1 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
4 * config/aarch64/aarch64.c (aarch64_override_options): Fix
5 misleading warning string.
7 2020-02-27 Martin Sebor <msebor@redhat.com>
9 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
11 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
14 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
15 Split the insn into two parts. This insn only does variable
16 extract from a register.
17 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
18 variable extract from memory.
19 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
20 only does variable extract from a register.
21 (vsx_extract_v4sf_var_load): New insn, do variable extract from
23 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
24 into two parts. This insn only does variable extract from a
26 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
27 do variable extract from memory.
29 2020-02-27 Martin Jambor <mjambor@suse.cz>
30 Feng Xue <fxue@os.amperecomputing.com>
33 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
34 new function calls_same_node_or_its_all_contexts_clone_p.
35 (cgraph_edge_brings_value_p): Use it.
36 (cgraph_edge_brings_value_p): Likewise.
37 (self_recursive_pass_through_p): Return false if caller is a clone.
38 (self_recursive_agg_pass_through_p): Likewise.
40 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
43 * alias.c (ends_tbaa_access_path_p): Break out from ...
44 (component_uses_parent_alias_set_from): ... here.
45 * alias.h (ends_tbaa_access_path_p): Declare.
46 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
47 handle trailing arrays past end of tbaa access path.
48 (aliasing_component_refs_p): ... here; likewise.
49 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
50 path; disambiguate also past end of it.
51 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
54 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
56 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
57 beginning of the file.
58 (vcreate_bf16, vcombine_bf16): New.
59 (vdup_n_bf16, vdupq_n_bf16): New.
60 (vdup_lane_bf16, vdup_laneq_bf16): New.
61 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
62 (vduph_lane_bf16, vduph_laneq_bf16): New.
63 (vset_lane_bf16, vsetq_lane_bf16): New.
64 (vget_lane_bf16, vgetq_lane_bf16): New.
65 (vget_high_bf16, vget_low_bf16): New.
66 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
67 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
68 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
69 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
70 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
71 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
72 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
73 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
74 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
75 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
76 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
77 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
78 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
79 (vreinterpretq_bf16_p128): New.
80 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
81 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
82 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
83 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
84 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
85 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
86 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
87 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
88 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
89 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
90 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
91 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
92 (vreinterpretq_p128_bf16): New.
93 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
99 (VDQX): Add V4BF and V8BF.
100 (V_two_elem, V_three_elem, V_four_elem): Likewise.
103 (V_double_vector_mode): Likewise.
104 (V_cmp_result): Likewise.
105 (V_uf_sclr): Likewise.
106 (V_sz_elem): Likewise.
107 (Is_d_reg): Likewise.
108 (V_mode_nunits): Likewise.
109 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
111 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
113 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
114 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
115 (<expander><mode>3<exec>): Likewise.
116 (<expander><mode>3): New.
117 (v<expander><mode>3): New.
118 (<expander><mode>3): New.
119 (<expander><mode>3<exec>): Rename to ...
120 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
121 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
123 2020-02-27 Alexandre Oliva <oliva@adacore.com>
125 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
128 2020-02-27 Richard Biener <rguenther@suse.de>
130 PR tree-optimization/93508
131 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
132 non-_CHK variants. Valueize their length arguments.
134 2020-02-27 Richard Biener <rguenther@suse.de>
136 PR tree-optimization/93953
137 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
138 to the hash-map entry.
140 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
142 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
144 2020-02-27 Mark Williams <mwilliams@fb.com>
146 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
147 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
148 -ffile-prefix-map and -fmacro-prefix-map.
149 * lto-streamer-out.c: Include file-prefix-map.h.
150 (lto_output_location): Remap the file part of locations.
152 2020-02-27 Jakub Jelinek <jakub@redhat.com>
155 * gimplify.c (gimplify_init_constructor): Don't promote readonly
156 DECL_REGISTER variables to TREE_STATIC.
158 PR tree-optimization/93582
159 PR tree-optimization/93945
160 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
161 non-zero INTEGER_CST second argument and ref->offset or ref->size
162 not a multiple of BITS_PER_UNIT.
164 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
166 * doc/install.texi (Binaries): Update description of BullFreeware.
168 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
172 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
173 C++ Language Options, Warning Options, and Static Analyzer
174 Options lists. Document negative form of options enabled by
175 default. Move some things around to more accurately sort
176 warnings by category.
177 (C++ Dialect Options, Warning Options, Static Analyzer
178 Options): Document negative form of options when enabled by
179 default. Move some things around to more accurately sort
180 warnings by category. Add some missing index entries.
183 2020-02-26 Carl Love <cel@us.ibm.com>
186 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
187 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
188 for the vector unsigned short arguments. It is also listed as the
189 name of the built-in for arguments vector unsigned short,
190 vector unsigned int and vector unsigned long long built-ins. The
191 name of the builtins for these arguments should be:
192 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
193 __builtin_crypto_vpmsumd respectively.
195 2020-02-26 Richard Biener <rguenther@suse.de>
197 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
198 and load permutation.
200 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
203 * optabs-tree.c (supportable_convert_operation): Reject types with
206 2020-02-26 David Malcolm <dmalcolm@redhat.com>
208 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
210 2020-02-26 Jakub Jelinek <jakub@redhat.com>
212 PR tree-optimization/93820
213 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
214 argument to ALL_INTEGER_CST_P boolean.
215 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
216 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
217 adjacent INTEGER_CST store into merged_store->only_constants like
220 2020-02-25 Jakub Jelinek <jakub@redhat.com>
223 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
225 * cfghooks.c (verify_flow_info): Likewise.
226 * predict.c (combine_predictions_for_bb): Likewise.
227 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
228 sucessor -> successor.
229 (find_traces_1_round): Fix comment typo, destinarion -> destination.
230 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
232 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
233 message typo, sucessors -> successors.
235 2020-02-25 Martin Sebor <msebor@redhat.com>
237 * doc/extend.texi (attribute access): Correct an example.
239 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
241 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
243 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
245 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
246 (VD): Enable for V4BF.
248 (VQ): Enable for V8BF.
251 (VDBL, Vdbl): Add V4BF.
252 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
253 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
254 (bfloat16x8x2_t): Likewise.
255 (bfloat16x4x3_t): Likewise.
256 (bfloat16x8x3_t): Likewise.
257 (bfloat16x4x4_t): Likewise.
258 (bfloat16x8x4_t): Likewise.
259 (vcombine_bf16): New.
260 (vld1_bf16, vld1_bf16_x2): New.
261 (vld1_bf16_x3, vld1_bf16_x4): New.
262 (vld1q_bf16, vld1q_bf16_x2): New.
263 (vld1q_bf16_x3, vld1q_bf16_x4): New.
264 (vld1_lane_bf16): New.
265 (vld1q_lane_bf16): New.
266 (vld1_dup_bf16): New.
267 (vld1q_dup_bf16): New.
270 (vld2_dup_bf16): New.
271 (vld2q_dup_bf16): New.
274 (vld3_dup_bf16): New.
275 (vld3q_dup_bf16): New.
278 (vld4_dup_bf16): New.
279 (vld4q_dup_bf16): New.
280 (vst1_bf16, vst1_bf16_x2): New.
281 (vst1_bf16_x3, vst1_bf16_x4): New.
282 (vst1q_bf16, vst1q_bf16_x2): New.
283 (vst1q_bf16_x3, vst1q_bf16_x4): New.
284 (vst1_lane_bf16): New.
285 (vst1q_lane_bf16): New.
293 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
295 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
296 (VALL_F16): Likewise.
297 (VALLDI_F16): Likewise.
300 (vswap_width_name): Likewise.
301 (VSWAP_WIDTH): Likewise.
305 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
306 (vget_lane_bf16, vgetq_lane_bf16): New.
308 (vdup_n_bf16, vdupq_n_bf16): New.
309 (vdup_lane_bf16, vdup_laneq_bf16): New.
310 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
311 (vduph_lane_bf16, vduph_laneq_bf16): New.
312 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
313 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
314 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
315 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
316 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
317 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
318 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
319 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
320 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
321 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
322 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
323 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
324 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
325 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
326 (vreinterpretq_bf16_p128): New.
327 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
328 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
329 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
330 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
331 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
332 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
333 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
334 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
335 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
336 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
337 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
338 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
339 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
340 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
341 (vreinterpretq_p128_bf16): New.
343 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
345 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
346 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
347 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
348 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
349 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
350 * config/arm/iterators.md (VSF2BF): New attribute.
351 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
352 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
353 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
355 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
357 * config/arm/arm.md (required_for_purecode): New attribute.
358 (enabled): Handle required_for_purecode.
359 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
360 work with -mpure-code.
362 2020-02-25 Jakub Jelinek <jakub@redhat.com>
364 PR rtl-optimization/93908
365 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
368 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
370 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
372 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
374 * doc/install.texi (--enable-checking): Adjust wording.
376 2020-02-25 Richard Biener <rguenther@suse.de>
378 PR tree-optimization/93868
379 * tree-vect-slp.c (slp_copy_subtree): New function.
380 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
381 re-arranging stmts in it.
383 2020-02-25 Jakub Jelinek <jakub@redhat.com>
386 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
387 dummy function and remove it at the end.
390 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
391 paramter -> parameter.
392 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
393 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
395 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
397 * doc/install.texi (--enable-checking): Properly document current
399 (--enable-stage1-checking): Minor clarification about bootstrap.
401 2020-02-24 David Malcolm <dmalcolm@redhat.com>
404 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
405 -fanalyzer-checker=taint is also required.
406 (-fanalyzer-checker=): Note that providing this option enables the
407 given checker, and doing so may be required for checkers that are
410 2020-02-24 David Malcolm <dmalcolm@redhat.com>
412 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
413 significant control flow events; add a "3" which shows all
414 control flow events; the old "3" becomes "4".
416 2020-02-24 Jakub Jelinek <jakub@redhat.com>
418 PR tree-optimization/93582
419 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
420 pd.offset and pd.size to be counted in bits rather than bytes, add
421 support for maxsizei that is not a multiple of BITS_PER_UNIT and
422 handle bitfield stores and loads.
423 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
424 uncomparable quantities - bytes vs. bits. Allow push_partial_def
425 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
426 pd.offset/pd.size to be counted in bits rather than bytes.
427 Formatting fix. Rename shadowed len variable to buflen.
429 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
430 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
433 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
434 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
435 * opts-common.c (parse_options_from_collect_gcc_options): New function.
436 (prepend_xassembler_to_collect_as_options): Likewise.
437 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
438 (prepend_xassembler_to_collect_as_options): Likewise.
439 * lto-opts.c (lto_write_options): Stream assembler options
440 in COLLECT_AS_OPTIONS.
441 * lto-wrapper.c (xassembler_options_error): New static variable.
442 (get_options_from_collect_gcc_options): Move parsing options code to
443 parse_options_from_collect_gcc_options and call it.
444 (merge_and_complain): Validate -Xassembler options.
445 (append_compiler_options): Handle OPT_Xassembler.
446 (run_gcc): Append command line -Xassembler options to
448 * doc/invoke.texi: Add documentation about using Xassembler
451 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
453 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
455 (riscv_rtx_costs): Update cost model for LTGT.
457 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
459 PR rtl-optimization/93564
460 * ira-color.c (struct update_cost_queue_elem): New member start.
461 (queue_update_cost, get_next_update_cost): Add new arg start.
462 (allocnos_conflict_p): New function.
463 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
464 Add checking conflicts with allocnos_conflict_p.
465 (update_costs_from_prefs, restore_costs_from_copies): Adjust
466 update_costs_from_allocno calls.
467 (update_conflict_hard_regno_costs): Add checking conflicts with
468 allocnos_conflict_p. Adjust calls of queue_update_cost and
469 get_next_update_cost.
470 (assign_hard_reg): Adjust calls of queue_update_cost. Add
472 (bucket_allocno_compare_func): Restore previous version.
474 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
476 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
477 double-word size when handling aggregate return values.
478 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
479 that homogeneous SFmode and DFmode aggregates are passed and returned
480 in general registers.
482 2020-02-21 Jakub Jelinek <jakub@redhat.com>
485 * opts.c (print_filtered_help): Translate help before appending
486 messages to it rather than after that.
488 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
490 PR rtl-optimization/PR92989
491 * lra-lives.c (process_bb_lives): Restore the original order
492 of the bb liveness update. Call make_hard_regno_dead for each
493 register clobbered at the start of an EH receiver.
495 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
498 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
499 self-recursively generated.
501 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
504 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
507 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
509 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
510 Document new target supports option.
512 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
514 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
515 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
516 * config/arm/iterators.md (MATMUL): New iterator.
517 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
518 (mmla_sfx): New attribute.
519 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
520 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
521 (UNSPEC_MATMUL_US): New.
523 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
525 * config/arm/arm.md: Prevent scalar shifts from being used when big
528 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
529 Richard Biener <rguenther@suse.de>
531 PR tree-optimization/93586
532 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
533 after mismatched array refs; do not sure type size information to
534 recover from unmatched referneces with !flag_strict_aliasing_p.
536 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
538 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
539 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
540 (scatter_store<mode>): Rename to ...
541 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
542 (scatter<mode>_exec): Delete. Move contents ...
543 (mask_scatter_store<mode>): ... here, and rename that to ...
544 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
545 Remove mode conversion.
546 (mask_gather_load<mode>): Rename to ...
547 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
548 Remove mode conversion.
549 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
551 2020-02-21 Martin Jambor <mjambor@suse.cz>
553 PR tree-optimization/93845
554 * tree-sra.c (verify_sra_access_forest): Only test access size of
557 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
559 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
560 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
561 (addv64di3_exec): Likewise.
562 (subv64di3): Likewise.
563 (subv64di3_exec): Likewise.
564 (addv64di3_zext): Likewise.
565 (addv64di3_zext_exec): Likewise.
566 (addv64di3_zext_dup): Likewise.
567 (addv64di3_zext_dup_exec): Likewise.
568 (addv64di3_zext_dup2): Likewise.
569 (addv64di3_zext_dup2_exec): Likewise.
570 (addv64di3_sext_dup2): Likewise.
571 (addv64di3_sext_dup2_exec): Likewise.
572 (<expander>v64di3): Likewise.
573 (<expander>v64di3_exec): Likewise.
574 (*<reduc_op>_dpp_shr_v64di): Likewise.
575 (*plus_carry_dpp_shr_v64di): Likewise.
576 * config/gcn/gcn.md (adddi3): Likewise.
577 (addptrdi3): Likewise.
578 (<expander>di3): Likewise.
580 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
582 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
584 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
586 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
587 support. Use aarch64_emit_mult instead of emitting multiplication
588 instructions directly.
589 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
590 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
592 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
594 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
595 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
596 instead of emitting multiplication instructions directly.
597 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
598 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
599 (@aarch64_frecps<mode>): New expanders.
601 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
603 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
604 on and produce uint64_ts rather than ints.
605 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
606 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
608 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
610 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
611 an unused xmsk register when handling approximate rsqrt.
613 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
615 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
616 flag_finite_math_only condition.
618 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
621 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
622 to destination operand for shufps alternative.
623 (*vec_extractv2si_1): Ditto.
625 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
628 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
631 2020-02-20 Martin Liska <mliska@suse.cz>
634 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
636 2020-02-20 Martin Liska <mliska@suse.cz>
639 * common/config/avr/avr-common.c: Remote trailing "|".
641 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
643 * collect2.c (maybe_run_lto_and_relink): Fix typo in
646 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
648 PR tree-optimization/93767
649 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
650 access-size bias from the offset calculations for negative strides.
652 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
654 * collect2.c (c_file, o_file): Make const again.
655 (ldout,lderrout, dump_ld_file): Remove.
656 (tool_cleanup): Avoid calling not signal-safe functions.
657 (maybe_run_lto_and_relink): Avoid possible signal handler
658 access to unintialzed memory (lto_o_files).
659 (main): Avoid leaking temp files in $TMPDIR.
660 Initialize c_file/o_file with concat, which avoids exposing
661 uninitialized memory to signal handler, which calls unlink(!).
662 Avoid calling maybe_unlink when the main function returns,
663 since the atexit handler is already doing this.
664 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
666 2020-02-19 Martin Jambor <mjambor@suse.cz>
668 PR tree-optimization/93776
669 * tree-sra.c (create_access): Do not create zero size accesses.
670 (get_access_for_expr): Do not search for zero sized accesses.
672 2020-02-19 Martin Jambor <mjambor@suse.cz>
674 PR tree-optimization/93667
675 * tree-sra.c (scalarizable_type_p): Return false if record fields
676 do not follow wach other.
678 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
680 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
681 rather than fmv.x.s/fmv.s.x.
683 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
685 * config/aarch64/aarch64-simd-builtins.def
686 (intrinsic_vec_smult_lo_): New.
687 (intrinsic_vec_umult_lo_): Likewise.
688 (vec_widen_smult_hi_): Likewise.
689 (vec_widen_umult_hi_): Likewise.
690 * config/aarch64/aarch64-simd.md
691 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
692 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
693 (vmull_high_s16): Likewise.
694 (vmull_high_s32): Likewise.
695 (vmull_high_u8): Likewise.
696 (vmull_high_u16): Likewise.
697 (vmull_high_u32): Likewise.
698 (vmull_s8): Likewise.
699 (vmull_s16): Likewise.
700 (vmull_s32): Likewise.
701 (vmull_u8): Likewise.
702 (vmull_u16): Likewise.
703 (vmull_u32): Likewise.
705 2020-02-18 Martin Liska <mliska@suse.cz>
707 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
708 bootstrap by missing removal of invalid sanity check.
710 2020-02-18 Martin Liska <mliska@suse.cz>
713 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
714 Always compare LHS of gimple_assign.
716 2020-02-18 Martin Liska <mliska@suse.cz>
719 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
720 and return type of functions.
721 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
722 Drop MALLOC attribute for void functions.
723 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
724 malloc_state for a new VOID clone.
726 2020-02-18 Martin Liska <mliska@suse.cz>
729 * common.opt: Add -fprofile-reproducibility.
730 * doc/invoke.texi: Document it.
731 * value-prof.c (dump_histogram_value):
732 Document and support behavior for counters[0]
733 being a negative value.
734 (get_nth_most_common_value): Handle negative
735 counters[0] in respect to flag_profile_reproducible.
737 2020-02-18 Jakub Jelinek <jakub@redhat.com>
740 * cgraph.c (verify_speculative_call): Use speculative_id instead of
741 speculative_uid in messages. Remove trailing whitespace from error
742 message. Use num_speculative_call_targets instead of
743 num_speculative_targets in a message.
744 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
745 edge messages and stmt instead of cal_stmt in reference message.
747 PR tree-optimization/93780
748 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
749 before calling build_vector_type.
750 (execute_update_addresses_taken): Likewise.
753 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
754 typo, functoin -> function.
755 * tree.c (free_lang_data_in_decl): Fix comment typo,
756 functoin -> function.
757 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
759 2020-02-17 David Malcolm <dmalcolm@redhat.com>
761 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
763 (print_option_information): Don't call get_option_url if URLs
766 2020-02-17 Alexandre Oliva <oliva@adacore.com>
768 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
769 handling of register_common-less targets.
771 2020-02-17 Martin Liska <mliska@suse.cz>
774 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
776 2020-02-17 Martin Liska <mliska@suse.cz>
779 * config/rs6000/rs6000.c (rs6000_option_override_internal):
782 2020-02-17 Martin Liska <mliska@suse.cz>
785 * config/rx/elf.opt: Fix typo.
787 2020-02-17 Richard Biener <rguenther@suse.de>
790 * opts-global.c (print_ignored_options): Use inform and
793 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
796 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
798 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
801 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
802 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
804 2020-02-15 Jason Merrill <jason@redhat.com>
806 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
808 2020-02-15 Jakub Jelinek <jakub@redhat.com>
810 PR tree-optimization/93744
811 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
812 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
813 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
814 sure @2 in the first and @1 in the other patterns has no side-effects.
816 2020-02-15 David Malcolm <dmalcolm@redhat.com>
817 Bernd Edlinger <bernd.edlinger@hotmail.de>
821 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
822 * configure.ac (--with-diagnostics-urls): New configuration
823 option, based on --with-diagnostics-color.
824 (DIAGNOSTICS_URLS_DEFAULT): New define.
825 * config.h: Regenerate.
826 * configure: Regenerate.
827 * diagnostic.c (diagnostic_urls_init): Handle -1 for
828 DIAGNOSTICS_URLS_DEFAULT from configure-time
829 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
830 and TERM_URLS environment variable.
831 * diagnostic-url.h (diagnostic_url_format): New enum type.
832 (diagnostic_urls_enabled_p): rename to...
833 (determine_url_format): ... this, and change return type.
834 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
835 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
836 the linux console, and mingw.
837 (diagnostic_urls_enabled_p): rename to...
838 (determine_url_format): ... this, and adjust.
839 * pretty-print.h (pretty_printer::show_urls): rename to...
840 (pretty_printer::url_format): ... this, and change to enum.
841 * pretty-print.c (pretty_printer::pretty_printer,
842 pp_begin_url, pp_end_url, test_urls): Adjust.
843 * doc/install.texi (--with-diagnostics-urls): Document the new
844 configuration option.
845 (--with-diagnostics-color): Document the existing interaction
846 with GCC_COLORS better.
847 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
848 vindex reference. Update description of defaults based on the above.
849 (-fdiagnostics-color): Update description of how -fdiagnostics-color
850 interacts with GCC_COLORS.
852 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
855 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
856 conjunction with TARGET_GNU_TLS in early return.
858 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
860 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
861 the mode is not wider than UNITS_PER_WORD.
863 2020-02-14 Martin Jambor <mjambor@suse.cz>
865 PR tree-optimization/93516
866 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
867 access of the same type as the parent.
868 (propagate_subaccesses_from_lhs): Likewise.
870 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
873 * config/i386/avx512vbmi2intrin.h
874 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
875 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
876 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
877 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
878 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
879 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
880 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
881 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
882 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
883 of lacking a closing parenthesis.
884 * config/i386/avx512vbmi2vlintrin.h
885 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
886 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
887 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
888 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
889 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
890 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
891 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
892 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
893 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
894 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
895 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
896 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
897 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
898 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
899 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
900 _mm_shldi_epi32, _mm_mask_shldi_epi32,
901 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
902 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
904 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
907 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
908 the target function entry.
910 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
912 * common/config/arc/arc-common.c (arc_option_optimization_table):
913 Disable if-conversion step when optimized for size.
915 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
917 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
918 R12-R15 are always in ARCOMPACT16_REGS register class.
919 * config/arc/arc.opt (mq-class): Deprecate.
920 * config/arc/constraint.md ("q"): Remove dependency on mq-class
922 * doc/invoke.texi (mq-class): Update text.
923 * common/config/arc/arc-common.c (arc_option_optimization_table):
926 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
928 * config/arc/arc.c (arc_insn_cost): New function.
929 (TARGET_INSN_COST): Define.
930 * config/arc/arc.md (cost): New attribute.
931 (add_n): Use arc_nonmemory_operand.
932 (ashlsi3_insn): Likewise, also update constraints.
933 (ashrsi3_insn): Likewise.
935 (add_shift): Likewise.
936 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
938 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
940 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
942 (umulsidi_600): Likewise.
944 2020-02-13 Jakub Jelinek <jakub@redhat.com>
947 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
948 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
949 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
950 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
951 pass __A to the builtin followed by __W instead of __A followed by
953 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
954 _mm512_mask_popcnt_epi64): Likewise.
955 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
956 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
957 _mm256_mask_popcnt_epi64): Likewise.
959 PR tree-optimization/93582
960 * fold-const.h (shift_bytes_in_array_left,
961 shift_bytes_in_array_right): Declare.
962 * fold-const.c (shift_bytes_in_array_left,
963 shift_bytes_in_array_right): New function, moved from
964 gimple-ssa-store-merging.c, no longer static.
965 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
966 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
967 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
968 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
969 shift_bytes_in_array.
970 (verify_shift_bytes_in_array): Rename to ...
971 (verify_shift_bytes_in_array_left): ... this. Use
972 shift_bytes_in_array_left instead of shift_bytes_in_array.
973 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
974 instead of verify_shift_bytes_in_array.
975 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
976 / native_interpret_expr where the store covers all needed bits,
977 punt on PDP-endian, otherwise allow all involved offsets and sizes
978 not to be byte-aligned.
981 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
982 use const_0_to_255_operand predicate instead of immediate_operand.
983 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
984 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
985 vgf2p8affineinvqb_<mode><mask_name>,
986 vgf2p8affineqb_<mode><mask_name>): Drop mode from
987 const_0_to_255_operand predicated operands.
989 2020-02-12 Jeff Law <law@redhat.com>
991 * config/h8300/h8300.md (comparison shortening peepholes): Use
992 a mode iterator to merge the HImode and SImode peepholes.
994 2020-02-12 Jakub Jelinek <jakub@redhat.com>
997 * real.c (is_even): Make static. Function comment fix.
998 (is_halfway_below): Make static, don't assert R is not inf/nan,
999 instead return false for those. Small formatting fixes.
1001 2020-02-12 Martin Sebor <msebor@redhat.com>
1004 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1005 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1006 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1007 (strlen_check_and_optimize_call): Adjust callee name.
1009 2020-02-12 Jeff Law <law@redhat.com>
1011 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1012 (and (xor)) variant. Combine other two into single peephole.
1014 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1016 PR rtl-optimization/93565
1017 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1019 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1021 * config/aarch64/aarch64-simd.md
1022 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1023 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1024 generating separate ADDV and zero_extend patterns.
1025 * config/aarch64/iterators.md (VDQV_E): New iterator.
1027 2020-02-12 Jeff Law <law@redhat.com>
1029 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1030 expanders, splits, etc.
1031 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1032 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1033 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1034 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1035 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1036 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1038 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1040 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1043 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1044 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1045 TARGET_AVX512DQ from condition.
1046 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1047 instead of <mask_mode512bit_condition> in condition. If
1048 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1050 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1053 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1056 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1058 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1060 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1061 where strlen is more legible.
1062 (rs6000_builtin_vectorized_libmass): Ditto.
1063 (rs6000_print_options_internal): Ditto.
1065 2020-02-11 Martin Sebor <msebor@redhat.com>
1067 PR tree-optimization/93683
1068 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1070 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1072 * config/rs6000/predicates.md (cint34_operand): Rename the
1073 -mprefixed-addr option to be -mprefixed.
1074 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1075 the -mprefixed-addr option to be -mprefixed.
1076 (OTHER_FUTURE_MASKS): Likewise.
1077 (POWERPC_MASKS): Likewise.
1078 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1079 the -mprefixed-addr option to be -mprefixed. Change error
1080 messages to refer to -mprefixed.
1081 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1083 (rs6000_legitimate_offset_address_p): Likewise.
1084 (rs6000_mode_dependent_address): Likewise.
1085 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1086 "-mprefixed" for target attributes and pragmas.
1087 (address_to_insn_form): Rename the -mprefixed-addr option to be
1089 (rs6000_adjust_insn_length): Likewise.
1090 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1091 -mprefixed-addr option to be -mprefixed.
1092 (ASM_OUTPUT_OPCODE): Likewise.
1093 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1094 -mprefixed-addr option to be -mprefixed.
1095 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1096 -mprefixed-addr option to be prefixed. Change the option from
1097 being undocumented to being documented.
1098 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1099 -mprefixed option. Update the -mpcrel documentation to mention
1102 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1104 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1105 including FIRST_PSEUDO_REGISTER - 1.
1106 * ira-color.c (print_hard_reg_set): Ditto.
1108 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1110 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1111 (USTERNOP_QUALIFIERS): New define.
1112 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1113 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1114 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1115 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1116 * config/arm/arm_neon.h (vusdot_s32): New.
1117 (vusdot_lane_s32): New.
1118 (vusdotq_lane_s32): New.
1119 (vsudot_lane_s32): New.
1120 (vsudotq_lane_s32): New.
1121 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1122 * config/arm/iterators.md (DOTPROD_I8MM): New.
1123 (sup, opsuffix): Add <us/su>.
1124 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1125 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1127 2020-02-11 Richard Biener <rguenther@suse.de>
1129 PR tree-optimization/93661
1130 PR tree-optimization/93662
1131 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1133 * tree-sra.c (get_access_for_expr): Likewise.
1135 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1138 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1139 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1140 Change condition from TARGET_AVX2 to TARGET_AVX.
1142 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1145 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1146 argument of strncmp.
1148 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1150 Try to generate zero-based comparisons.
1151 * config/cris/cris.c (cris_reduce_compare): New function.
1152 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1153 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1154 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1156 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1159 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1160 in Thumb state and also as a destination in Arm state. Add T16
1163 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1165 * md.texi (Define Subst): Match closing paren in example.
1167 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1171 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1172 arguments of strncmp.
1174 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1177 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1178 but different source value.
1179 (adjust_callers_for_value_intersection): New function.
1180 (gather_edges_for_value): Adjust order of callers to let a
1181 non-self-recursive caller be the first element.
1182 (self_recursive_pass_through_p): Add a new parameter "simple", and
1183 check generalized self-recursive pass-through jump function.
1184 (self_recursive_agg_pass_through_p): Likewise.
1185 (find_more_scalar_values_for_callers_subset): Compute value from
1186 pass-through jump function for self-recursive.
1187 (intersect_with_plats): Cleanup previous implementation code for value
1188 itersection with self-recursive call edge.
1189 (intersect_with_agg_replacements): Likewise.
1190 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1191 function for self-recursive call edge. Cleanup previous implementation
1192 code for value intersection with self-recursive call edge.
1193 (decide_whether_version_node): Remove dead callers and adjust order
1194 to let a non-self-recursive caller be the first element.
1196 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1198 * recog.c: Move pass_split_before_sched2 code in front of
1199 pass_split_before_regstack.
1200 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1201 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1202 (rest_of_handle_split_before_sched2): Remove.
1203 (pass_split_before_sched2::execute): Unconditionally call
1205 (enable_split_before_sched2): New function.
1206 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1207 (pass_split_before_regstack::gate): Ditto.
1208 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1209 Update name check for renamed split4 pass.
1210 * config/sh/sh.c (register_sh_passes): Update pass insertion
1211 point for renamed split4 pass.
1213 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1215 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1216 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1217 copying them around between host and target.
1219 2020-02-08 Andrew Pinski <apinski@marvell.com>
1222 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1223 STRICT_ALIGNMENT also.
1225 2020-02-08 Jim Wilson <jimw@sifive.com>
1228 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1230 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1231 Jakub Jelinek <jakub@redhat.com>
1234 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1235 xmm16-xmm31 call-used even in 64-bit ms-abi.
1237 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1239 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1240 (simd_ummla, simd_usmmla): Likewise.
1241 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1242 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1243 (vusmmlaq_s32): New.
1245 2020-02-07 Richard Biener <rguenther@suse.de>
1248 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1249 skipping unreachable regions.
1250 (optimize_inline_calls): Skip folding stmts when we didn't
1253 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1256 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1257 Don't return aggregates with only SFmode and DFmode in SSE
1259 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1261 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1264 * config/rs6000/rs6000-logue.c
1265 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1266 if it fails, move rs into end_addr and retry. Add
1267 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1268 the insn pattern doesn't describe well what exactly happens to
1272 * config/i386/predicates.md (avx_identity_operand): Remove.
1273 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1274 (avx_<castmode><avxsizesuffix>_<castmode>,
1275 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1276 a VEC_CONCAT of the operand and UNSPEC_CAST.
1277 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1278 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1282 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1283 recog_data.insn if distance_non_agu_define changed it.
1285 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1288 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1289 we only had X-FORM (reg+reg) addressing for vectors. Also before
1290 ISA 3.0, we only had X-FORM addressing for scalars in the
1291 traditional Altivec registers.
1293 2020-02-06 <zhongyunde@huawei.com>
1294 Vladimir Makarov <vmakarov@redhat.com>
1296 PR rtl-optimization/93561
1297 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1298 hard register range.
1300 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1302 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1305 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1307 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1308 where the low and the high 32 bits are equal to each other specially,
1309 with an rldimi instruction.
1311 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1313 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1315 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1317 * config/arm/arm-tables.opt: Regenerate.
1319 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1322 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1323 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1324 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1326 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1328 PR rtl-optimization/87763
1329 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1331 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1333 * config/aarch64/aarch64-simd-builtins.def
1334 (bfmlaq): New built-in function.
1335 (bfmlalb): New built-in function.
1336 (bfmlalt): New built-in function.
1337 (bfmlalb_lane): New built-in function.
1338 (bfmlalt_lane): New built-in function.
1339 * config/aarch64/aarch64-simd.md
1340 (aarch64_bfmmlaqv4sf): New pattern.
1341 (aarch64_bfmlal<bt>v4sf): New pattern.
1342 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1343 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1344 (vbfmlalbq_f32): New intrinsic.
1345 (vbfmlaltq_f32): New intrinsic.
1346 (vbfmlalbq_lane_f32): New intrinsic.
1347 (vbfmlaltq_lane_f32): New intrinsic.
1348 (vbfmlalbq_laneq_f32): New intrinsic.
1349 (vbfmlaltq_laneq_f32): New intrinsic.
1350 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1351 (bt): New int attribute.
1353 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1355 * config/i386/i386.md (*pushtf): Emit "#" instead of
1356 calling gcc_unreachable in insn output.
1359 (*pushsf_rex64): Ditto for alternatives other than 1.
1360 (*pushsf): Ditto for alternatives other than 1.
1362 2020-02-06 Martin Liska <mliska@suse.cz>
1364 PR gcov-profile/91971
1365 PR gcov-profile/93466
1366 * coverage.c (coverage_init): Revert mangling of
1367 path into filename. It can lead to huge filename length.
1368 Creation of subfolders seem more natural.
1370 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1373 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1374 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1375 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1377 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1380 * config/i386/predicates.md (avx_identity_operand): New predicate.
1381 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1382 define_insn_and_split.
1385 * omp-low.c (use_pointer_for_field): For nested constructs, also
1386 look for map clauses on target construct.
1387 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1388 taskreg_nesting_level.
1391 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1392 shared clause, call omp_notice_variable on outer context if any.
1394 2020-02-05 Jason Merrill <jason@redhat.com>
1397 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1398 non-zero address even if weak and not yet defined.
1400 2020-02-05 Martin Sebor <msebor@redhat.com>
1402 PR tree-optimization/92765
1403 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1404 * tree-ssa-strlen.c (compute_string_length): Remove.
1405 (determine_min_objsize): Remove.
1406 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1407 Avoid using type size as the upper bound on string length.
1408 (handle_builtin_string_cmp): Add an argument. Adjust.
1409 (strlen_check_and_optimize_call): Pass additional argument to
1410 handle_builtin_string_cmp.
1412 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1414 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1415 (*pushdi2_rex64 peephole2): Unconditionally split after
1417 (*ashl<mode>3_doubleword): Ditto.
1418 (*<shift_insn><mode>3_doubleword): Ditto.
1420 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1423 * config/rs6000/rs6000.c (get_vector_offset): Fix
1425 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1427 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1429 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1432 (Special Functions for Debugging the Analyzer): Update description
1433 of __analyzer_dump_exploded_nodes.
1435 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1438 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1439 include sets and not clobbers in the vzeroupper pattern.
1440 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1441 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1442 (*avx_vzeroupper_1): New define_insn_and_split.
1445 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1446 don't run when !optimize.
1447 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1450 2020-02-05 Richard Biener <rguenther@suse.de>
1453 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1454 checks before matching calls.
1456 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1458 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1459 function comment typo.
1462 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1463 simd_clone_create failed when i == 0, adjust clone->nargs by
1466 2020-02-05 Martin Liska <mliska@suse.cz>
1469 * doc/invoke.texi: Document that one should
1470 not combine ASLR and -fpch.
1472 2020-02-04 Richard Biener <rguenther@suse.de>
1474 PR tree-optimization/93538
1475 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1477 2020-02-04 Richard Biener <rguenther@suse.de>
1479 PR tree-optimization/91123
1480 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1481 (vn_walk_cb_data::last_vuse): New member.
1482 (vn_walk_cb_data::saved_operands): Likewsie.
1483 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1484 (vn_walk_cb_data::push_partial_def): Use finish.
1485 (vn_reference_lookup_2): Update last_vuse and use finish if
1486 we've saved operands.
1487 (vn_reference_lookup_3): Use finish and update calls to
1488 push_partial_defs everywhere. When translating through
1489 memcpy or aggregate copies save off operands and alias-set.
1490 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1491 operation for redundant store removal.
1493 2020-02-04 Richard Biener <rguenther@suse.de>
1495 PR tree-optimization/92819
1496 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1497 generating more stmts than before.
1499 2020-02-04 Martin Liska <mliska@suse.cz>
1501 * config/arm/arm.c (arm_gen_far_branch): Move the function
1502 outside of selftests.
1504 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1506 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1507 function to adjust PC-relative vector addresses.
1508 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1509 handle vectors with PC-relative addresses.
1511 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1513 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1515 (hard_reg_and_mode_to_addr_mask): Delete.
1516 (rs6000_adjust_vec_address): If the original vector address
1517 was REG+REG or REG+OFFSET and the element is not zero, do the add
1518 of the elements in the original address before adding the offset
1519 for the vector element. Use address_to_insn_form to validate the
1520 address using the register being loaded, rather than guessing
1521 whether the address is a DS-FORM or DQ-FORM address.
1523 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1525 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1526 to calculate the offset in memory from the start of a vector of a
1527 particular element. Add code to keep the element number in
1528 bounds if the element number is variable.
1529 (rs6000_adjust_vec_address): Move calculation of offset of the
1530 vector element to get_vector_offset.
1531 (rs6000_split_vec_extract_var): Do not do the initial AND of
1532 element here, move the code to get_vector_offset.
1534 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1536 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1539 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1541 * config/rs6000/constraints.md: Improve documentation.
1543 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1546 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1547 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1549 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1551 * config.gcc: Remove "carrizo" support.
1552 * config/gcn/gcn-opts.h (processor_type): Likewise.
1553 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1554 * config/gcn/gcn.opt (gpu_type): Likewise.
1555 * config/gcn/t-omp-device: Likewise.
1557 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1560 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1561 * config/arm/arm.c (arm_gen_far_branch): New function
1563 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1565 2020-02-03 Julian Brown <julian@codesourcery.com>
1566 Tobias Burnus <tobias@codesourcery.com>
1568 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1570 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1573 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1574 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1577 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1579 PR rtl-optimization/91333
1580 * ira-color.c (struct allocno_color_data): Add member
1582 (init_allocno_threads): Set the member up.
1583 (bucket_allocno_compare_func): Add compare hard reg
1586 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1588 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1590 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1591 * config.in: Regenerated.
1592 * configure: Regenerated.
1593 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1594 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1595 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1597 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1599 * configure: Regenerate.
1601 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1603 PR rtl-optimization/91333
1604 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1605 reg preferences comparison up.
1607 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1609 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1610 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1611 aarch64-sve-builtins-base.h.
1612 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1613 aarch64-sve-builtins-base.cc.
1614 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1615 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1617 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1618 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1619 (svcvtnt): New functions.
1620 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1621 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1622 (svcvtnt): New functions.
1623 (svcvt): Add a form that converts f32 to bf16.
1624 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1625 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1627 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1628 Treat B as bfloat16_t.
1629 (ternary_bfloat_lane_base): New class.
1630 (ternary_bfloat_def): Likewise.
1631 (ternary_bfloat): New shape.
1632 (ternary_bfloat_lane_def): New class.
1633 (ternary_bfloat_lane): New shape.
1634 (ternary_bfloat_lanex2_def): New class.
1635 (ternary_bfloat_lanex2): New shape.
1636 (ternary_bfloat_opt_n_def): New class.
1637 (ternary_bfloat_opt_n): New shape.
1638 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1639 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1640 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1641 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1642 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1643 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1644 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1645 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1646 the pattern off the narrow mode instead of the wider one.
1647 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1648 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1649 (sve_fp_op): Handle them.
1650 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1651 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1653 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1655 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1656 * config/aarch64/aarch64-modes.def (BF): Move definition before
1657 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1658 (SVE_MODES): Handle BF modes.
1659 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1661 (aarch64_full_sve_mode): Likewise.
1662 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1664 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1665 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1666 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1667 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1669 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1671 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1672 (TYPES_all_data): Add bf16.
1673 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1674 (register_tuple_type): Increase buffer size.
1675 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1676 (bf16): New type suffix.
1677 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1678 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1679 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1680 Change type from all_data to all_arith.
1681 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1684 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1685 Matthew Malcomson <matthew.malcomson@arm.com>
1686 Richard Sandiford <richard.sandiford@arm.com>
1688 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1689 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1690 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1691 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1692 __ARM_FEATURE_MATMUL_FP64.
1693 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1694 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1695 be disabled at the same time.
1696 (f32mm): New extension.
1697 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1698 (AARCH64_FL_F64MM): Bump to the next bit up.
1699 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1700 (TARGET_SVE_F64MM): New macros.
1701 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1702 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1703 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1704 (UNSPEC_ZIP2Q): New unspeccs.
1705 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1706 (optab, sur, perm_insn): Handle the new unspecs.
1707 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1708 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1709 TARGET_SVE_F64MM instead of separate tests.
1710 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1711 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1712 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1713 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1714 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1715 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1716 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1717 (TYPES_s_signed): New macro.
1718 (TYPES_s_integer): Use it.
1719 (TYPES_d_float): New macro.
1720 (TYPES_d_data): Use it.
1721 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1722 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1723 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1724 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1725 (svmmla): New shape.
1726 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1727 template parameters.
1728 (ternary_resize2_lane_base): Likewise.
1729 (ternary_resize2_base): New class.
1730 (ternary_qq_lane_base): Likewise.
1731 (ternary_intq_uintq_lane_def): Likewise.
1732 (ternary_intq_uintq_lane): New shape.
1733 (ternary_intq_uintq_opt_n_def): New class
1734 (ternary_intq_uintq_opt_n): New shape.
1735 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1736 (ternary_uintq_intq_def): New class.
1737 (ternary_uintq_intq): New shape.
1738 (ternary_uintq_intq_lane_def): New class.
1739 (ternary_uintq_intq_lane): New shape.
1740 (ternary_uintq_intq_opt_n_def): New class.
1741 (ternary_uintq_intq_opt_n): New shape.
1742 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1743 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1744 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1745 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1747 (svdotprod_lane_impl): ...this new class.
1748 (svmmla_impl, svusdot_impl): New classes.
1749 (svdot_lane): Update to use svdotprod_lane_impl.
1750 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1751 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1753 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1754 function, with no types defined.
1755 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1756 AARCH64_FL_I8MM functions.
1757 (svmmla): New AARCH64_FL_F32MM function.
1758 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1759 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1760 AARCH64_FL_F64MM function.
1761 (REQUIRED_EXTENSIONS):
1763 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1765 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1768 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1770 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1771 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1772 (*movti_internal): Do not check for
1773 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1774 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1775 just after check for TARGET_AVX.
1776 (*movdf_internal): Ditto.
1777 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1778 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1779 * config/i386/sse.md (mov<mode>_internal): Only check
1780 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1781 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1782 (<sse>_andnot<mode>3<mask_name>): Move check for
1783 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1784 (<code><mode>3<mask_name>): Ditto.
1785 (*andnot<mode>3): Ditto.
1786 (*andnottf3): Ditto.
1787 (*<code><mode>3): Ditto.
1788 (*<code>tf3): Ditto.
1789 (*andnot<VI:mode>3): Remove
1790 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1791 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1792 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1793 (sse4_1_blendv<ssemodesuffix>): Ditto.
1794 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1795 Explain that tune applies to 128bit instructions only.
1797 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1799 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1800 to definition of hsa_kernel_description. Parse assembly to find SGPR
1801 and VGPR count of kernel and store in hsa_kernel_description.
1803 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1805 PR rtl-optimization/91838
1806 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1807 to truncate if allowed or reject combination.
1809 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1811 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1812 (find_inv_vars_cb): Likewise.
1814 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1816 * calls.c (special_function_p): Split out the check for DECL_NAME
1817 being non-NULL and fndecl being extern at file scope into a
1818 new maybe_special_function_p and call it. Drop check for fndecl
1819 being non-NULL that was after a usage of DECL_NAME (fndecl).
1820 * tree.h (maybe_special_function_p): New inline function.
1822 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1824 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1825 (mask_gather_load<mode>): ... here, and zero-initialize the
1827 (maskload<mode>di): Zero-initialize the destination.
1830 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1833 * doc/analyzer.texi (Limitations): Note that constraints on
1834 floating-point values are currently ignored.
1836 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1839 * symtab.c (symtab_node::noninterposable_alias): If localalias
1840 already exists, but is not usable, append numbers after it until
1841 a unique name is found. Formatting fix.
1844 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1847 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1849 * config/gcn/gcn.c (print_operand): Handle LTGT.
1850 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1852 2020-01-30 Richard Biener <rguenther@suse.de>
1854 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1855 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1857 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1859 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1860 without a DECL in .data.rel.ro.local.
1862 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1865 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1869 * config/i386/sse.md
1870 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1871 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1872 any_extend code iterator instead of always zero_extend.
1873 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1874 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1875 Use any_extend code iterator instead of always zero_extend.
1876 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1877 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1878 Use any_extend code iterator instead of always zero_extend.
1879 (*sse2_pmovmskb_ext): New define_insn.
1880 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1883 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1884 (*popcountsi2_zext_falsedep): New define_insn.
1886 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1888 * config.in: Regenerated.
1889 * configure: Regenerated.
1891 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1894 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1895 LLVM's assembler changed the default in version 9.
1897 2020-01-24 Jeff Law <law@redhat.com>
1899 PR tree-optimization/89689
1900 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1902 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1906 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1908 PR rtl-optimization/87763
1909 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1910 simplification to handle subregs as well as bare regs.
1911 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1913 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1916 * ira.c (ira): Revert use of simplified LRA algorithm.
1918 2020-01-29 Martin Jambor <mjambor@suse.cz>
1920 PR tree-optimization/92706
1921 * tree-sra.c (struct access): Fields first_link, last_link,
1922 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1923 next_rhs_queued and grp_rhs_queued respectively, new fields
1924 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1925 (struct assign_link): Field next renamed to next_rhs, new field
1926 next_lhs. Updated comment.
1927 (work_queue_head): Renamed to rhs_work_queue_head.
1928 (lhs_work_queue_head): New variable.
1929 (add_link_to_lhs): New function.
1930 (relink_to_new_repr): Also relink LHS lists.
1931 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1932 (add_access_to_lhs_work_queue): New function.
1933 (pop_access_from_work_queue): Renamed to
1934 pop_access_from_rhs_work_queue.
1935 (pop_access_from_lhs_work_queue): New function.
1936 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1938 (child_would_conflict_in_lacc): Renamed to
1939 child_would_conflict_in_acc. Adjusted parameter names.
1940 (create_artificial_child_access): New parameter set_grp_read, use it.
1941 (subtree_mark_written_and_enqueue): Renamed to
1942 subtree_mark_written_and_rhs_enqueue.
1943 (propagate_subaccesses_across_link): Renamed to
1944 propagate_subaccesses_from_rhs.
1945 (propagate_subaccesses_from_lhs): New function.
1946 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1949 2020-01-29 Martin Jambor <mjambor@suse.cz>
1951 PR tree-optimization/92706
1952 * tree-sra.c (struct access): Adjust comment of
1953 grp_total_scalarization.
1954 (find_access_in_subtree): Look for single children spanning an entire
1956 (scalarizable_type_p): Allow register accesses, adjust callers.
1957 (completely_scalarize): Remove function.
1958 (scalarize_elem): Likewise.
1959 (create_total_scalarization_access): Likewise.
1960 (sort_and_splice_var_accesses): Do not track total scalarization
1962 (analyze_access_subtree): New parameter totally, adjust to new meaning
1963 of grp_total_scalarization.
1964 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1965 (can_totally_scalarize_forest_p): New function.
1966 (create_total_scalarization_access): Likewise.
1967 (create_total_access_and_reshape): Likewise.
1968 (total_should_skip_creating_access): Likewise.
1969 (totally_scalarize_subtree): Likewise.
1970 (analyze_all_variable_accesses): Perform total scalarization after
1971 subaccess propagation using the new functions above.
1972 (initialize_constant_pool_replacements): Output initializers by
1973 traversing the access tree.
1975 2020-01-29 Martin Jambor <mjambor@suse.cz>
1977 * tree-sra.c (verify_sra_access_forest): New function.
1978 (verify_all_sra_access_forests): Likewise.
1979 (create_artificial_child_access): Set parent.
1980 (analyze_all_variable_accesses): Call the verifier.
1982 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1984 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1985 if called on indirect edge.
1986 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1987 speculative call if needed.
1989 2020-01-29 Richard Biener <rguenther@suse.de>
1991 PR tree-optimization/93428
1992 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1993 permutation when the load node is created.
1994 (vect_analyze_slp_instance): Re-use it here.
1996 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1998 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2000 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2002 PR rtl-optimization/93272
2003 * ira-lives.c (process_out_of_region_eh_regs): New function.
2004 (process_bb_node_lives): Call it.
2006 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2008 * coverage.c (read_counts_file): Make error message lowercase.
2010 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2012 * profile-count.c (profile_quality_display_names): Fix ordering.
2014 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2017 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2018 hash only when edge is first within the sequence.
2019 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2020 (symbol_table::create_edge): Do not set target_prob.
2021 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2023 (cgraph_edge::make_speculative): Drop target_prob parameter.
2024 (cgraph_edge::speculative_call_info): Remove.
2025 (cgraph_edge::first_speculative_call_target): New member function.
2026 (update_call_stmt_hash_for_removing_direct_edge): New function.
2027 (cgraph_edge::resolve_speculation): Rewrite to new API.
2028 (cgraph_edge::speculative_call_for_target): New member function.
2029 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2030 multiple speculation targets.
2031 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2033 (verify_speculative_call): Verify that targets form an interval.
2034 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2035 (cgraph_edge::first_speculative_call_target): New member function.
2036 (cgraph_edge::next_speculative_call_target): New member function.
2037 (cgraph_edge::speculative_call_target_ref): New member function.
2038 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2039 (cgraph_edge): Remove target_prob.
2040 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2041 Fix handling of speculative calls.
2042 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2043 * ipa-fnsummary.c (analyze_function_body): Likewise.
2044 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2045 * ipa-profile.c (dump_histogram): Fix formating.
2046 (ipa_profile_generate_summary): Watch for overflows.
2047 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2048 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2049 (update_indirect_edges_after_inlining): Update to new API.
2050 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2052 * profile-count.h: (profile_probability::adjusted): New.
2053 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2054 updating of profile.
2055 * value-prof.c (gimple_ic_transform): Rename to ...
2056 (dump_ic_profile): ... this one; update dumping.
2057 (stream_in_histogram_value): Fix formating.
2058 (gimple_value_profile_transformations): Update.
2060 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2063 * config/i386/i386.md (*movoi_internal_avx): Remove
2064 TARGET_SSE_TYPELESS_STORES check.
2065 (*movti_internal): Prefer TARGET_AVX over
2066 TARGET_SSE_TYPELESS_STORES.
2067 (*movtf_internal): Likewise.
2068 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2069 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2070 from TARGET_SSE_TYPELESS_STORES.
2072 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2074 * diagnostic-core.h (warning_at): Rename overload to...
2075 (warning_meta): ...this.
2076 (emit_diagnostic_valist): Delete decl of overload taking
2077 diagnostic_metadata.
2078 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2079 (warning_at): Rename overload taking diagnostic_metadata to...
2080 (warning_meta): ...this.
2082 2020-01-28 Richard Biener <rguenther@suse.de>
2084 PR tree-optimization/93439
2085 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2086 * tree-cfg.c (move_sese_region_to_fn): ... here.
2087 (verify_types_in_gimple_reference): Verify used cliques are
2090 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2093 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2094 argument of a pointer to struct gcc_options and pass it to
2095 parse_mtune_ctrl_str.
2096 (ix86_function_specific_restore): Pass opts to
2097 set_ix86_tune_features.
2098 (ix86_option_override_internal): Likewise.
2099 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2100 gcc_options and use it for x_ix86_tune_ctrl_string.
2102 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2104 PR rtl-optimization/87763
2105 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2106 simplification to handle subregs as well as bare regs.
2107 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2109 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2111 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2112 for reduction chains that (now) include a call.
2114 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2116 PR tree-optimization/92822
2117 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2118 out the don't-care elements of a vector whose significant elements
2119 are duplicates, make the don't-care elements duplicates too.
2121 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2123 PR tree-optimization/93434
2124 * tree-predcom.c (split_data_refs_to_components): Record which
2125 components have had aliasing loads removed. Prevent store-store
2126 commoning for all such components.
2128 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2131 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2132 -1 or is_vshift is true, use new_vector with number of elts npatterns
2133 rather than new_unary_operation.
2135 PR tree-optimization/93454
2136 * gimple-fold.c (fold_array_ctor_reference): Perform
2137 elt_size.to_uhwi () just once, instead of calling it in every
2138 iteration. Punt if that value is above size of the temporary
2139 buffer. Decrease third native_encode_expr argument when
2140 bufoff + elt_sz is above size of buf.
2142 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2144 * config/mips/mips.c (mips_declare_object_name)
2145 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2147 2020-01-27 Martin Liska <mliska@suse.cz>
2149 PR gcov-profile/93403
2150 * tree-profile.c (gimple_init_gcov_profiler): Generate
2151 both __gcov_indirect_call_profiler_v4 and
2152 __gcov_indirect_call_profiler_v4_atomic.
2154 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2157 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2159 (@aarch64_split_simd_mov<mode>): Use it.
2160 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2161 Leave the vec_extract patterns to handle 2-element vectors.
2162 (aarch64_simd_mov_from_<mode>high): Likewise.
2163 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2164 (vec_extractv2dfv1df): Likewise.
2166 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2168 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2169 jump conditions for *compare_condjump<GPI:mode>.
2171 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2174 * digraph.cc (test_edge::test_edge): Specify template for base
2177 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2179 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2181 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2183 * config/arc/arc-protos.h (gen_mlo): Remove.
2184 (gen_mhi): Likewise.
2185 * config/arc/arc.c (AUX_MULHI): Define.
2186 (arc_must_save_reister): Special handling for r58/59.
2187 (arc_compute_frame_size): Consider mlo/mhi registers.
2188 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2190 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2191 mlo/mhi name selection.
2192 (arc_restore_callee_saves): Don't early restore blink when ISR.
2193 (arc_expand_prologue): Add mlo/mhi saving.
2194 (arc_expand_epilogue): Add mlo/mhi restoring.
2197 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2198 numbering when MUL64 option is used.
2199 (DWARF2_FRAME_REG_OUT): Define.
2200 * config/arc/arc.md (arc600_stall): New pattern.
2201 (VUNSPEC_ARC_ARC600_STALL): Define.
2202 (mulsi64): Use correct mlo/mhi registers.
2203 (mulsi_600): Clean it up.
2204 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2206 (mhi_operand): Likewise.
2208 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2209 Petro Karashchenko <petro.karashchenko@ring.com>
2211 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2212 attributes if needed.
2213 (prepare_move_operands): Generate special unspec instruction for
2215 (arc_isuncached_mem_p): Propagate uncached attribute to each
2217 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2218 (VUNSPEC_ARC_STDI): Likewise.
2219 (ALLI): New mode iterator.
2220 (mALLI): New mode attribute.
2221 (lddi): New instruction pattern.
2223 (stdidi_split): Split instruction for architectures which are not
2224 supporting ll64 option.
2225 (lddidi_split): Likewise.
2227 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2229 PR rtl-optimization/92989
2230 * lra-lives.c (process_bb_lives): Update the live-in set before
2231 processing additional clobbers.
2233 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2235 PR rtl-optimization/93170
2236 * cselib.c (cselib_invalidate_regno_val): New function, split out
2238 (cselib_invalidate_regno): ...here.
2239 (cselib_invalidated_by_call_p): New function.
2240 (cselib_process_insn): Iterate over all the hard-register entries in
2241 REG_VALUES and invalidate any that cross call-clobbered registers.
2243 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2245 * dojump.c (split_comparison): Use HONOR_NANS rather than
2246 HONOR_SNANS when splitting LTGT.
2248 2020-01-27 Martin Liska <mliska@suse.cz>
2251 * opts.c (print_filtered_help): Exclude language-specific
2252 options from --help=common unless enabled in all FEs.
2254 2020-01-27 Martin Liska <mliska@suse.cz>
2256 * opts.c (print_help): Exclude params from
2257 all except --help=param.
2259 2020-01-27 Martin Liska <mliska@suse.cz>
2262 * config/i386/i386-features.c (make_resolver_func):
2263 Align the code with ppc64 target implementation.
2264 Do not generate a unique name for resolver function.
2266 2020-01-27 Richard Biener <rguenther@suse.de>
2268 PR tree-optimization/93397
2269 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2270 converted reduction chain SLP graph adjustment.
2272 2020-01-26 Marek Polacek <polacek@redhat.com>
2275 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2278 2020-01-26 Jason Merrill <jason@redhat.com>
2281 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2284 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2286 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2287 (rx_setmem): Likewise.
2289 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2292 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2293 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2294 drop <di> from constraint of last operand.
2297 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2298 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2299 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2301 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2304 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2307 2020-01-24 Jeff Law <law@redhat.com>
2309 PR tree-optimization/92788
2310 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2313 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2316 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2317 *avx_vperm_broadcast_<mode>,
2318 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2319 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2320 Move before avx2_perm<mode>/avx512f_perm<mode>.
2323 * simplify-rtx.c (simplify_const_unary_operation,
2324 simplify_const_binary_operation): Punt for mode precision above
2325 MAX_BITSIZE_MODE_ANY_INT.
2327 2020-01-24 Andrew Pinski <apinski@marvell.com>
2329 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2332 2020-01-24 Jeff Law <law@redhat.com>
2335 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2336 for REGs. Call output_operand_lossage to get more reasonable
2339 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2341 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2342 gcn_fp_compare_operator.
2343 (vec_cmpu<mode>di): Use gcn_compare_operator.
2344 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2345 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2346 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2347 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2348 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2349 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2350 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2351 gcn_fp_compare_operator.
2352 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2353 gcn_fp_compare_operator.
2354 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2355 gcn_fp_compare_operator.
2356 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2357 gcn_fp_compare_operator.
2359 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2361 * doc/install.texi (Cross-Compiler-Specific Options): Document
2362 `--with-toolexeclibdir' option.
2364 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2366 * target.def (flags_regnum): Also mention effect on delay slot filling.
2367 * doc/tm.texi: Regenerate.
2369 2020-01-23 Jeff Law <law@redhat.com>
2371 PR translation/90162
2372 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2374 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2377 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2380 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2382 PR rtl-optimization/93402
2383 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2386 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2388 * config.in: Regenerated.
2389 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2390 for TARGET_LIBC_GNUSTACK.
2391 * configure: Regenerated.
2392 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2393 found to be 2.31 or greater.
2395 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2397 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2399 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2400 (mips_asm_file_end): New function. Delegate to
2401 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2402 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2404 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2407 * config/i386/i386-modes.def (POImode): New mode.
2408 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2409 * config/i386/i386.md (DPWI): New mode attribute.
2410 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2412 (QPWI): ... this. Use POI instead of OI for TImode.
2413 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2414 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2417 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2420 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2422 (speculation_tracker_rev): New pattern.
2423 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2424 Use speculation_tracker_rev to track the inverse condition.
2426 2020-01-23 Richard Biener <rguenther@suse.de>
2428 PR tree-optimization/93381
2429 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2430 alias-set of the def as argument and record the first one.
2431 (vn_walk_cb_data::first_set): New member.
2432 (vn_reference_lookup_3): Pass the alias-set of the current def
2433 to push_partial_def. Fix alias-set used in the aggregate copy
2435 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2436 * real.c (clear_significand_below): Fix out-of-bound access.
2438 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2441 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2442 New define_insn patterns.
2444 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2446 * doc/sourcebuild.texi (check-function-bodies): Add an
2447 optional target/xfail selector.
2449 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2451 PR rtl-optimization/93124
2452 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2453 bare USE and CLOBBER insns.
2455 2020-01-22 Andrew Pinski <apinski@marvell.com>
2457 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2459 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2462 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2463 diagnostic_manager into "ana" namespace.
2464 * selftest-run-tests.c (selftest::run_tests): Update for move of
2465 selftest::run_analyzer_selftests to
2466 ana::selftest::run_analyzer_selftests.
2468 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2470 * cfgexpand.c (union_stack_vars): Update the size.
2472 2020-01-22 Richard Biener <rguenther@suse.de>
2474 PR tree-optimization/93381
2475 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2476 throughout, handle all conversions the same.
2478 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2481 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2482 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2483 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2484 Call force_reg on high_in2 unconditionally.
2486 2020-01-22 Martin Liska <mliska@suse.cz>
2488 PR tree-optimization/92924
2489 * profile.c (compute_value_histograms): Divide
2492 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2495 * output.h (assemble_name_resolve): Declare.
2496 * varasm.c (assemble_name_resolve): New function.
2497 (assemble_name): Use it.
2498 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2500 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2502 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2503 update_web_docs_git instead of update_web_docs_svn.
2505 2020-01-21 Andrew Pinski <apinski@marvell.com>
2508 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2509 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2510 (*tlsgd_small_<mode>): Likewise.
2511 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2512 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2513 register. Convert that register back to dest using convert_mode.
2515 2020-01-21 Jim Wilson <jimw@sifive.com>
2517 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2520 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2521 Uros Bizjak <ubizjak@gmail.com>
2524 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2526 (legitimize_tls_address): Do GNU2 TLS address computation in
2527 ptr_mode and zero-extend result to Pmode.
2528 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2529 :P with :PTR and Pmode with ptr_mode.
2530 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2531 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2532 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2534 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2537 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2538 the last two operands are CONST_INT_P before using them as such.
2540 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2542 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2543 to get the integer element types.
2545 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2547 * config/aarch64/aarch64-sve-builtins.h
2548 (function_expander::convert_to_pmode): Declare.
2549 * config/aarch64/aarch64-sve-builtins.cc
2550 (function_expander::convert_to_pmode): New function.
2551 (function_expander::get_contiguous_base): Use it.
2552 (function_expander::prepare_gather_address_operands): Likewise.
2553 * config/aarch64/aarch64-sve-builtins-sve2.cc
2554 (svwhilerw_svwhilewr_impl::expand): Likewise.
2556 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2559 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2560 cfun->machine->label_is_assembled.
2561 (aarch64_print_patchable_function_entry): New.
2562 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2563 * config/aarch64/aarch64.h (struct machine_function): New field,
2566 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2569 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2572 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2575 * cgraph.c (cgraph_edge::resolve_speculation,
2576 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2577 call_stmt_site_hash.
2579 2020-01-21 Martin Liska <mliska@suse.cz>
2581 * config/rs6000/rs6000.c (common_mode_defined): Remove
2584 2020-01-21 Richard Biener <rguenther@suse.de>
2586 PR tree-optimization/92328
2587 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2588 type when value-numbering same-sized store by inserting a
2590 (eliminate_dom_walker::eliminate_stmt): When eliminating
2591 a redundant store handle bit-reinterpretation of the same value.
2593 2020-01-21 Andrew Pinski <apinski@marvel.com>
2596 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2598 (prepare_block_for_update): This. Use a worklist instead of
2601 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2603 * gcc/config/arm/arm.c (clear_operation_p):
2604 Initialise last_regno, skip first iteration
2605 based on the first_set value and use ints instead
2606 of the unnecessary HOST_WIDE_INTs.
2608 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2611 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2612 compare_mode other than SFmode or DFmode.
2614 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2617 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2618 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2619 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2621 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2623 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2625 2020-01-20 Andrew Pinski <apinski@marvell.com>
2628 * targhooks.c (default_print_patchable_function_entry): Use
2629 output_asm_insn to emit the nop instruction.
2631 2020-01-20 Fangrui Song <maskray@google.com>
2634 * targhooks.c (default_print_patchable_function_entry): Align to
2637 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2640 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2641 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2642 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2643 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2644 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2645 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2646 Remove the {q} suffix from lea.
2647 (*tls_dynamic_gnu2_call_64): Renamed to ...
2648 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2649 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2650 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2651 Pass Pmode to gen_tls_dynamic_gnu2_64.
2653 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2655 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2657 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2659 * config/aarch64/aarch64-sve-builtins-base.cc
2660 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2662 2020-01-20 Richard Biener <rguenther@suse.de>
2665 * dwarf2out.c (prune_unused_types): Unconditionally mark
2666 called function DIEs.
2668 2020-01-20 Martin Liska <mliska@suse.cz>
2670 PR tree-optimization/93199
2671 * tree-eh.c (struct leh_state): Add
2672 new field outer_non_cleanup.
2673 (cleanup_is_dead_in): Pass leh_state instead
2674 of eh_region. Add a checking that state->outer_non_cleanup
2675 points to outer non-clean up region.
2676 (lower_try_finally): Record outer_non_cleanup
2678 (lower_catch): Likewise.
2679 (lower_eh_filter): Likewise.
2680 (lower_eh_must_not_throw): Likewise.
2681 (lower_cleanup): Likewise.
2683 2020-01-20 Richard Biener <rguenther@suse.de>
2685 PR tree-optimization/93094
2686 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2687 (vect_transform_loop): Likewise.
2688 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2689 loop_vectorized_call to vect_transform_loop.
2690 * tree-vect-loop.c (vect_transform_loop): Pass down
2691 loop_vectorized_call to vect_loop_versioning.
2692 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2693 the earlier discovered loop_vectorized_call.
2695 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2697 * doc/contribute.texi: Update for SVN -> Git transition.
2698 * doc/install.texi: Likewise.
2700 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2703 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2704 speculative targets.
2705 (verify_speculative_call): New function
2706 (cgraph_node::verify_node): Use it.
2707 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2710 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2713 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2714 (cgraph_edge::make_direct): Remove all indirect targets.
2715 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2716 (cgraph_node::verify_node): Verify that only one call_stmt or
2717 lto_stmt_uid is set.
2718 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2720 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2721 (lto_output_ref): Simplify streaming of stmt.
2722 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2724 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2726 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2727 Mark parameter unused.
2729 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2731 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2733 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2735 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2737 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2739 * Makefile.in: Add coroutine-passes.o.
2740 * builtin-types.def (BT_CONST_SIZE): New.
2741 (BT_FN_BOOL_PTR): New.
2742 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2743 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2744 * coroutine-builtins.def: New file.
2745 * coroutine-passes.cc: New file.
2746 * function.h (struct GTY function): Add a bit to indicate that the
2747 function is a coroutine component.
2748 * internal-fn.c (expand_CO_FRAME): New.
2749 (expand_CO_YIELD): New.
2750 (expand_CO_SUSPN): New.
2751 (expand_CO_ACTOR): New.
2752 * internal-fn.def (CO_ACTOR): New.
2756 * passes.def: Add pass_coroutine_lower_builtins,
2757 pass_coroutine_early_expand_ifns.
2758 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2759 (make_pass_coroutine_early_expand_ifns): New.
2760 * doc/invoke.texi: Document the fcoroutines command line
2763 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2765 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2768 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2769 after checking the argument is a REG. Don't use REGNO (reg)
2770 again to set last_regno, reuse regno variable instead.
2772 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2774 * doc/analyzer.texi (Limitations): Add note about NaN.
2776 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2777 Sudakshina Das <sudi.das@arm.com>
2779 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2780 and valid immediate.
2781 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2782 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2783 * config/arm/constraints.md (Pg): New.
2784 * config/arm/predicates.md (long_shift_imm): New.
2785 (arm_reg_or_long_shift_imm): Likewise.
2786 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2787 (thumb2_lsll): Likewise.
2790 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2791 Sudakshina Das <sudi.das@arm.com>
2793 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2794 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2795 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2796 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2797 * config/arm/thumb2.md (thumb2_asrl): New.
2798 (thumb2_lsll): Likewise.
2800 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2802 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2805 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2807 * gdbinit.in (help-gcc-hooks): New command.
2808 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2809 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2812 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2814 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2815 correct target macro.
2817 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2819 * config/aarch64/aarch64-protos.h
2820 (aarch64_sve_ld1ro_operand_p): New.
2821 * config/aarch64/aarch64-sve-builtins-base.cc
2822 (class load_replicate): New.
2823 (class svld1ro_impl): New.
2824 (class svld1rq_impl): Change to inherit from load_replicate.
2825 (svld1ro): New sve intrinsic function base.
2826 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2827 New DEF_SVE_FUNCTION.
2828 * config/aarch64/aarch64-sve-builtins-base.h
2829 (svld1ro): New decl.
2830 * config/aarch64/aarch64-sve-builtins.cc
2831 (function_expander::add_mem_operand): Modify assert to allow
2833 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2835 * config/aarch64/aarch64.c
2836 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2837 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2838 (aarch64_sve_ld1ro_operand_p): New.
2839 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2840 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2841 * config/aarch64/predicates.md
2842 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2844 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2846 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2847 Introduce this ACLE specified predefined macro.
2848 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2849 (fp): Disabling this disables f64mm.
2850 (simd): Disabling this disables f64mm.
2851 (fp16): Disabling this disables f64mm.
2852 (sve): Disabling this disables f64mm.
2853 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2854 (AARCH64_ISA_F64MM): New.
2855 (TARGET_F64MM): New.
2856 * doc/invoke.texi (f64mm): Document new option.
2858 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2860 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2861 (neoversen1_tunings): Likewise.
2863 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2866 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2867 Add assert to ensure prolog has been emitted.
2868 (aarch64_split_atomic_op): Likewise.
2869 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2870 Use epilogue_completed rather than reload_completed.
2871 (aarch64_atomic_exchange<mode>): Likewise.
2872 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2873 (atomic_nand<mode>): Likewise.
2874 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2875 (atomic_fetch_nand<mode>): Likewise.
2876 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2877 (atomic_nand_fetch<mode>): Likewise.
2879 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2882 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2884 (REVERSE_CONDITION): Delete.
2885 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2886 (CCFP_CCFPE): Likewise.
2887 (e): New mode attribute.
2888 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2889 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2890 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2891 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2892 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2893 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2894 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2895 name of generator from gen_ccmpdi to gen_ccmpccdi.
2896 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2897 the previous comparison but aren't able to, use the new ccmp_rev
2900 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2902 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2903 than testing directly for INTEGER_CST.
2904 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2906 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2908 PR tree-optimization/93292
2909 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2910 get_vectype_for_scalar_type returns NULL.
2912 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2914 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2915 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2917 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2919 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2921 * params.opt: (max-predicted-iterations): Set bounds.
2922 * predict.c (real_almost_one, real_br_prob_base,
2923 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2924 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2925 probabilities; do not truncate to reg_br_prob_bases.
2926 (estimate_loops_at_level): Pass max_cyclic_prob.
2927 (estimate_loops): Compute max_cyclic_prob.
2928 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2930 * profile-count.c (profile_probability::to_sreal): New.
2931 * profile-count.h (class sreal): Move up in file.
2932 (profile_probability::to_sreal): Declare.
2934 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2937 (arm_invalid_conversion): New function for target hook.
2938 (arm_invalid_unary_op): New function for target hook.
2939 (arm_invalid_binary_op): New function for target hook.
2941 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2943 * config.gcc: Add arm_bf16.h.
2944 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2945 (arm_simd_builtin_std_type): Add BFmode.
2946 (arm_init_simd_builtin_types): Define element types for vector types.
2947 (arm_init_bf16_types): New function.
2948 (arm_init_builtins): Add arm_init_bf16_types function call.
2949 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2950 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2951 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2952 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2953 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2954 (arm_mangle_type): Add __bf16.
2955 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2956 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2957 arm_bf16_ptr_type_node.
2958 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2959 define_split between ARM registers.
2960 * config/arm/arm_bf16.h: New file.
2961 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2962 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2964 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2965 * config/arm/vfp.md: Add BFmode to movhf patterns.
2967 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2968 Andre Vieira <andre.simoesdiasvieira@arm.com>
2970 * config/arm/arm-cpus.in (mve, mve_float): New features.
2971 (dsp, mve, mve.fp): New options.
2972 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2973 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2974 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2976 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2977 Thomas Preud'homme <thomas.preudhomme@arm.com>
2979 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2981 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2982 error for using -mcmse when targeting Armv8.1-M Mainline.
2984 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2985 Thomas Preud'homme <thomas.preudhomme@arm.com>
2987 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2988 address in r4 when targeting Armv8.1-M Mainline.
2989 (nonsecure_call_value_internal): Likewise.
2990 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2991 a register match_operand again. Emit BLXNS when targeting
2993 (nonsecure_call_value_reg_thumb2): Likewise.
2995 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2996 Thomas Preud'homme <thomas.preudhomme@arm.com>
2998 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2999 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3000 variable as true when floating-point ABI is not hard. Replace
3001 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3002 Generate VLSTM and VLLDM instruction respectively before and
3003 after a function call to cmse_nonsecure_call function.
3004 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3005 (VUNSPEC_VLLDM): Likewise.
3006 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3007 (lazy_load_multiple_insn): Likewise.
3009 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3010 Thomas Preud'homme <thomas.preudhomme@arm.com>
3012 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3013 (arm_emit_vfp_multi_reg_pop): Likewise.
3014 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3015 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3016 restore callee-saved VFP registers.
3018 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3019 Thomas Preud'homme <thomas.preudhomme@arm.com>
3021 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3022 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3023 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3024 callee-saved GPRs as well as clear ip register before doing a nonsecure
3025 call then restore callee-saved GPRs after it when targeting
3027 (arm_reorg): Adapt to function rename.
3029 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3030 Thomas Preud'homme <thomas.preudhomme@arm.com>
3032 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3033 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3034 clear_vfp_multiple pattern based on a new vfp parameter.
3035 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3036 targeting Armv8.1-M Mainline.
3037 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3038 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3039 whether VFP registers are available before looking call_used_regs for a
3041 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3042 of prototype of clear_operation_p.
3043 (clear_vfp_multiple_operation): New predicate.
3044 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3045 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3047 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3048 Thomas Preud'homme <thomas.preudhomme@arm.com>
3050 * config/arm/arm-protos.h (clear_operation_p): Declare.
3051 * config/arm/arm.c (clear_operation_p): New function.
3052 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3053 targeting Armv8.1-M Mainline or successor.
3054 (output_return_instruction): Only output APSR register clearing if
3055 Armv8.1-M Mainline instructions not available.
3056 (thumb_exit): Likewise.
3057 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3058 * config/arm/thumb2.md (clear_apsr): New define_insn.
3059 (clear_multiple): Likewise.
3060 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3062 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3063 Thomas Preud'homme <thomas.preudhomme@arm.com>
3065 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3066 (use_return_insn): Also return false for Armv8.1-M Mainline.
3067 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3068 Mainline instructions are available.
3069 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3070 when targeting Armv8.1-M Mainline Security Extensions.
3071 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3072 Mainline entry function.
3073 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3074 targeting Armv8.1-M Mainline or successor.
3075 (arm_expand_epilogue): Fix indentation of caller-saved register
3076 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3078 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3079 (FP_SYSREGS): Likewise.
3080 (enum vfp_sysregs_encoding): Define enum.
3081 (fp_sysreg_names): Declare.
3082 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3083 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3084 (pop_fpsysreg_insn): Likewise.
3086 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3087 Thomas Preud'homme <thomas.preudhomme@arm.com>
3089 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3090 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3091 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3092 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3093 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3094 (ARMv8_1m_main): New feature group.
3095 (armv8.1-m.main): New architecture.
3096 * config/arm/arm-tables.opt: Regenerate.
3097 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3098 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3099 (arm_options_perform_arch_sanity_checks): Error out when targeting
3100 Armv8.1-M Mainline Security Extensions.
3101 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3103 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3105 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3106 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3107 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3108 aarch64_bfdot_laneq): New.
3109 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3110 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3111 vbfdotq_laneq_f32): New.
3112 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3113 VBFMLA_W, VBF): New.
3114 (isquadop): Add V4BF, V8BF.
3116 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3118 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3119 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3120 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3121 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3122 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3123 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3124 usdot_laneq, sudot_lane,sudot_laneq): New.
3125 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3126 (aarch64_<sur>dot_lane): New.
3127 * config/aarch64/arm_neon.h (vusdot_s32): New.
3129 (vusdot_lane_s32): New.
3130 (vsudot_lane_s32): New.
3131 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3132 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3134 2020-01-16 Martin Liska <mliska@suse.cz>
3136 * value-prof.c (dump_histogram_value): Fix
3137 obvious spacing issue.
3139 2020-01-16 Andrew Pinski <apinski@marvell.com>
3141 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3142 !storage_order_barrier_p.
3144 2020-01-16 Andrew Pinski <apinski@marvell.com>
3146 * sched-int.h (_dep): Add unused bit-field field for the padding.
3147 * sched-deps.c (init_dep_1): Init unused field.
3149 2020-01-16 Andrew Pinski <apinski@marvell.com>
3151 * optabs.h (create_expand_operand): Initialize target field also.
3153 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3155 PR tree-optimization/92429
3156 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3157 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3159 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3162 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3164 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3165 aarch64_sve_int_mode to each mode.
3167 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3169 * doc/analyzer.texi (Overview): Add note about
3170 -fdump-ipa-analyzer.
3172 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3174 PR tree-optimization/93231
3175 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3176 input_type is unsigned. Use tree_to_shwi for shift constant.
3177 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3178 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3179 inputs more efficiently.
3181 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3183 * config/i386/i386.md (*movsf_internal): Do not require
3184 SSE2 ISA for alternatives 14 and 15.
3186 2020-01-15 Richard Biener <rguenther@suse.de>
3189 * tree-eh.c (sink_clobbers): If we already visited the destination
3190 block do not defer insertion.
3191 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3192 the purpose of defered insertion.
3194 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3196 * BASE-VER: Bump to 10.0.1.
3198 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3200 PR tree-optimization/93247
3201 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3202 type of the stmt that we're going to vectorize.
3204 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3206 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3207 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3210 2020-01-15 Martin Liska <mliska@suse.cz>
3212 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3213 2 calls of streamer_read_hwi in a function call.
3215 2020-01-15 Richard Biener <rguenther@suse.de>
3217 * alias.c (record_alias_subset): Avoid redundant work when
3218 subset is already recorded.
3220 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3222 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3223 the analyzer options provide CWE identifiers.
3225 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3227 * tree-diagnostic-path.cc (path_summary::event_range::print):
3228 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3229 using get_pure_location.
3231 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3233 PR tree-optimization/93262
3234 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3235 perform head trimming only if the last argument is constant,
3236 either all ones, or larger or equal to head trim, in the latter
3237 case decrease the last argument by head_trim.
3239 PR tree-optimization/93249
3240 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3241 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3242 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3243 perform head trim unless we can prove there are no '\0' chars
3244 from the source among the first head_trim chars.
3246 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3248 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3250 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3253 * config/i386/sse.md
3254 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3255 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3256 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3257 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3258 just a single alternative instead of two, make operands 1 and 2
3261 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3264 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3267 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3269 * Makefile.in (lang_opt_files): Add analyzer.opt.
3270 (ANALYZER_OBJS): New.
3271 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3272 tristate.o and ANALYZER_OBJS.
3273 (TEXI_GCCINT_FILES): Add analyzer.texi.
3274 * common.opt (-fanalyzer): New driver option.
3275 * config.in: Regenerate.
3276 * configure: Regenerate.
3277 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3278 (gccdepdir): Also create depdir for "analyzer" subdir.
3279 * digraph.cc: New file.
3280 * digraph.h: New file.
3281 * doc/analyzer.texi: New file.
3282 * doc/gccint.texi ("Static Analyzer") New menu item.
3283 (analyzer.texi): Include it.
3284 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3285 ("Warning Options"): Add static analysis warnings to the list.
3286 (-Wno-analyzer-double-fclose): New option.
3287 (-Wno-analyzer-double-free): New option.
3288 (-Wno-analyzer-exposure-through-output-file): New option.
3289 (-Wno-analyzer-file-leak): New option.
3290 (-Wno-analyzer-free-of-non-heap): New option.
3291 (-Wno-analyzer-malloc-leak): New option.
3292 (-Wno-analyzer-possible-null-argument): New option.
3293 (-Wno-analyzer-possible-null-dereference): New option.
3294 (-Wno-analyzer-null-argument): New option.
3295 (-Wno-analyzer-null-dereference): New option.
3296 (-Wno-analyzer-stale-setjmp-buffer): New option.
3297 (-Wno-analyzer-tainted-array-index): New option.
3298 (-Wno-analyzer-use-after-free): New option.
3299 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3300 (-Wno-analyzer-use-of-uninitialized-value): New option.
3301 (-Wanalyzer-too-complex): New option.
3302 (-fanalyzer-call-summaries): New warning.
3303 (-fanalyzer-checker=): New warning.
3304 (-fanalyzer-fine-grained): New warning.
3305 (-fno-analyzer-state-merge): New warning.
3306 (-fno-analyzer-state-purge): New warning.
3307 (-fanalyzer-transitivity): New warning.
3308 (-fanalyzer-verbose-edges): New warning.
3309 (-fanalyzer-verbose-state-changes): New warning.
3310 (-fanalyzer-verbosity=): New warning.
3311 (-fdump-analyzer): New warning.
3312 (-fdump-analyzer-callgraph): New warning.
3313 (-fdump-analyzer-exploded-graph): New warning.
3314 (-fdump-analyzer-exploded-nodes): New warning.
3315 (-fdump-analyzer-exploded-nodes-2): New warning.
3316 (-fdump-analyzer-exploded-nodes-3): New warning.
3317 (-fdump-analyzer-supergraph): New warning.
3318 * doc/sourcebuild.texi (dg-require-dot): New.
3319 (dg-check-dot): New.
3320 * gdbinit.in (break-on-saved-diagnostic): New command.
3321 * graphviz.cc: New file.
3322 * graphviz.h: New file.
3323 * ordered-hash-map-tests.cc: New file.
3324 * ordered-hash-map.h: New file.
3325 * passes.def (pass_analyzer): Add before
3326 pass_ipa_whole_program_visibility.
3327 * selftest-run-tests.c (selftest::run_tests): Call
3328 selftest::ordered_hash_map_tests_cc_tests.
3329 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3331 * shortest-paths.h: New file.
3332 * timevar.def (TV_ANALYZER): New timevar.
3333 (TV_ANALYZER_SUPERGRAPH): Likewise.
3334 (TV_ANALYZER_STATE_PURGE): Likewise.
3335 (TV_ANALYZER_PLAN): Likewise.
3336 (TV_ANALYZER_SCC): Likewise.
3337 (TV_ANALYZER_WORKLIST): Likewise.
3338 (TV_ANALYZER_DUMP): Likewise.
3339 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3340 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3341 * tree-pass.h (make_pass_analyzer): New decl.
3342 * tristate.cc: New file.
3343 * tristate.h: New file.
3345 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3348 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3349 alternatives 9 and 10.
3351 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3353 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3354 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3355 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3356 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3357 (selftest::hash_map_tests_c_tests): Call it.
3358 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3359 New static constant, using the value of = H::empty_zero_p.
3360 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3361 from default_hash_traits <Value>.
3362 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3364 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3365 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3366 calls to mark_empty with !Descriptor::empty_zero_p.
3367 (hash_table::empty_slow): Conditionalize the memset call with a
3368 check that Descriptor::empty_zero_p; otherwise, loop through the
3369 entries calling mark_empty on them.
3370 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3371 (pointer_hash::empty_zero_p): Likewise.
3372 (pair_hash::empty_zero_p): Likewise.
3373 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3375 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3376 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3377 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3378 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3379 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3380 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3381 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3382 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3384 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3387 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3389 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3390 fix typo on return value.
3392 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3395 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3397 (cgraph_edge::make_speculative): Add param for setting speculative_id
3399 (cgraph_edge::speculative_call_info): Update comments and find reference
3400 by speculative_id for multiple indirect targets.
3401 (cgraph_edge::resolve_speculation): Decrease the speculations
3402 for indirect edge, drop it's speculative if not direct target
3403 left. Update comments.
3404 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3405 (cgraph_node::dump): Print num_speculative_call_targets.
3406 (cgraph_node::verify_node): Don't report error if speculative
3407 edge not include statement.
3408 (cgraph_edge::num_speculative_call_targets_p): New function.
3409 * cgraph.h (int common_target_id): Remove.
3410 (int common_target_probability): Remove.
3411 (num_speculative_call_targets): New variable.
3412 (make_speculative): Add param for setting speculative_id.
3413 (cgraph_edge::num_speculative_call_targets_p): New declare.
3414 (target_prob): New variable.
3415 (speculative_id): New variable.
3416 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3417 call summaries for multiple speculative call targets.
3418 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3419 * ipa-profile.c (struct speculative_call_target): New struct.
3420 (class speculative_call_summary): New class.
3421 (class speculative_call_summaries): New class.
3422 (call_sums): New variable.
3423 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3424 (ipa_profile_write_edge_summary): New function.
3425 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3426 (ipa_profile_dump_all_summaries): New function.
3427 (ipa_profile_read_edge_summary): New function.
3428 (ipa_profile_read_summary_section): New function.
3429 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3430 (ipa_profile): Generate num_speculative_call_targets from
3432 * ipa-ref.h (speculative_id): New variable.
3433 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3434 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3435 common_target_probability. Stream out speculative_id and
3436 num_speculative_call_targets.
3437 (input_edge): Likewise.
3438 * predict.c (dump_prediction): Remove edges count assert to be
3440 * symtab.c (symtab_node::create_reference): Init speculative_id.
3441 (symtab_node::clone_references): Clone speculative_id.
3442 (symtab_node::clone_referring): Clone speculative_id.
3443 (symtab_node::clone_reference): Clone speculative_id.
3444 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3445 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3446 if indirect call contains multiple speculative targets.
3447 * value-prof.h (check_ic_target): Remove.
3448 * value-prof.c (gimple_value_profile_transformations):
3449 Use void function gimple_ic_transform.
3450 * value-prof.c (gimple_ic_transform): Handle topn case.
3451 Fix comment typos. Change it to a void function.
3453 2020-01-13 Andrew Pinski <apinski@marvell.com>
3455 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3456 (octeontx2t98): New define.
3457 (octeontx2t96): New define.
3458 (octeontx2t93): New define.
3459 (octeontx2f95): New define.
3460 (octeontx2f95n): New define.
3461 (octeontx2f95mm): New define.
3462 * config/aarch64/aarch64-tune.md: Regenerate.
3463 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3465 2020-01-13 Jason Merrill <jason@redhat.com>
3467 PR c++/33799 - destroy return value if local cleanup throws.
3468 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3470 2020-01-13 Martin Liska <mliska@suse.cz>
3472 * ipa-cp.c (get_max_overall_size): Use newly
3473 renamed param param_ipa_cp_unit_growth.
3474 * params.opt: Remove legacy param name.
3476 2020-01-13 Martin Sebor <msebor@redhat.com>
3478 PR tree-optimization/93213
3479 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3480 stores to be eliminated.
3482 2020-01-13 Martin Liska <mliska@suse.cz>
3484 * opts.c (print_help): Do not print CL_PARAM
3485 and CL_WARNING for CL_OPTIMIZATION.
3487 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3490 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3491 depending on optimization settings.
3493 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3495 PR tree-optimization/90838
3496 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3497 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3498 argument rather than to initialize temporary for targets that
3499 don't use the mode argument at all. Initialize ctzval to avoid
3502 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3504 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3505 * tree-core.h: Document it.
3506 * gimplify.c (gimplify_omp_workshare): Set it.
3507 * omp-low.c (lower_omp_target): Use it.
3508 * tree-pretty-print.c (dump_omp_clause): Print it.
3510 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3511 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3513 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3515 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3516 * common.opt (fdiagnostics-path-format=): New option.
3517 (diagnostic_path_format): New enum.
3518 (fdiagnostics-show-path-depths): New option.
3519 * coretypes.h (diagnostic_event_id_t): New forward decl.
3520 * diagnostic-color.c (color_dict): Add "path".
3521 * diagnostic-event-id.h: New file.
3522 * diagnostic-format-json.cc (json_from_expanded_location): Make
3524 (json_end_diagnostic): Call context->make_json_for_path if it
3525 exists and the diagnostic has a path.
3526 (diagnostic_output_format_init): Clear context->print_path.
3527 * diagnostic-path.h: New file.
3528 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3529 when printing a run of events in a diagnostic_path so that they
3530 all get the same color.
3531 (layout::m_diagnostic_path_p): New field.
3532 (layout::layout): Initialize it.
3533 (layout::print_any_labels): Don't colorize the label text for an
3534 event in a diagnostic_path.
3535 (gcc_rich_location::add_location_if_nearby): Add
3536 "restrict_to_current_line_spans" and "label" params. Pass the
3537 former to layout.maybe_add_location_range; pass the latter
3538 when calling add_range.
3539 * diagnostic.c: Include "diagnostic-path.h".
3540 (diagnostic_initialize): Initialize context->path_format and
3541 context->show_path_depths.
3542 (diagnostic_show_any_path): New function.
3543 (diagnostic_path::interprocedural_p): New function.
3544 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3545 (simple_diagnostic_path::num_events): New function.
3546 (simple_diagnostic_path::get_event): New function.
3547 (simple_diagnostic_path::add_event): New function.
3548 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3549 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3550 (debug): New overload taking a diagnostic_path *.
3551 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3552 * diagnostic.h (enum diagnostic_path_format): New enum.
3553 (json::value): New forward decl.
3554 (diagnostic_context::path_format): New field.
3555 (diagnostic_context::show_path_depths): New field.
3556 (diagnostic_context::print_path): New callback field.
3557 (diagnostic_context::make_json_for_path): New callback field.
3558 (diagnostic_show_any_path): New decl.
3559 (json_from_expanded_location): New decl.
3560 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3561 (-fdiagnostics-show-path-depths): New option.
3562 (-fdiagnostics-color): Add "path" to description of default
3563 GCC_COLORS; describe it.
3564 (-fdiagnostics-format=json): Document how diagnostic paths are
3565 represented in the JSON output format.
3566 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3567 Add optional params "restrict_to_current_line_spans" and "label".
3568 * opts.c (common_handle_option): Handle
3569 OPT_fdiagnostics_path_format_ and
3570 OPT_fdiagnostics_show_path_depths.
3571 * pretty-print.c: Include "diagnostic-event-id.h".
3572 (pp_format): Implement "%@" format code for printing
3573 diagnostic_event_id_t *.
3574 (selftest::test_pp_format): Add tests for "%@".
3575 * selftest-run-tests.c (selftest::run_tests): Call
3576 selftest::tree_diagnostic_path_cc_tests.
3577 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3578 * toplev.c (general_init): Initialize global_dc->path_format and
3579 global_dc->show_path_depths.
3580 * tree-diagnostic-path.cc: New file.
3581 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3582 non-static. Drop "diagnostic" param in favor of storing the
3583 original value of "where" and re-using it.
3584 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3585 maybe_unwind_expanded_macro_loc.
3586 (tree_diagnostics_defaults): Initialize context->print_path and
3587 context->make_json_for_path.
3588 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3590 (default_tree_make_json_for_path): New decl.
3591 (maybe_unwind_expanded_macro_loc): New decl.
3593 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3595 PR tree-optimization/93210
3596 * fold-const.h (native_encode_initializer,
3597 can_native_interpret_type_p): Declare.
3598 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3600 (native_encode_initializer): New function, moved from dwarf2out.c.
3601 Adjust to native_encode_expr compatible arguments, including dry-run
3602 and partial extraction modes. Don't handle STRING_CST.
3603 (can_native_interpret_type_p): No longer static.
3604 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3605 offset / BITS_PER_UNIT fits into int and don't call it if
3606 can_native_interpret_type_p fails. If suboff is NULL and for
3607 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3608 native_encode_initializer.
3609 (fold_const_aggregate_ref_1): Formatting fix.
3610 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3611 (tree_add_const_value_attribute): Adjust caller.
3613 PR tree-optimization/90838
3614 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3615 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3616 CTZ_DEFINED_VALUE_AT_ZERO.
3618 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3621 * lra-constraints.c (match_reload): Permit input operands have the
3622 same mode as output while other input operands have a different
3625 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3627 PR tree-optimization/90838
3628 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3629 (check_ctz_string): Likewise.
3630 (optimize_count_trailing_zeroes): Likewise.
3631 (simplify_count_trailing_zeroes): Likewise.
3632 (pass_forwprop::execute): Try ctz simplification.
3633 * match.pd: Add matching for ctz idioms.
3635 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3637 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3639 (aarch64_invalid_unary_op): New function for target hook.
3640 (aarch64_invalid_binary_op): New function for target hook.
3642 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3644 * config.gcc: Add arm_bf16.h.
3645 * config/aarch64/aarch64-builtins.c
3646 (aarch64_simd_builtin_std_type): Add BFmode.
3647 (aarch64_init_simd_builtin_types): Define element types for vector
3649 (aarch64_init_bf16_types): New function.
3650 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3651 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3653 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3654 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3656 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3657 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3658 * config/aarch64/aarch64.c
3659 (aarch64_classify_vector_mode): Add support for BF types.
3660 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3661 (aarch64_vq_mode): Add support for BF types.
3662 (aarch64_simd_container_mode): Add support for BF types.
3663 (aarch64_mangle_type): Add support for BF scalar type.
3664 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3665 * config/aarch64/arm_bf16.h: New file.
3666 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3667 * config/aarch64/iterators.md: Add BF types to mode attributes.
3668 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3670 2020-01-10 Jason Merrill <jason@redhat.com>
3672 PR c++/93173 - incorrect tree sharing.
3673 * gimplify.c (copy_if_shared): No longer static.
3674 * gimplify.h: Declare it.
3676 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3678 * doc/invoke.texi (-msve-vector-bits=): Document that
3679 -msve-vector-bits=128 now generates VL-specific code for
3680 little-endian targets.
3681 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3682 build_vector_type_for_mode to construct the data vector types.
3683 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3684 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3685 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3686 for 128-bit vectors.
3688 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3690 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3693 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3695 * config/aarch64/aarch64-builtins.c
3696 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3697 rather than checking the number of elements and the element mode.
3699 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3701 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3702 get_related_vectype_for_scalar_type rather than build_vector_type
3703 to create the index type for a conditional reduction.
3705 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3707 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3708 for any type of gather or scatter, including strided accesses.
3710 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3712 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3715 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3717 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3719 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3720 parameter and its use to reset DR_OFFSET's.
3721 (vect_transform_loop): Remove orig_drs_init argument.
3722 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3723 member of dr_vec_info rather than the offset of the associated
3724 data_reference's innermost_loop_behavior.
3725 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3726 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3727 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3728 get_dr_vinfo_offset.
3729 (vectorizable_store): Likewise.
3730 (vectorizable_load): Likewise.
3732 2020-01-10 Richard Biener <rguenther@suse.de>
3734 * gimple-ssa-store-merging
3735 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3737 2020-01-10 Martin Liska <mliska@suse.cz>
3740 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3741 encapsulation that was there before r280040.
3743 2020-01-10 Richard Biener <rguenther@suse.de>
3746 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3747 sequences to avoid walking them again for secondary opportunities.
3748 (pass_lower_eh_dispatch::execute): Instead actually insert
3751 2020-01-10 Richard Biener <rguenther@suse.de>
3754 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3755 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3756 avoid quadraticness.
3758 2020-01-10 Martin Jambor <mjambor@suse.cz>
3760 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3761 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3762 to get param_ipa_sra_max_replacements.
3763 (param_splitting_across_edge): Pass the caller to
3764 pull_accesses_from_callee.
3766 2020-01-10 Martin Jambor <mjambor@suse.cz>
3768 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3769 * ipa-cp.c (max_new_size): Removed.
3770 (orig_overall_size): New variable.
3771 (get_max_overall_size): New function.
3772 (estimate_local_effects): Use it. Adjust dump.
3773 (decide_about_value): Likewise.
3774 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3775 orig_overall_size. Adjust dump.
3776 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3778 2020-01-10 Martin Jambor <mjambor@suse.cz>
3780 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3781 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3782 instead of param_ipa_max_agg_items.
3783 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3784 optimization info for the callee.
3786 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3788 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3789 markers if debug_inline_points is false.
3791 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3793 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3795 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3796 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3797 aarch64-sve-builtins-sve2.h.
3798 (aarch64-sve-builtins-sve2.o): New rule.
3799 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3800 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3801 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3802 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3803 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3804 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3806 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3807 instructions that are handled here.
3808 (@cond_asrd<mode>): Generalize to...
3809 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3810 (*cond_asrd<mode>_2): Generalize to...
3811 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3812 (*cond_asrd<mode>_z): Generalize to...
3813 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3814 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3815 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3816 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3817 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3819 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3820 (@aarch64_scatter_stnt<mode>): Likewise.
3821 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3822 (@aarch64_mul_lane_<mode>): Likewise.
3823 (@aarch64_sve_suqadd<mode>_const): Likewise.
3824 (*<sur>h<addsub><mode>): Generalize to...
3825 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3827 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3828 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3829 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3830 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3831 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3832 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3833 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3834 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3835 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3836 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3837 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3838 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3839 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3840 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3841 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3842 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3843 (@aarch64_sve2_xar<mode>): Likewise.
3844 (@aarch64_sve2_bcax<mode>): Likewise.
3845 (*aarch64_sve2_eor3<mode>): Rename to...
3846 (@aarch64_sve2_eor3<mode>): ...this.
3847 (@aarch64_sve2_bsl<mode>): New expander.
3848 (@aarch64_sve2_nbsl<mode>): Likewise.
3849 (@aarch64_sve2_bsl1n<mode>): Likewise.
3850 (@aarch64_sve2_bsl2n<mode>): Likewise.
3851 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3852 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3853 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3854 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3855 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3856 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3857 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3858 (<su>mull<bt><Vwide>): Generalize to...
3859 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3861 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3862 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3863 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3864 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3865 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3866 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3867 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3868 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3869 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3870 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3871 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3872 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3873 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3874 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3875 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3876 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3877 (<SHRNB:r>shrnb<mode>): Generalize to...
3878 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3880 (<SHRNT:r>shrnt<mode>): Generalize to...
3881 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3883 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3884 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3885 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3886 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3887 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3888 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3889 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3890 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3891 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3892 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3893 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3894 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3895 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3896 (@aarch64_sve2_cvtnt<mode>): Likewise.
3897 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3898 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3899 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3900 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3901 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3902 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3903 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3904 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3905 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3906 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3907 (@aarch64_sve2_pmul<mode>): Likewise.
3908 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3909 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3910 (@aarch64_sve2_tbl2<mode>): Likewise.
3911 (@aarch64_sve2_tbx<mode>): Likewise.
3912 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3913 (@aarch64_sve2_histcnt<mode>): Likewise.
3914 (@aarch64_sve2_histseg<mode>): Likewise.
3915 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3916 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3917 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3918 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3919 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3920 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3921 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3922 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3923 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3924 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3925 (SVE2_PMULL_PAIR_I): New mode iterators.
3926 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3927 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3928 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3929 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3930 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3931 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3932 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3933 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3934 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3935 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3936 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3937 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3938 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3939 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3940 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3941 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3942 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3943 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3944 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3945 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3946 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3947 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3948 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3949 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3950 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3951 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3952 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3953 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3954 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3955 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3956 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3958 (VNARROW, Ventype): New mode attributes.
3959 (Vewtype): Handle VNx2DI. Fix typo in comment.
3960 (VDOUBLE): New mode attribute.
3961 (sve_lane_con): Handle VNx8HI.
3962 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3963 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3964 (sve_int_op, sve_int_op_rev): Handle the above codes.
3965 (sve_pred_int_rhs2_operand): Likewise.
3966 (MULLBT, SHRNB, SHRNT): Delete.
3967 (SVE_INT_SHIFT_IMM): New int iterator.
3968 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3969 and UNSPEC_WHILEHS for TARGET_SVE2.
3970 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3971 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3972 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3973 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3974 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3975 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3976 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3977 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3978 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3979 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3980 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3981 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3982 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3983 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3984 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3985 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3986 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3987 (optab): Handle the new unspecs.
3988 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3990 (lr): Handle the new unspecs.
3992 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3993 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3994 (sve_int_qsub_op): New int attributes.
3995 (sve_fp_op, rot): Handle the new unspecs.
3996 * config/aarch64/aarch64-sve-builtins.h
3997 (function_resolver::require_matching_pointer_type): Declare.
3998 (function_resolver::resolve_unary): Add an optional boolean argument.
3999 (function_resolver::finish_opt_n_resolution): Add an optional
4000 type_suffix_index argument.
4001 (gimple_folder::redirect_call): Declare.
4002 (gimple_expander::prepare_gather_address_operands): Add an optional
4004 * config/aarch64/aarch64-sve-builtins.cc: Include
4005 aarch64-sve-builtins-sve2.h.
4006 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4007 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4008 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4009 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4010 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4011 (TYPES_s_unsigned): Likewise.
4012 (TYPES_s_integer): Use TYPES_s_unsigned.
4013 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4014 (TYPES_sd_integer): Use them.
4015 (TYPES_d_unsigned): New macro.
4016 (TYPES_d_integer): Use it.
4017 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4018 (TYPES_cvt_narrow): Likewise.
4019 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4020 (preds_mx): New variable.
4021 (function_builder::add_overloaded_function): Allow the new feature
4022 set to be more restrictive than the original one.
4023 (function_resolver::infer_pointer_type): Remove qualifiers from
4024 the pointer type before printing it.
4025 (function_resolver::require_matching_pointer_type): New function.
4026 (function_resolver::resolve_sv_displacement): Handle functions
4027 that don't support 32-bit vector indices or svint32_t vector offsets.
4028 (function_resolver::finish_opt_n_resolution): Take the inferred type
4029 as a separate argument.
4030 (function_resolver::resolve_unary): Optionally treat all forms in
4031 the same way as normal merging functions.
4032 (gimple_folder::redirect_call): New function.
4033 (function_expander::prepare_gather_address_operands): Add an argument
4034 that says whether scaled forms are available. If they aren't,
4035 handle scaling of vector indices and don't add the extension and
4037 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4038 fall back to using cond_* instead.
4039 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4040 Split out the member variables into...
4041 (rtx_code_function_base): ...this new base class.
4042 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4043 (unspec_based_function): Split out the member variables into...
4044 (unspec_based_function_base): ...this new base class.
4045 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4046 (unspec_based_function_exact_insn): New class.
4047 (unspec_based_add_function, unspec_based_add_lane_function)
4048 (unspec_based_lane_function, unspec_based_pred_function)
4049 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4050 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4051 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4053 (unspec_based_fused_function): New class.
4054 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4055 (unspec_based_fused_lane_function): New class.
4056 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4058 (CODE_FOR_MODE1): New macro.
4059 (fixed_insn_function): New class.
4060 (while_comparison): Likewise.
4061 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4062 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4063 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4064 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4065 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4066 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4067 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4068 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4069 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4070 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4071 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4072 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4073 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4074 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4075 Also add an initial argument for unary_convert_narrowt, regardless
4076 of the predication type.
4077 (build_32_64): Allow loads and stores to specify MODE_none.
4078 (build_sv_index64, build_sv_uint_offset): New functions.
4079 (long_type_suffix): New function.
4080 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4081 (binary_imm_long_base, load_gather_sv_base): Likewise.
4082 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4083 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4084 (unary_narrowb_base, unary_narrowt_base): Likewise.
4085 (binary_long_lane_def, binary_long_lane): New shape.
4086 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4087 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4088 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4089 (binary_to_uint_def, binary_to_uint): Likewise.
4090 (binary_wide_def, binary_wide): Likewise.
4091 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4092 (compare_def, compare): Likewise.
4093 (compare_ptr_def, compare_ptr): Likewise.
4094 (load_ext_gather_index_restricted_def,
4095 load_ext_gather_index_restricted): Likewise.
4096 (load_ext_gather_offset_restricted_def,
4097 load_ext_gather_offset_restricted): Likewise.
4098 (load_gather_sv_def): Inherit from load_gather_sv_base.
4099 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4100 (shift_left_imm_def, shift_left_imm): Likewise.
4101 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4102 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4103 (store_scatter_index_restricted_def,
4104 store_scatter_index_restricted): Likewise.
4105 (store_scatter_offset_restricted_def,
4106 store_scatter_offset_restricted): Likewise.
4107 (tbl_tuple_def, tbl_tuple): Likewise.
4108 (ternary_long_lane_def, ternary_long_lane): Likewise.
4109 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4110 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4111 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4112 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4113 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4114 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4115 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4116 (ternary_uint_def, ternary_uint): Likewise.
4117 (unary_convert): Fix typo in comment.
4118 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4119 (unary_long_def, unary_long): Likewise.
4120 (unary_narrowb_def, unary_narrowb): Likewise.
4121 (unary_narrowt_def, unary_narrowt): Likewise.
4122 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4123 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4124 (unary_to_int_def, unary_to_int): Likewise.
4125 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4126 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4127 (svasrd_impl): Delete.
4128 (svcadd_impl::expand): Handle integer operations too.
4129 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4130 new functions to derive the unspec numbers.
4131 (svmla_svmls_lane_impl): Replace with...
4132 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4133 integer operations too.
4134 (svwhile_impl): Rename to...
4135 (svwhilelx_impl): ...this and inherit from while_comparison.
4136 (svasrd): Use unspec_based_function.
4137 (svmla_lane): Use svmla_lane_impl.
4138 (svmls_lane): Use svmls_lane_impl.
4139 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4140 (svwhilele, svwhilelt): Use svwhilelx_impl.
4141 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4142 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4143 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4144 * config/aarch64/aarch64-sve-builtins.def: Include
4145 aarch64-sve-builtins-sve2.def.
4147 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4149 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4150 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4151 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4152 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4153 immediates as well as vector ones.
4154 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4155 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4156 (aarch64_sve_qsub_immediate): Update calls accordingly.
4158 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4160 * config/aarch64/aarch64-sve2.md: Add banner comments.
4161 (<su>mulh<r>s<mode>3): Move further up file.
4162 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4163 (*aarch64_sve2_sra<mode>): Move further down file.
4164 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4166 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4168 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4170 (while_optab_cmp): Handle them.
4171 * config/aarch64/aarch64-sve.md
4172 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4173 and add a "@" marker.
4174 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4175 instead of gen_aarch64_sve2_while_ptest.
4176 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4178 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4180 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4181 (UNSPEC_WHILELE): ...this.
4182 (UNSPEC_WHILE_LO): Rename to...
4183 (UNSPEC_WHILELO): ...this.
4184 (UNSPEC_WHILE_LS): Rename to...
4185 (UNSPEC_WHILELS): ...this.
4186 (UNSPEC_WHILE_LT): Rename to...
4187 (UNSPEC_WHILELT): ...this.
4188 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4189 (cmp_op, while_optab_cmp): Likewise.
4190 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4191 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4192 (svwhilelt): Likewise.
4194 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4196 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4197 (unary_to_uint): Define.
4198 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4199 (unary_count): Rename to...
4200 (unary_to_uint_def, unary_to_uint): ...this.
4201 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4203 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4205 * config/aarch64/aarch64-sve-builtins-functions.h
4206 (code_for_mode_function): New class.
4207 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4208 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4209 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4210 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4211 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4213 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4215 * config/aarch64/iterators.md (addsub): New code attribute.
4216 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4218 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4219 in the asm string and attributes. Fix indentation.
4220 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4222 (@aarch64_sve_<optab><mode>): ...this.
4223 * config/aarch64/aarch64-sve-builtins.h
4224 (function_expander::expand_signed_unpred_op): Delete.
4225 * config/aarch64/aarch64-sve-builtins.cc
4226 (function_expander::expand_signed_unpred_op): Likewise.
4227 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4228 try using code_for_aarch64_sve instead.
4229 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4230 (svqsub_impl): Likewise.
4231 (svqadd, svqsub): Use rtx_code_function instead.
4233 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4235 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4236 (HADDSUB, sur, addsub): Remove them.
4238 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4240 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4241 functions the same way, rather than singling out those that
4242 aren't mapped directly to optabs.
4244 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4246 * target.def (compatible_vector_types_p): New target hook.
4247 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4248 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4249 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4250 * doc/tm.texi: Regenerate.
4251 * gimple-expr.c: Include target.h.
4252 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4253 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4255 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4256 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4257 Use the original predicate if it already has a suitable type.
4259 2020-01-09 Martin Jambor <mjambor@suse.cz>
4261 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4262 resolve_speculation and redirect_call_stmt_to_callee static. Change
4263 return type of set_call_stmt to cgraph_edge *.
4264 * auto-profile.c (afdo_indirect_call): Adjust call to
4265 redirect_call_stmt_to_callee.
4266 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4267 make the this pointer explicit, adjust self-recursive calls and the
4268 call top make_direct. Return the resulting edge.
4269 (cgraph_edge::remove): Make this pointer explicit.
4270 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4271 (cgraph_edge::make_direct): Likewise, adjust call to
4272 resolve_speculation.
4273 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4274 call to set_call_stmt.
4275 (cgraph_update_edges_for_call_stmt_node): Update call to
4276 set_call_stmt and remove.
4277 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4278 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4279 (cgraph_node::create_edge_including_clones): Moved "first" definition
4280 of edge to the block where it was used. Adjusted calls to
4282 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4283 cgraph_edge::remove.
4284 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4285 make_direct and redirect_call_stmt_to_callee.
4286 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4287 resolve_speculation and make_direct.
4288 * ipa-inline-transform.c (inline_transform): Adjust call to
4289 redirect_call_stmt_to_callee.
4290 (check_speculations_1):: Adjust call to resolve_speculation.
4291 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4292 resolve-speculation.
4293 (inline_small_functions): Adjust call to resolve_speculation.
4294 (ipa_inline): Likewise.
4295 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4297 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4298 safe with regards to edge removal, adjust calls to
4299 redirect_call_stmt_to_callee.
4300 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4301 and redirect_call_stmt_to_callee.
4302 * multiple_target.c (create_dispatcher_calls): Adjust call to
4303 redirect_call_stmt_to_callee
4304 (redirect_to_specific_clone): Likewise.
4305 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4306 Adjust calls to cgraph_edge::remove.
4307 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4308 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4309 (expand_call_inline): Adjust call to cgraph_edge::remove.
4311 2020-01-09 Martin Liska <mliska@suse.cz>
4313 * params.opt: Set Optimization for
4314 param_max_speculative_devirt_maydefs.
4316 2020-01-09 Martin Sebor <msebor@redhat.com>
4320 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4322 2020-01-09 Martin Liska <mliska@suse.cz>
4324 * auto-profile.c (auto_profile): Use opt_for_fn
4326 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4327 (propagate_vals_across_arith_jfunc): Likewise.
4328 (hint_time_bonus): Likewise.
4329 (incorporate_penalties): Likewise.
4330 (good_cloning_opportunity_p): Likewise.
4331 (perform_estimation_of_a_value): Likewise.
4332 (estimate_local_effects): Likewise.
4333 (ipcp_propagate_stage): Likewise.
4334 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4335 (set_switch_stmt_execution_predicate): Likewise.
4336 (analyze_function_body): Likewise.
4337 * ipa-inline-analysis.c (offline_size): Likewise.
4338 * ipa-inline.c (early_inliner): Likewise.
4339 * ipa-prop.c (ipa_analyze_node): Likewise.
4340 (ipcp_transform_function): Likewise.
4341 * ipa-sra.c (process_scan_results): Likewise.
4342 (ipa_sra_summarize_function): Likewise.
4343 * params.opt: Rename ipcp-unit-growth to
4344 ipa-cp-unit-growth. Add Optimization for various
4345 IPA-related parameters.
4347 2020-01-09 Richard Biener <rguenther@suse.de>
4350 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4352 2020-01-09 Richard Biener <rguenther@suse.de>
4354 PR tree-optimization/93040
4355 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4357 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4359 * common/config/avr/avr-common.c (avr_option_optimization_table)
4360 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4362 2020-01-09 Martin Liska <mliska@suse.cz>
4364 * cgraphclones.c (symbol_table::materialize_all_clones):
4365 Use cgraph_node::dump_name.
4367 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4370 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4371 output_operand_lossage instead of gcc_unreachable.
4372 * doc/md.texi (riscv f constraint): Fix typo.
4375 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4376 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4377 CONST_SCALAR_INT_P instead of CONST_INT_P.
4378 (*subv<mode>4_1): Rename to ...
4379 (subv<mode>4_1): ... this.
4380 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4381 define_insn_and_split patterns.
4382 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4385 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4387 * vec.c (class selftest::count_dtor): New class.
4388 (selftest::test_auto_delete_vec): New test.
4389 (selftest::vec_c_tests): Call it.
4390 * vec.h (class auto_delete_vec): New class template.
4391 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4393 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4395 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4397 2020-01-08 Jim Wilson <jimw@sifive.com>
4399 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4400 use of TLS_MODEL_LOCAL_EXEC when not pic.
4402 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4404 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4407 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4410 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4411 *stack_protect_set_3 peephole2): Also check that the second
4412 insns source is general_operand.
4415 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4416 predicate for output operand instead of register_operand.
4417 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4418 memory destination and non-memory operands[2].
4420 2020-01-08 Martin Liska <mliska@suse.cz>
4422 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4423 ::dump_asm_name instead of (::name or ::asm_name).
4424 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4425 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4426 (analyze_functions): Likewise.
4427 (expand_all_functions): Likewise.
4428 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4429 (propagate_bits_across_jump_function): Likewise.
4430 (dump_profile_updates): Likewise.
4431 (ipcp_store_bits_results): Likewise.
4432 (ipcp_store_vr_results): Likewise.
4433 * ipa-devirt.c (dump_targets): Likewise.
4434 * ipa-fnsummary.c (analyze_function_body): Likewise.
4435 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4436 (process_hsa_functions): Likewise.
4437 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4438 (set_alias_uids): Likewise.
4439 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4440 * ipa-inline.c (recursive_inlining): Likewise.
4441 (inline_to_all_callers_1): Likewise.
4442 (ipa_inline): Likewise.
4443 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4444 (ipa_propagate_frequency): Likewise.
4445 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4446 (remove_described_reference): Likewise.
4447 * ipa-pure-const.c (worse_state): Likewise.
4448 (check_retval_uses): Likewise.
4449 (analyze_function): Likewise.
4450 (propagate_pure_const): Likewise.
4451 (propagate_nothrow): Likewise.
4452 (dump_malloc_lattice): Likewise.
4453 (propagate_malloc): Likewise.
4454 (pass_local_pure_const::execute): Likewise.
4455 * ipa-visibility.c (optimize_weakref): Likewise.
4456 (function_and_variable_visibility): Likewise.
4457 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4458 (ipa_discover_variable_flags): Likewise.
4459 * lto-streamer-out.c (output_function): Likewise.
4460 (output_constructor): Likewise.
4461 * tree-inline.c (copy_bb): Likewise.
4462 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4463 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4465 2020-01-08 Richard Biener <rguenther@suse.de>
4468 * tree-eh.c (sink_clobbers): Update virtual operands for
4469 the first and last stmt only. Add a dry-run capability.
4470 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4471 after CFG manipulations and in RPO order to catch all
4472 secondary opportunities reliably.
4474 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4477 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4479 2019-01-08 Richard Biener <rguenther@suse.de>
4482 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4483 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4484 virtual operand, also updating SSA use.
4485 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4486 Update stmt after resetting virtual operand.
4487 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4488 * gimple-iterator.c (gsi_remove): When not removing the stmt
4489 permanently do not delink immediate uses or mark the stmt modified.
4491 2020-01-08 Martin Liska <mliska@suse.cz>
4493 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4494 (ipa_call_context::estimate_size_and_time): Likewise.
4495 (inline_analyze_function): Likewise.
4497 2020-01-08 Martin Liska <mliska@suse.cz>
4499 * cgraph.c (cgraph_node::dump): Use systematically
4502 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4504 Add -nodevicespecs option for avr.
4507 * config/avr/avr.opt (-nodevicespecs): New driver option.
4508 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4509 "-specs=device-specs/..." if that option is not set.
4510 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4512 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4514 Implement 64-bit double functions for avr.
4517 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4518 --with-double-comparison.
4519 * doc/install.texi: Document them.
4520 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4521 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4522 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4523 * doc/invoke.texi (AVR Built-in Macros): Document them.
4524 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4525 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4526 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4528 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4531 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4532 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4533 when only building rm-profile multilibs.
4535 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4538 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4539 lattice for a value to check.
4540 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4541 finite propagation in self-recursive scc.
4543 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4545 * ipa-inline.c (caller_growth_limits): Restore the AND.
4547 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4549 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4550 (VEC_ALLREG_ALT): New iterator.
4551 (VEC_ALLREG_INT_MODE): New iterator.
4552 (VCMP_MODE): New iterator.
4553 (VCMP_MODE_INT): New iterator.
4554 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4555 (vec_cmp<u>v64qidi): New define_expand.
4556 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4557 (vec_cmpu<mode>di_exec): New define_expand.
4558 (vec_cmp<u>v64qidi_exec): New define_expand.
4559 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4560 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4561 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4562 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4563 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4564 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4565 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4566 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4567 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4568 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4570 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4571 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4573 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4575 * config/gcn/constraints.md (DA): Update description and match.
4577 (Db): New constraint.
4578 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4580 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4581 Implement 'Db' mixed immediate type.
4582 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4583 (addcv64si3_dup<exec_vcc>): Delete.
4584 (subcv64si3<exec_vcc>): Rework constraints.
4585 (addv64di3): Rework constraints.
4586 (addv64di3_exec): Rework constraints.
4587 (subv64di3): Rework constraints.
4588 (addv64di3_dup): Delete.
4589 (addv64di3_dup_exec): Delete.
4590 (addv64di3_zext): Rework constraints.
4591 (addv64di3_zext_exec): Rework constraints.
4592 (addv64di3_zext_dup): Rework constraints.
4593 (addv64di3_zext_dup_exec): Rework constraints.
4594 (addv64di3_zext_dup2): Rework constraints.
4595 (addv64di3_zext_dup2_exec): Rework constraints.
4596 (addv64di3_sext_dup2): Rework constraints.
4597 (addv64di3_sext_dup2_exec): Rework constraints.
4599 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4601 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4602 existing target checks.
4604 2020-01-07 Richard Biener <rguenther@suse.de>
4606 * doc/install.texi: Bump minimal supported MPC version.
4608 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4610 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4611 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4612 * langhooks.c: Include stor-layout.h.
4613 (lhd_simulate_enum_decl): New function.
4614 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4615 handle_arm_sve_h for the LTO frontend.
4616 (register_vector_type): Cope with null returns from pushdecl.
4618 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4620 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4621 (aarch64_sve::nvectors_if_data_type): Replace with...
4622 (aarch64_sve::builtin_type_p): ...this.
4623 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4624 (find_vector_type): Delete.
4625 (add_sve_type_attribute): New function.
4626 (lookup_sve_type_attribute): Likewise.
4627 (register_builtin_types): Add an "SVE type" attribute to each type.
4628 (register_tuple_type): Likewise.
4629 (svbool_type_p, nvectors_if_data_type): Delete.
4630 (mangle_builtin_type): Use lookup_sve_type_attribute.
4631 (builtin_type_p): Likewise. Add an overload that returns the
4632 number of constituent vector and predicate registers.
4633 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4634 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4635 instead of aarch64_sve_argument_p.
4636 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4637 (aarch64_pass_by_reference): Likewise.
4638 (aarch64_function_value_1): Likewise.
4639 (aarch64_return_in_memory): Likewise.
4640 (aarch64_layout_arg): Likewise.
4642 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4644 PR tree-optimization/93156
4645 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4646 least significant bit is always clear.
4648 PR tree-optimization/93118
4649 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4650 simplifier with two intermediate conversions.
4652 2020-01-07 Martin Liska <mliska@suse.cz>
4654 * params.opt: Add Optimization for various parameters.
4656 2020-01-07 Martin Liska <mliska@suse.cz>
4659 * doc/extend.texi: Explain cloning for target_clone
4662 2020-01-07 Martin Liska <mliska@suse.cz>
4664 PR tree-optimization/92860
4665 * common.opt: Make in Optimization option
4666 as it is affected by -O0, which is an Optimization
4668 * tree-inline.c (tree_inlinable_function_p):
4669 Use opt_for_fn for warn_inline.
4670 (expand_call_inline): Likewise.
4672 2020-01-07 Martin Liska <mliska@suse.cz>
4674 PR tree-optimization/92860
4675 * common.opt: Make flag_ree as optimization
4678 2020-01-07 Martin Liska <mliska@suse.cz>
4680 PR optimization/92860
4681 * params.opt: Mark param_min_crossjump_insns with Optimization
4684 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4686 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4687 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4689 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4691 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4692 helper function to return the valid addressing formats for a given
4693 hard register and mode.
4694 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4696 * config/rs6000/constraints.md (Q constraint): Update
4698 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4701 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4702 Use 'Q' for doing vector extract from memory.
4703 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4705 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4706 doing vector extract from memory.
4707 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4708 extract from memory.
4710 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4711 for the offset being 34-bits when -mcpu=future is used.
4713 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4715 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4716 instead of cmpib_comparison_operator in cmpib patterns.
4717 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4718 of cmpib_comparison_operator. Revise comment.
4720 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4722 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4723 in an IFN_DIV_POW2 node to be equal.
4725 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4727 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4728 (vect_check_scalar_mask): ...this.
4729 (vectorizable_store, vectorizable_load): Update call accordingly.
4730 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4731 argument in calls to conditional internal functions.
4733 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4735 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4736 '0' matching inputs.
4737 (subv64di3_exec): Likewise.
4739 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4741 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4742 * doc/md.texi (movstr): Likewise.
4744 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4746 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4749 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4751 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4753 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4754 to a temporary file and use move-if-change to update the real
4755 file where necessary.
4757 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4759 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4760 rather than Upa for CPY /M.
4762 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4764 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4767 2020-01-06 Martin Liska <mliska@suse.cz>
4769 PR tree-optimization/92860
4770 * params.opt: Mark param_max_combine_insns with Optimization
4773 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4776 * config/i386/i386.md (SWIDWI): New mode iterator.
4777 (DWI, dwi): Add TImode variants.
4778 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4779 <general_hilo_operand> instead of <general_operand>. Use
4780 CONST_SCALAR_INT_P instead of CONST_INT_P.
4781 (*addv<mode>4_1): Rename to ...
4782 (addv<mode>4_1): ... this.
4783 (QWI): New mode attribute.
4784 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4785 define_insn_and_split patterns.
4786 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4788 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4789 <general_hilo_operand> instead of <general_operand>.
4790 (*addcarry<mode>_1): New define_insn.
4791 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4793 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4795 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4796 Use "call" instead of "set".
4798 2020-01-03 Martin Jambor <mjambor@suse.cz>
4801 * ipa-cp.c (print_all_lattices): Skip functions without info.
4803 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4806 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4807 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4808 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4809 for 'e' simd clones.
4812 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4814 (mprefer-vector-width=): Add Save.
4815 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4816 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4817 (ix86_debug_options, ix86_function_specific_print): Adjust
4818 ix86_target_string callers.
4819 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4820 (ix86_valid_target_attribute_tree): Likewise.
4821 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4822 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4823 ix86_target_string caller.
4826 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4827 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4828 instead of gen_int_shift_amount + convert_modes.
4830 PR rtl-optimization/93088
4831 * loop-iv.c (find_single_def_src): Punt after looking through
4832 128 reg copies for regs with single definitions. Move definitions
4835 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4837 * config/arm/arm-c.c (arm_cpu_builtins): Define
4838 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4839 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4840 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4841 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4842 * config/arm/arm-tables.opt: Regenerated.
4843 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4844 arm_arch_i8mm and arm_arch_bf16 when enabled.
4845 * config/arm/arm.h (TARGET_I8MM): New macro.
4846 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4847 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4848 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4849 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4850 (v8_6_a_simd_variants): New.
4851 (v8_*_a_simd_variants): Add i8mm and bf16.
4852 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4854 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4857 * predict.c (compute_function_frequency): Don't call
4858 warn_function_cold on functions that already have cold attribute.
4860 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4863 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4864 COMDAT group function labels in .data.rel.ro.local section.
4865 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4868 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4869 comparison_operator in B and S integer comparisons. Likewise, use
4870 ordered_comparison_operator instead of cmpib_comparison_operator in
4872 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4874 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4876 Update copyright years.
4878 * gcc.c (process_command): Update copyright notice dates.
4879 * gcov-dump.c (print_version): Ditto.
4880 * gcov.c (print_version): Ditto.
4881 * gcov-tool.c (print_version): Ditto.
4882 * gengtype.c (create_file): Ditto.
4883 * doc/cpp.texi: Bump @copying's copyright year.
4884 * doc/cppinternals.texi: Ditto.
4885 * doc/gcc.texi: Ditto.
4886 * doc/gccint.texi: Ditto.
4887 * doc/gcov.texi: Ditto.
4888 * doc/install.texi: Ditto.
4889 * doc/invoke.texi: Ditto.
4891 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4893 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4896 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4898 PR tree-optimization/93098
4899 * match.pd (popcount): For shift amounts, use integer_onep
4900 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4901 tests. Make sure that precision is power of two larger than or equal
4902 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4903 instead of ULL suffixed constants. Formatting fixes.
4905 Copyright (C) 2020 Free Software Foundation, Inc.
4907 Copying and distribution of this file, with or without modification,
4908 are permitted in any medium without royalty provided the copyright
4909 notice and this notice are preserved.