mangle_builtin_type): Use lookup_sve_type_attribute.
3395 (builtin_type_p): Likewise. Add an overload that returns the
3396 number of constituent vector and predicate registers.
3397 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3398 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3399 instead of aarch64_sve_argument_p.
3400 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3401 (aarch64_pass_by_reference): Likewise.
3402 (aarch64_function_value_1): Likewise.
3403 (aarch64_return_in_memory): Likewise.
3404 (aarch64_layout_arg): Likewise.
3406 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3408 PR tree-optimization/93156
3409 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3410 least significant bit is always clear.
3412 PR tree-optimization/93118
3413 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3414 simplifier with two intermediate conversions.
3416 2020-01-07 Martin Liska <mliska@suse.cz>
3418 * params.opt: Add Optimization for various parameters.
3420 2020-01-07 Martin Liska <mliska@suse.cz>
3423 * doc/extend.texi: Explain cloning for target_clone
3426 2020-01-07 Martin Liska <mliska@suse.cz>
3428 PR tree-optimization/92860
3429 * common.opt: Make in Optimization option
3430 as it is affected by -O0, which is an Optimization
3432 * tree-inline.c (tree_inlinable_function_p):
3433 Use opt_for_fn for warn_inline.
3434 (expand_call_inline): Likewise.
3436 2020-01-07 Martin Liska <mliska@suse.cz>
3438 PR tree-optimization/92860
3439 * common.opt: Make flag_ree as optimization
3442 2020-01-07 Martin Liska <mliska@suse.cz>
3444 PR optimization/92860
3445 * params.opt: Mark param_min_crossjump_insns with Optimization
3448 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3450 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3451 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3453 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3455 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3456 helper function to return the valid addressing formats for a given
3457 hard register and mode.
3458 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3460 * config/rs6000/constraints.md (Q constraint): Update
3462 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3465 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3466 Use 'Q' for doing vector extract from memory.
3467 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3469 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3470 doing vector extract from memory.
3471 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3472 extract from memory.
3474 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3475 for the offset being 34-bits when -mcpu=future is used.
3477 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3479 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3480 instead of cmpib_comparison_operator in cmpib patterns.
3481 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3482 of cmpib_comparison_operator. Revise comment.
3484 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3486 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3487 in an IFN_DIV_POW2 node to be equal.
3489 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3491 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3492 (vect_check_scalar_mask): ...this.
3493 (vectorizable_store, vectorizable_load): Update call accordingly.
3494 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3495 argument in calls to conditional internal functions.
3497 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3499 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
3500 '0' matching inputs.
3501 (subv64di3_exec): Likewise.
3503 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
3505 * config/mips/mips.c (vr4130_align_insns): Fix typo.
3506 * doc/md.texi (movstr): Likewise.
3508 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3510 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
3513 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3515 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
3517 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
3518 to a temporary file and use move-if-change to update the real
3519 file where necessary.
3521 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3523 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
3524 rather than Upa for CPY /M.
3526 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3528 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
3531 2020-01-06 Martin Liska <mliska@suse.cz>
3533 PR tree-optimization/92860
3534 * params.opt: Mark param_max_combine_insns with Optimization
3537 2020-01-05 Jakub Jelinek <jakub@redhat.com>
3540 * config/i386/i386.md (SWIDWI): New mode iterator.
3541 (DWI, dwi): Add TImode variants.
3542 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
3543 <general_hilo_operand> instead of <general_operand>. Use
3544 CONST_SCALAR_INT_P instead of CONST_INT_P.
3545 (*addv<mode>4_1): Rename to ...
3546 (addv<mode>4_1): ... this.
3547 (QWI): New mode attribute.
3548 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3549 define_insn_and_split patterns.
3550 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3552 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
3553 <general_hilo_operand> instead of <general_operand>.
3554 (*addcarry<mode>_1): New define_insn.
3555 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
3557 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
3559 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
3560 Use "call" instead of "set".
3562 2020-01-03 Martin Jambor <mjambor@suse.cz>
3565 * ipa-cp.c (print_all_lattices): Skip functions without info.
3567 2020-01-03 Jakub Jelinek <jakub@redhat.com>
3570 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
3571 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
3572 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
3573 for 'e' simd clones.
3576 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
3578 (mprefer-vector-width=): Add Save.
3579 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
3580 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
3581 (ix86_debug_options, ix86_function_specific_print): Adjust
3582 ix86_target_string callers.
3583 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
3584 (ix86_valid_target_attribute_tree): Likewise.
3585 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
3586 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
3587 ix86_target_string caller.
3590 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
3591 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
3592 instead of gen_int_shift_amount + convert_modes.
3594 PR rtl-optimization/93088
3595 * loop-iv.c (find_single_def_src): Punt after looking through
3596 128 reg copies for regs with single definitions. Move definitions
3599 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
3601 * config/arm/arm-c.c (arm_cpu_builtins): Define
3602 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
3603 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
3604 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
3605 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
3606 * config/arm/arm-tables.opt: Regenerated.
3607 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
3608 arm_arch_i8mm and arm_arch_bf16 when enabled.
3609 * config/arm/arm.h (TARGET_I8MM): New macro.
3610 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
3611 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
3612 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
3613 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
3614 (v8_6_a_simd_variants): New.
3615 (v8_*_a_simd_variants): Add i8mm and bf16.
3616 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
3618 2020-01-02 Jakub Jelinek <jakub@redhat.com>
3621 * predict.c (compute_function_frequency): Don't call
3622 warn_function_cold on functions that already have cold attribute.
3624 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
3627 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
3628 COMDAT group function labels in .data.rel.ro.local section.
3629 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
3632 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
3633 comparison_operator in B and S integer comparisons. Likewise, use
3634 ordered_comparison_operator instead of cmpib_comparison_operator in
3636 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
3638 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3640 Update copyright years.
3642 * gcc.c (process_command): Update copyright notice dates.
3643 * gcov-dump.c (print_version): Ditto.
3644 * gcov.c (print_version): Ditto.
3645 * gcov-tool.c (print_version): Ditto.
3646 * gengtype.c (create_file): Ditto.
3647 * doc/cpp.texi: Bump @copying's copyright year.
3648 * doc/cppinternals.texi: Ditto.
3649 * doc/gcc.texi: Ditto.
3650 * doc/gccint.texi: Ditto.
3651 * doc/gcov.texi: Ditto.
3652 * doc/install.texi: Ditto.
3653 * doc/invoke.texi: Ditto.
3655 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
3657 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
3660 2020-01-01 Jakub Jelinek <jakub@redhat.com>
3662 PR tree-optimization/93098
3663 * match.pd (popcount): For shift amounts, use integer_onep
3664 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
3665 tests. Make sure that precision is power of two larger than or equal
3666 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
3667 instead of ULL suffixed constants. Formatting fixes.
3669 Copyright (C) 2020 Free Software Foundation, Inc.
3671 Copying and distribution of this file, with or without modification,
3672 are permitted in any medium without royalty provided the copyright
3673 notice and this notice are preserved.