1 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
4 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
7 2020-02-28 Martin Liska <mliska@suse.cz>
10 * configure.ac: Improve detection of ld_date by requiring
11 either two dashes or none.
12 * configure: Regenerate.
14 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
16 PR rtl-optimization/93564
17 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
18 do not honor reg alloc order.
20 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
23 * config/aarch64/aarch64.c (aarch64_override_options): Fix
24 misleading warning string.
26 2020-02-27 Martin Sebor <msebor@redhat.com>
28 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
30 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
33 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
34 Split the insn into two parts. This insn only does variable
35 extract from a register.
36 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
37 variable extract from memory.
38 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
39 only does variable extract from a register.
40 (vsx_extract_v4sf_var_load): New insn, do variable extract from
42 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
43 into two parts. This insn only does variable extract from a
45 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
46 do variable extract from memory.
48 2020-02-27 Martin Jambor <mjambor@suse.cz>
49 Feng Xue <fxue@os.amperecomputing.com>
52 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
53 new function calls_same_node_or_its_all_contexts_clone_p.
54 (cgraph_edge_brings_value_p): Use it.
55 (cgraph_edge_brings_value_p): Likewise.
56 (self_recursive_pass_through_p): Return false if caller is a clone.
57 (self_recursive_agg_pass_through_p): Likewise.
59 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
62 * alias.c (ends_tbaa_access_path_p): Break out from ...
63 (component_uses_parent_alias_set_from): ... here.
64 * alias.h (ends_tbaa_access_path_p): Declare.
65 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
66 handle trailing arrays past end of tbaa access path.
67 (aliasing_component_refs_p): ... here; likewise.
68 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
69 path; disambiguate also past end of it.
70 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
73 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
75 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
76 beginning of the file.
77 (vcreate_bf16, vcombine_bf16): New.
78 (vdup_n_bf16, vdupq_n_bf16): New.
79 (vdup_lane_bf16, vdup_laneq_bf16): New.
80 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
81 (vduph_lane_bf16, vduph_laneq_bf16): New.
82 (vset_lane_bf16, vsetq_lane_bf16): New.
83 (vget_lane_bf16, vgetq_lane_bf16): New.
84 (vget_high_bf16, vget_low_bf16): New.
85 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
86 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
87 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
88 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
89 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
90 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
91 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
92 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
93 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
94 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
95 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
96 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
97 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
98 (vreinterpretq_bf16_p128): New.
99 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
100 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
101 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
102 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
103 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
104 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
105 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
106 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
107 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
108 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
109 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
110 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
111 (vreinterpretq_p128_bf16): New.
112 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
114 (V_elem_l): Likewise.
117 (V_DOUBLE): Likewise.
118 (VDQX): Add V4BF and V8BF.
119 (V_two_elem, V_three_elem, V_four_elem): Likewise.
122 (V_double_vector_mode): Likewise.
123 (V_cmp_result): Likewise.
124 (V_uf_sclr): Likewise.
125 (V_sz_elem): Likewise.
126 (Is_d_reg): Likewise.
127 (V_mode_nunits): Likewise.
128 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
130 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
132 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
133 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
134 (<expander><mode>3<exec>): Likewise.
135 (<expander><mode>3): New.
136 (v<expander><mode>3): New.
137 (<expander><mode>3): New.
138 (<expander><mode>3<exec>): Rename to ...
139 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
140 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
142 2020-02-27 Alexandre Oliva <oliva@adacore.com>
144 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
147 2020-02-27 Richard Biener <rguenther@suse.de>
149 PR tree-optimization/93508
150 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
151 non-_CHK variants. Valueize their length arguments.
153 2020-02-27 Richard Biener <rguenther@suse.de>
155 PR tree-optimization/93953
156 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
157 to the hash-map entry.
159 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
161 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
163 2020-02-27 Mark Williams <mwilliams@fb.com>
165 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
166 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
167 -ffile-prefix-map and -fmacro-prefix-map.
168 * lto-streamer-out.c: Include file-prefix-map.h.
169 (lto_output_location): Remap the file part of locations.
171 2020-02-27 Jakub Jelinek <jakub@redhat.com>
174 * gimplify.c (gimplify_init_constructor): Don't promote readonly
175 DECL_REGISTER variables to TREE_STATIC.
177 PR tree-optimization/93582
178 PR tree-optimization/93945
179 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
180 non-zero INTEGER_CST second argument and ref->offset or ref->size
181 not a multiple of BITS_PER_UNIT.
183 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
185 * doc/install.texi (Binaries): Update description of BullFreeware.
187 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
191 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
192 C++ Language Options, Warning Options, and Static Analyzer
193 Options lists. Document negative form of options enabled by
194 default. Move some things around to more accurately sort
195 warnings by category.
196 (C++ Dialect Options, Warning Options, Static Analyzer
197 Options): Document negative form of options when enabled by
198 default. Move some things around to more accurately sort
199 warnings by category. Add some missing index entries.
202 2020-02-26 Carl Love <cel@us.ibm.com>
205 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
206 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
207 for the vector unsigned short arguments. It is also listed as the
208 name of the built-in for arguments vector unsigned short,
209 vector unsigned int and vector unsigned long long built-ins. The
210 name of the builtins for these arguments should be:
211 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
212 __builtin_crypto_vpmsumd respectively.
214 2020-02-26 Richard Biener <rguenther@suse.de>
216 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
217 and load permutation.
219 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
222 * optabs-tree.c (supportable_convert_operation): Reject types with
225 2020-02-26 David Malcolm <dmalcolm@redhat.com>
227 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
229 2020-02-26 Jakub Jelinek <jakub@redhat.com>
231 PR tree-optimization/93820
232 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
233 argument to ALL_INTEGER_CST_P boolean.
234 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
235 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
236 adjacent INTEGER_CST store into merged_store->only_constants like
239 2020-02-25 Jakub Jelinek <jakub@redhat.com>
242 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
244 * cfghooks.c (verify_flow_info): Likewise.
245 * predict.c (combine_predictions_for_bb): Likewise.
246 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
247 sucessor -> successor.
248 (find_traces_1_round): Fix comment typo, destinarion -> destination.
249 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
251 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
252 message typo, sucessors -> successors.
254 2020-02-25 Martin Sebor <msebor@redhat.com>
256 * doc/extend.texi (attribute access): Correct an example.
258 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
260 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
262 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
264 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
265 (VD): Enable for V4BF.
267 (VQ): Enable for V8BF.
270 (VDBL, Vdbl): Add V4BF.
271 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
272 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
273 (bfloat16x8x2_t): Likewise.
274 (bfloat16x4x3_t): Likewise.
275 (bfloat16x8x3_t): Likewise.
276 (bfloat16x4x4_t): Likewise.
277 (bfloat16x8x4_t): Likewise.
278 (vcombine_bf16): New.
279 (vld1_bf16, vld1_bf16_x2): New.
280 (vld1_bf16_x3, vld1_bf16_x4): New.
281 (vld1q_bf16, vld1q_bf16_x2): New.
282 (vld1q_bf16_x3, vld1q_bf16_x4): New.
283 (vld1_lane_bf16): New.
284 (vld1q_lane_bf16): New.
285 (vld1_dup_bf16): New.
286 (vld1q_dup_bf16): New.
289 (vld2_dup_bf16): New.
290 (vld2q_dup_bf16): New.
293 (vld3_dup_bf16): New.
294 (vld3q_dup_bf16): New.
297 (vld4_dup_bf16): New.
298 (vld4q_dup_bf16): New.
299 (vst1_bf16, vst1_bf16_x2): New.
300 (vst1_bf16_x3, vst1_bf16_x4): New.
301 (vst1q_bf16, vst1q_bf16_x2): New.
302 (vst1q_bf16_x3, vst1q_bf16_x4): New.
303 (vst1_lane_bf16): New.
304 (vst1q_lane_bf16): New.
312 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
314 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
315 (VALL_F16): Likewise.
316 (VALLDI_F16): Likewise.
319 (vswap_width_name): Likewise.
320 (VSWAP_WIDTH): Likewise.
324 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
325 (vget_lane_bf16, vgetq_lane_bf16): New.
327 (vdup_n_bf16, vdupq_n_bf16): New.
328 (vdup_lane_bf16, vdup_laneq_bf16): New.
329 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
330 (vduph_lane_bf16, vduph_laneq_bf16): New.
331 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
332 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
333 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
334 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
335 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
336 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
337 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
338 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
339 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
340 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
341 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
342 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
343 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
344 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
345 (vreinterpretq_bf16_p128): New.
346 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
347 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
348 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
349 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
350 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
351 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
352 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
353 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
354 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
355 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
356 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
357 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
358 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
359 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
360 (vreinterpretq_p128_bf16): New.
362 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
364 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
365 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
366 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
367 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
368 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
369 * config/arm/iterators.md (VSF2BF): New attribute.
370 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
371 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
372 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
374 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
376 * config/arm/arm.md (required_for_purecode): New attribute.
377 (enabled): Handle required_for_purecode.
378 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
379 work with -mpure-code.
381 2020-02-25 Jakub Jelinek <jakub@redhat.com>
383 PR rtl-optimization/93908
384 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
387 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
389 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
391 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
393 * doc/install.texi (--enable-checking): Adjust wording.
395 2020-02-25 Richard Biener <rguenther@suse.de>
397 PR tree-optimization/93868
398 * tree-vect-slp.c (slp_copy_subtree): New function.
399 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
400 re-arranging stmts in it.
402 2020-02-25 Jakub Jelinek <jakub@redhat.com>
405 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
406 dummy function and remove it at the end.
409 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
410 paramter -> parameter.
411 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
412 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
414 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
416 * doc/install.texi (--enable-checking): Properly document current
418 (--enable-stage1-checking): Minor clarification about bootstrap.
420 2020-02-24 David Malcolm <dmalcolm@redhat.com>
423 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
424 -fanalyzer-checker=taint is also required.
425 (-fanalyzer-checker=): Note that providing this option enables the
426 given checker, and doing so may be required for checkers that are
429 2020-02-24 David Malcolm <dmalcolm@redhat.com>
431 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
432 significant control flow events; add a "3" which shows all
433 control flow events; the old "3" becomes "4".
435 2020-02-24 Jakub Jelinek <jakub@redhat.com>
437 PR tree-optimization/93582
438 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
439 pd.offset and pd.size to be counted in bits rather than bytes, add
440 support for maxsizei that is not a multiple of BITS_PER_UNIT and
441 handle bitfield stores and loads.
442 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
443 uncomparable quantities - bytes vs. bits. Allow push_partial_def
444 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
445 pd.offset/pd.size to be counted in bits rather than bytes.
446 Formatting fix. Rename shadowed len variable to buflen.
448 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
449 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
452 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
453 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
454 * opts-common.c (parse_options_from_collect_gcc_options): New function.
455 (prepend_xassembler_to_collect_as_options): Likewise.
456 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
457 (prepend_xassembler_to_collect_as_options): Likewise.
458 * lto-opts.c (lto_write_options): Stream assembler options
459 in COLLECT_AS_OPTIONS.
460 * lto-wrapper.c (xassembler_options_error): New static variable.
461 (get_options_from_collect_gcc_options): Move parsing options code to
462 parse_options_from_collect_gcc_options and call it.
463 (merge_and_complain): Validate -Xassembler options.
464 (append_compiler_options): Handle OPT_Xassembler.
465 (run_gcc): Append command line -Xassembler options to
467 * doc/invoke.texi: Add documentation about using Xassembler
470 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
472 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
474 (riscv_rtx_costs): Update cost model for LTGT.
476 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
478 PR rtl-optimization/93564
479 * ira-color.c (struct update_cost_queue_elem): New member start.
480 (queue_update_cost, get_next_update_cost): Add new arg start.
481 (allocnos_conflict_p): New function.
482 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
483 Add checking conflicts with allocnos_conflict_p.
484 (update_costs_from_prefs, restore_costs_from_copies): Adjust
485 update_costs_from_allocno calls.
486 (update_conflict_hard_regno_costs): Add checking conflicts with
487 allocnos_conflict_p. Adjust calls of queue_update_cost and
488 get_next_update_cost.
489 (assign_hard_reg): Adjust calls of queue_update_cost. Add
491 (bucket_allocno_compare_func): Restore previous version.
493 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
495 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
496 double-word size when handling aggregate return values.
497 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
498 that homogeneous SFmode and DFmode aggregates are passed and returned
499 in general registers.
501 2020-02-21 Jakub Jelinek <jakub@redhat.com>
504 * opts.c (print_filtered_help): Translate help before appending
505 messages to it rather than after that.
507 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
509 PR rtl-optimization/PR92989
510 * lra-lives.c (process_bb_lives): Restore the original order
511 of the bb liveness update. Call make_hard_regno_dead for each
512 register clobbered at the start of an EH receiver.
514 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
517 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
518 self-recursively generated.
520 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
523 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
526 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
528 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
529 Document new target supports option.
531 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
533 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
534 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
535 * config/arm/iterators.md (MATMUL): New iterator.
536 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
537 (mmla_sfx): New attribute.
538 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
539 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
540 (UNSPEC_MATMUL_US): New.
542 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
544 * config/arm/arm.md: Prevent scalar shifts from being used when big
547 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
548 Richard Biener <rguenther@suse.de>
550 PR tree-optimization/93586
551 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
552 after mismatched array refs; do not sure type size information to
553 recover from unmatched referneces with !flag_strict_aliasing_p.
555 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
557 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
558 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
559 (scatter_store<mode>): Rename to ...
560 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
561 (scatter<mode>_exec): Delete. Move contents ...
562 (mask_scatter_store<mode>): ... here, and rename that to ...
563 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
564 Remove mode conversion.
565 (mask_gather_load<mode>): Rename to ...
566 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
567 Remove mode conversion.
568 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
570 2020-02-21 Martin Jambor <mjambor@suse.cz>
572 PR tree-optimization/93845
573 * tree-sra.c (verify_sra_access_forest): Only test access size of
576 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
578 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
579 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
580 (addv64di3_exec): Likewise.
581 (subv64di3): Likewise.
582 (subv64di3_exec): Likewise.
583 (addv64di3_zext): Likewise.
584 (addv64di3_zext_exec): Likewise.
585 (addv64di3_zext_dup): Likewise.
586 (addv64di3_zext_dup_exec): Likewise.
587 (addv64di3_zext_dup2): Likewise.
588 (addv64di3_zext_dup2_exec): Likewise.
589 (addv64di3_sext_dup2): Likewise.
590 (addv64di3_sext_dup2_exec): Likewise.
591 (<expander>v64di3): Likewise.
592 (<expander>v64di3_exec): Likewise.
593 (*<reduc_op>_dpp_shr_v64di): Likewise.
594 (*plus_carry_dpp_shr_v64di): Likewise.
595 * config/gcn/gcn.md (adddi3): Likewise.
596 (addptrdi3): Likewise.
597 (<expander>di3): Likewise.
599 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
601 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
603 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
605 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
606 support. Use aarch64_emit_mult instead of emitting multiplication
607 instructions directly.
608 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
609 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
611 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
613 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
614 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
615 instead of emitting multiplication instructions directly.
616 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
617 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
618 (@aarch64_frecps<mode>): New expanders.
620 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
622 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
623 on and produce uint64_ts rather than ints.
624 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
625 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
627 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
629 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
630 an unused xmsk register when handling approximate rsqrt.
632 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
634 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
635 flag_finite_math_only condition.
637 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
640 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
641 to destination operand for shufps alternative.
642 (*vec_extractv2si_1): Ditto.
644 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
647 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
650 2020-02-20 Martin Liska <mliska@suse.cz>
653 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
655 2020-02-20 Martin Liska <mliska@suse.cz>
658 * common/config/avr/avr-common.c: Remote trailing "|".
660 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
662 * collect2.c (maybe_run_lto_and_relink): Fix typo in
665 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
667 PR tree-optimization/93767
668 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
669 access-size bias from the offset calculations for negative strides.
671 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
673 * collect2.c (c_file, o_file): Make const again.
674 (ldout,lderrout, dump_ld_file): Remove.
675 (tool_cleanup): Avoid calling not signal-safe functions.
676 (maybe_run_lto_and_relink): Avoid possible signal handler
677 access to unintialzed memory (lto_o_files).
678 (main): Avoid leaking temp files in $TMPDIR.
679 Initialize c_file/o_file with concat, which avoids exposing
680 uninitialized memory to signal handler, which calls unlink(!).
681 Avoid calling maybe_unlink when the main function returns,
682 since the atexit handler is already doing this.
683 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
685 2020-02-19 Martin Jambor <mjambor@suse.cz>
687 PR tree-optimization/93776
688 * tree-sra.c (create_access): Do not create zero size accesses.
689 (get_access_for_expr): Do not search for zero sized accesses.
691 2020-02-19 Martin Jambor <mjambor@suse.cz>
693 PR tree-optimization/93667
694 * tree-sra.c (scalarizable_type_p): Return false if record fields
695 do not follow wach other.
697 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
699 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
700 rather than fmv.x.s/fmv.s.x.
702 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
704 * config/aarch64/aarch64-simd-builtins.def
705 (intrinsic_vec_smult_lo_): New.
706 (intrinsic_vec_umult_lo_): Likewise.
707 (vec_widen_smult_hi_): Likewise.
708 (vec_widen_umult_hi_): Likewise.
709 * config/aarch64/aarch64-simd.md
710 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
711 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
712 (vmull_high_s16): Likewise.
713 (vmull_high_s32): Likewise.
714 (vmull_high_u8): Likewise.
715 (vmull_high_u16): Likewise.
716 (vmull_high_u32): Likewise.
717 (vmull_s8): Likewise.
718 (vmull_s16): Likewise.
719 (vmull_s32): Likewise.
720 (vmull_u8): Likewise.
721 (vmull_u16): Likewise.
722 (vmull_u32): Likewise.
724 2020-02-18 Martin Liska <mliska@suse.cz>
726 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
727 bootstrap by missing removal of invalid sanity check.
729 2020-02-18 Martin Liska <mliska@suse.cz>
732 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
733 Always compare LHS of gimple_assign.
735 2020-02-18 Martin Liska <mliska@suse.cz>
738 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
739 and return type of functions.
740 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
741 Drop MALLOC attribute for void functions.
742 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
743 malloc_state for a new VOID clone.
745 2020-02-18 Martin Liska <mliska@suse.cz>
748 * common.opt: Add -fprofile-reproducibility.
749 * doc/invoke.texi: Document it.
750 * value-prof.c (dump_histogram_value):
751 Document and support behavior for counters[0]
752 being a negative value.
753 (get_nth_most_common_value): Handle negative
754 counters[0] in respect to flag_profile_reproducible.
756 2020-02-18 Jakub Jelinek <jakub@redhat.com>
759 * cgraph.c (verify_speculative_call): Use speculative_id instead of
760 speculative_uid in messages. Remove trailing whitespace from error
761 message. Use num_speculative_call_targets instead of
762 num_speculative_targets in a message.
763 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
764 edge messages and stmt instead of cal_stmt in reference message.
766 PR tree-optimization/93780
767 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
768 before calling build_vector_type.
769 (execute_update_addresses_taken): Likewise.
772 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
773 typo, functoin -> function.
774 * tree.c (free_lang_data_in_decl): Fix comment typo,
775 functoin -> function.
776 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
778 2020-02-17 David Malcolm <dmalcolm@redhat.com>
780 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
782 (print_option_information): Don't call get_option_url if URLs
785 2020-02-17 Alexandre Oliva <oliva@adacore.com>
787 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
788 handling of register_common-less targets.
790 2020-02-17 Martin Liska <mliska@suse.cz>
793 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
795 2020-02-17 Martin Liska <mliska@suse.cz>
798 * config/rs6000/rs6000.c (rs6000_option_override_internal):
801 2020-02-17 Martin Liska <mliska@suse.cz>
804 * config/rx/elf.opt: Fix typo.
806 2020-02-17 Richard Biener <rguenther@suse.de>
809 * opts-global.c (print_ignored_options): Use inform and
812 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
815 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
817 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
820 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
821 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
823 2020-02-15 Jason Merrill <jason@redhat.com>
825 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
827 2020-02-15 Jakub Jelinek <jakub@redhat.com>
829 PR tree-optimization/93744
830 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
831 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
832 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
833 sure @2 in the first and @1 in the other patterns has no side-effects.
835 2020-02-15 David Malcolm <dmalcolm@redhat.com>
836 Bernd Edlinger <bernd.edlinger@hotmail.de>
840 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
841 * configure.ac (--with-diagnostics-urls): New configuration
842 option, based on --with-diagnostics-color.
843 (DIAGNOSTICS_URLS_DEFAULT): New define.
844 * config.h: Regenerate.
845 * configure: Regenerate.
846 * diagnostic.c (diagnostic_urls_init): Handle -1 for
847 DIAGNOSTICS_URLS_DEFAULT from configure-time
848 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
849 and TERM_URLS environment variable.
850 * diagnostic-url.h (diagnostic_url_format): New enum type.
851 (diagnostic_urls_enabled_p): rename to...
852 (determine_url_format): ... this, and change return type.
853 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
854 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
855 the linux console, and mingw.
856 (diagnostic_urls_enabled_p): rename to...
857 (determine_url_format): ... this, and adjust.
858 * pretty-print.h (pretty_printer::show_urls): rename to...
859 (pretty_printer::url_format): ... this, and change to enum.
860 * pretty-print.c (pretty_printer::pretty_printer,
861 pp_begin_url, pp_end_url, test_urls): Adjust.
862 * doc/install.texi (--with-diagnostics-urls): Document the new
863 configuration option.
864 (--with-diagnostics-color): Document the existing interaction
865 with GCC_COLORS better.
866 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
867 vindex reference. Update description of defaults based on the above.
868 (-fdiagnostics-color): Update description of how -fdiagnostics-color
869 interacts with GCC_COLORS.
871 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
874 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
875 conjunction with TARGET_GNU_TLS in early return.
877 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
879 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
880 the mode is not wider than UNITS_PER_WORD.
882 2020-02-14 Martin Jambor <mjambor@suse.cz>
884 PR tree-optimization/93516
885 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
886 access of the same type as the parent.
887 (propagate_subaccesses_from_lhs): Likewise.
889 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
892 * config/i386/avx512vbmi2intrin.h
893 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
894 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
895 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
896 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
897 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
898 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
899 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
900 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
901 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
902 of lacking a closing parenthesis.
903 * config/i386/avx512vbmi2vlintrin.h
904 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
905 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
906 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
907 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
908 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
909 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
910 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
911 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
912 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
913 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
914 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
915 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
916 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
917 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
918 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
919 _mm_shldi_epi32, _mm_mask_shldi_epi32,
920 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
921 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
923 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
926 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
927 the target function entry.
929 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
931 * common/config/arc/arc-common.c (arc_option_optimization_table):
932 Disable if-conversion step when optimized for size.
934 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
936 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
937 R12-R15 are always in ARCOMPACT16_REGS register class.
938 * config/arc/arc.opt (mq-class): Deprecate.
939 * config/arc/constraint.md ("q"): Remove dependency on mq-class
941 * doc/invoke.texi (mq-class): Update text.
942 * common/config/arc/arc-common.c (arc_option_optimization_table):
945 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
947 * config/arc/arc.c (arc_insn_cost): New function.
948 (TARGET_INSN_COST): Define.
949 * config/arc/arc.md (cost): New attribute.
950 (add_n): Use arc_nonmemory_operand.
951 (ashlsi3_insn): Likewise, also update constraints.
952 (ashrsi3_insn): Likewise.
954 (add_shift): Likewise.
955 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
957 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
959 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
961 (umulsidi_600): Likewise.
963 2020-02-13 Jakub Jelinek <jakub@redhat.com>
966 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
967 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
968 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
969 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
970 pass __A to the builtin followed by __W instead of __A followed by
972 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
973 _mm512_mask_popcnt_epi64): Likewise.
974 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
975 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
976 _mm256_mask_popcnt_epi64): Likewise.
978 PR tree-optimization/93582
979 * fold-const.h (shift_bytes_in_array_left,
980 shift_bytes_in_array_right): Declare.
981 * fold-const.c (shift_bytes_in_array_left,
982 shift_bytes_in_array_right): New function, moved from
983 gimple-ssa-store-merging.c, no longer static.
984 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
985 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
986 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
987 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
988 shift_bytes_in_array.
989 (verify_shift_bytes_in_array): Rename to ...
990 (verify_shift_bytes_in_array_left): ... this. Use
991 shift_bytes_in_array_left instead of shift_bytes_in_array.
992 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
993 instead of verify_shift_bytes_in_array.
994 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
995 / native_interpret_expr where the store covers all needed bits,
996 punt on PDP-endian, otherwise allow all involved offsets and sizes
997 not to be byte-aligned.
1000 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1001 use const_0_to_255_operand predicate instead of immediate_operand.
1002 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1003 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1004 vgf2p8affineinvqb_<mode><mask_name>,
1005 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1006 const_0_to_255_operand predicated operands.
1008 2020-02-12 Jeff Law <law@redhat.com>
1010 * config/h8300/h8300.md (comparison shortening peepholes): Use
1011 a mode iterator to merge the HImode and SImode peepholes.
1013 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1016 * real.c (is_even): Make static. Function comment fix.
1017 (is_halfway_below): Make static, don't assert R is not inf/nan,
1018 instead return false for those. Small formatting fixes.
1020 2020-02-12 Martin Sebor <msebor@redhat.com>
1023 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1024 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1025 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1026 (strlen_check_and_optimize_call): Adjust callee name.
1028 2020-02-12 Jeff Law <law@redhat.com>
1030 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1031 (and (xor)) variant. Combine other two into single peephole.
1033 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1035 PR rtl-optimization/93565
1036 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1038 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1040 * config/aarch64/aarch64-simd.md
1041 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1042 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1043 generating separate ADDV and zero_extend patterns.
1044 * config/aarch64/iterators.md (VDQV_E): New iterator.
1046 2020-02-12 Jeff Law <law@redhat.com>
1048 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1049 expanders, splits, etc.
1050 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1051 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1052 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1053 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1054 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1055 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1057 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1059 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1062 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1063 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1064 TARGET_AVX512DQ from condition.
1065 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1066 instead of <mask_mode512bit_condition> in condition. If
1067 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1069 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1072 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1075 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1077 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1079 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1080 where strlen is more legible.
1081 (rs6000_builtin_vectorized_libmass): Ditto.
1082 (rs6000_print_options_internal): Ditto.
1084 2020-02-11 Martin Sebor <msebor@redhat.com>
1086 PR tree-optimization/93683
1087 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1089 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1091 * config/rs6000/predicates.md (cint34_operand): Rename the
1092 -mprefixed-addr option to be -mprefixed.
1093 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1094 the -mprefixed-addr option to be -mprefixed.
1095 (OTHER_FUTURE_MASKS): Likewise.
1096 (POWERPC_MASKS): Likewise.
1097 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1098 the -mprefixed-addr option to be -mprefixed. Change error
1099 messages to refer to -mprefixed.
1100 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1102 (rs6000_legitimate_offset_address_p): Likewise.
1103 (rs6000_mode_dependent_address): Likewise.
1104 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1105 "-mprefixed" for target attributes and pragmas.
1106 (address_to_insn_form): Rename the -mprefixed-addr option to be
1108 (rs6000_adjust_insn_length): Likewise.
1109 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1110 -mprefixed-addr option to be -mprefixed.
1111 (ASM_OUTPUT_OPCODE): Likewise.
1112 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1113 -mprefixed-addr option to be -mprefixed.
1114 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1115 -mprefixed-addr option to be prefixed. Change the option from
1116 being undocumented to being documented.
1117 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1118 -mprefixed option. Update the -mpcrel documentation to mention
1121 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1123 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1124 including FIRST_PSEUDO_REGISTER - 1.
1125 * ira-color.c (print_hard_reg_set): Ditto.
1127 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1129 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1130 (USTERNOP_QUALIFIERS): New define.
1131 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1132 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1133 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1134 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1135 * config/arm/arm_neon.h (vusdot_s32): New.
1136 (vusdot_lane_s32): New.
1137 (vusdotq_lane_s32): New.
1138 (vsudot_lane_s32): New.
1139 (vsudotq_lane_s32): New.
1140 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1141 * config/arm/iterators.md (DOTPROD_I8MM): New.
1142 (sup, opsuffix): Add <us/su>.
1143 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1144 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1146 2020-02-11 Richard Biener <rguenther@suse.de>
1148 PR tree-optimization/93661
1149 PR tree-optimization/93662
1150 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1152 * tree-sra.c (get_access_for_expr): Likewise.
1154 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1157 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1158 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1159 Change condition from TARGET_AVX2 to TARGET_AVX.
1161 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1164 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1165 argument of strncmp.
1167 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1169 Try to generate zero-based comparisons.
1170 * config/cris/cris.c (cris_reduce_compare): New function.
1171 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1172 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1173 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1175 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1178 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1179 in Thumb state and also as a destination in Arm state. Add T16
1182 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1184 * md.texi (Define Subst): Match closing paren in example.
1186 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1190 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1191 arguments of strncmp.
1193 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1196 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1197 but different source value.
1198 (adjust_callers_for_value_intersection): New function.
1199 (gather_edges_for_value): Adjust order of callers to let a
1200 non-self-recursive caller be the first element.
1201 (self_recursive_pass_through_p): Add a new parameter "simple", and
1202 check generalized self-recursive pass-through jump function.
1203 (self_recursive_agg_pass_through_p): Likewise.
1204 (find_more_scalar_values_for_callers_subset): Compute value from
1205 pass-through jump function for self-recursive.
1206 (intersect_with_plats): Cleanup previous implementation code for value
1207 itersection with self-recursive call edge.
1208 (intersect_with_agg_replacements): Likewise.
1209 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1210 function for self-recursive call edge. Cleanup previous implementation
1211 code for value intersection with self-recursive call edge.
1212 (decide_whether_version_node): Remove dead callers and adjust order
1213 to let a non-self-recursive caller be the first element.
1215 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1217 * recog.c: Move pass_split_before_sched2 code in front of
1218 pass_split_before_regstack.
1219 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1220 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1221 (rest_of_handle_split_before_sched2): Remove.
1222 (pass_split_before_sched2::execute): Unconditionally call
1224 (enable_split_before_sched2): New function.
1225 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1226 (pass_split_before_regstack::gate): Ditto.
1227 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1228 Update name check for renamed split4 pass.
1229 * config/sh/sh.c (register_sh_passes): Update pass insertion
1230 point for renamed split4 pass.
1232 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1234 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1235 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1236 copying them around between host and target.
1238 2020-02-08 Andrew Pinski <apinski@marvell.com>
1241 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1242 STRICT_ALIGNMENT also.
1244 2020-02-08 Jim Wilson <jimw@sifive.com>
1247 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1249 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1250 Jakub Jelinek <jakub@redhat.com>
1253 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1254 xmm16-xmm31 call-used even in 64-bit ms-abi.
1256 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1258 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1259 (simd_ummla, simd_usmmla): Likewise.
1260 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1261 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1262 (vusmmlaq_s32): New.
1264 2020-02-07 Richard Biener <rguenther@suse.de>
1267 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1268 skipping unreachable regions.
1269 (optimize_inline_calls): Skip folding stmts when we didn't
1272 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1275 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1276 Don't return aggregates with only SFmode and DFmode in SSE
1278 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1280 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1283 * config/rs6000/rs6000-logue.c
1284 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1285 if it fails, move rs into end_addr and retry. Add
1286 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1287 the insn pattern doesn't describe well what exactly happens to
1291 * config/i386/predicates.md (avx_identity_operand): Remove.
1292 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1293 (avx_<castmode><avxsizesuffix>_<castmode>,
1294 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1295 a VEC_CONCAT of the operand and UNSPEC_CAST.
1296 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1297 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1301 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1302 recog_data.insn if distance_non_agu_define changed it.
1304 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1307 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1308 we only had X-FORM (reg+reg) addressing for vectors. Also before
1309 ISA 3.0, we only had X-FORM addressing for scalars in the
1310 traditional Altivec registers.
1312 2020-02-06 <zhongyunde@huawei.com>
1313 Vladimir Makarov <vmakarov@redhat.com>
1315 PR rtl-optimization/93561
1316 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1317 hard register range.
1319 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1321 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1324 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1326 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1327 where the low and the high 32 bits are equal to each other specially,
1328 with an rldimi instruction.
1330 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1332 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1334 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1336 * config/arm/arm-tables.opt: Regenerate.
1338 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1341 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1342 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1343 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1345 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1347 PR rtl-optimization/87763
1348 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1350 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1352 * config/aarch64/aarch64-simd-builtins.def
1353 (bfmlaq): New built-in function.
1354 (bfmlalb): New built-in function.
1355 (bfmlalt): New built-in function.
1356 (bfmlalb_lane): New built-in function.
1357 (bfmlalt_lane): New built-in function.
1358 * config/aarch64/aarch64-simd.md
1359 (aarch64_bfmmlaqv4sf): New pattern.
1360 (aarch64_bfmlal<bt>v4sf): New pattern.
1361 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1362 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1363 (vbfmlalbq_f32): New intrinsic.
1364 (vbfmlaltq_f32): New intrinsic.
1365 (vbfmlalbq_lane_f32): New intrinsic.
1366 (vbfmlaltq_lane_f32): New intrinsic.
1367 (vbfmlalbq_laneq_f32): New intrinsic.
1368 (vbfmlaltq_laneq_f32): New intrinsic.
1369 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1370 (bt): New int attribute.
1372 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1374 * config/i386/i386.md (*pushtf): Emit "#" instead of
1375 calling gcc_unreachable in insn output.
1378 (*pushsf_rex64): Ditto for alternatives other than 1.
1379 (*pushsf): Ditto for alternatives other than 1.
1381 2020-02-06 Martin Liska <mliska@suse.cz>
1383 PR gcov-profile/91971
1384 PR gcov-profile/93466
1385 * coverage.c (coverage_init): Revert mangling of
1386 path into filename. It can lead to huge filename length.
1387 Creation of subfolders seem more natural.
1389 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1392 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1393 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1394 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1396 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1399 * config/i386/predicates.md (avx_identity_operand): New predicate.
1400 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1401 define_insn_and_split.
1404 * omp-low.c (use_pointer_for_field): For nested constructs, also
1405 look for map clauses on target construct.
1406 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1407 taskreg_nesting_level.
1410 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1411 shared clause, call omp_notice_variable on outer context if any.
1413 2020-02-05 Jason Merrill <jason@redhat.com>
1416 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1417 non-zero address even if weak and not yet defined.
1419 2020-02-05 Martin Sebor <msebor@redhat.com>
1421 PR tree-optimization/92765
1422 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1423 * tree-ssa-strlen.c (compute_string_length): Remove.
1424 (determine_min_objsize): Remove.
1425 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1426 Avoid using type size as the upper bound on string length.
1427 (handle_builtin_string_cmp): Add an argument. Adjust.
1428 (strlen_check_and_optimize_call): Pass additional argument to
1429 handle_builtin_string_cmp.
1431 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1433 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1434 (*pushdi2_rex64 peephole2): Unconditionally split after
1436 (*ashl<mode>3_doubleword): Ditto.
1437 (*<shift_insn><mode>3_doubleword): Ditto.
1439 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1442 * config/rs6000/rs6000.c (get_vector_offset): Fix
1444 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1446 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1448 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1451 (Special Functions for Debugging the Analyzer): Update description
1452 of __analyzer_dump_exploded_nodes.
1454 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1457 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1458 include sets and not clobbers in the vzeroupper pattern.
1459 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1460 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1461 (*avx_vzeroupper_1): New define_insn_and_split.
1464 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1465 don't run when !optimize.
1466 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1469 2020-02-05 Richard Biener <rguenther@suse.de>
1472 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1473 checks before matching calls.
1475 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1477 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1478 function comment typo.
1481 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1482 simd_clone_create failed when i == 0, adjust clone->nargs by
1485 2020-02-05 Martin Liska <mliska@suse.cz>
1488 * doc/invoke.texi: Document that one should
1489 not combine ASLR and -fpch.
1491 2020-02-04 Richard Biener <rguenther@suse.de>
1493 PR tree-optimization/93538
1494 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1496 2020-02-04 Richard Biener <rguenther@suse.de>
1498 PR tree-optimization/91123
1499 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1500 (vn_walk_cb_data::last_vuse): New member.
1501 (vn_walk_cb_data::saved_operands): Likewsie.
1502 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1503 (vn_walk_cb_data::push_partial_def): Use finish.
1504 (vn_reference_lookup_2): Update last_vuse and use finish if
1505 we've saved operands.
1506 (vn_reference_lookup_3): Use finish and update calls to
1507 push_partial_defs everywhere. When translating through
1508 memcpy or aggregate copies save off operands and alias-set.
1509 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1510 operation for redundant store removal.
1512 2020-02-04 Richard Biener <rguenther@suse.de>
1514 PR tree-optimization/92819
1515 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1516 generating more stmts than before.
1518 2020-02-04 Martin Liska <mliska@suse.cz>
1520 * config/arm/arm.c (arm_gen_far_branch): Move the function
1521 outside of selftests.
1523 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1525 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1526 function to adjust PC-relative vector addresses.
1527 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1528 handle vectors with PC-relative addresses.
1530 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1532 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1534 (hard_reg_and_mode_to_addr_mask): Delete.
1535 (rs6000_adjust_vec_address): If the original vector address
1536 was REG+REG or REG+OFFSET and the element is not zero, do the add
1537 of the elements in the original address before adding the offset
1538 for the vector element. Use address_to_insn_form to validate the
1539 address using the register being loaded, rather than guessing
1540 whether the address is a DS-FORM or DQ-FORM address.
1542 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1544 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1545 to calculate the offset in memory from the start of a vector of a
1546 particular element. Add code to keep the element number in
1547 bounds if the element number is variable.
1548 (rs6000_adjust_vec_address): Move calculation of offset of the
1549 vector element to get_vector_offset.
1550 (rs6000_split_vec_extract_var): Do not do the initial AND of
1551 element here, move the code to get_vector_offset.
1553 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1555 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1558 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1560 * config/rs6000/constraints.md: Improve documentation.
1562 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1565 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1566 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1568 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1570 * config.gcc: Remove "carrizo" support.
1571 * config/gcn/gcn-opts.h (processor_type): Likewise.
1572 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1573 * config/gcn/gcn.opt (gpu_type): Likewise.
1574 * config/gcn/t-omp-device: Likewise.
1576 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1579 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1580 * config/arm/arm.c (arm_gen_far_branch): New function
1582 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1584 2020-02-03 Julian Brown <julian@codesourcery.com>
1585 Tobias Burnus <tobias@codesourcery.com>
1587 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1589 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1592 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1593 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1596 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1598 PR rtl-optimization/91333
1599 * ira-color.c (struct allocno_color_data): Add member
1601 (init_allocno_threads): Set the member up.
1602 (bucket_allocno_compare_func): Add compare hard reg
1605 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1607 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1609 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1610 * config.in: Regenerated.
1611 * configure: Regenerated.
1612 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1613 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1614 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1616 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1618 * configure: Regenerate.
1620 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1622 PR rtl-optimization/91333
1623 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1624 reg preferences comparison up.
1626 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1628 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1629 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1630 aarch64-sve-builtins-base.h.
1631 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1632 aarch64-sve-builtins-base.cc.
1633 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1634 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1636 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1637 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1638 (svcvtnt): New functions.
1639 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1640 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1641 (svcvtnt): New functions.
1642 (svcvt): Add a form that converts f32 to bf16.
1643 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1644 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1646 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1647 Treat B as bfloat16_t.
1648 (ternary_bfloat_lane_base): New class.
1649 (ternary_bfloat_def): Likewise.
1650 (ternary_bfloat): New shape.
1651 (ternary_bfloat_lane_def): New class.
1652 (ternary_bfloat_lane): New shape.
1653 (ternary_bfloat_lanex2_def): New class.
1654 (ternary_bfloat_lanex2): New shape.
1655 (ternary_bfloat_opt_n_def): New class.
1656 (ternary_bfloat_opt_n): New shape.
1657 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1658 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1659 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1660 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1661 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1662 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1663 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1664 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1665 the pattern off the narrow mode instead of the wider one.
1666 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1667 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1668 (sve_fp_op): Handle them.
1669 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1670 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1672 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1674 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1675 * config/aarch64/aarch64-modes.def (BF): Move definition before
1676 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1677 (SVE_MODES): Handle BF modes.
1678 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1680 (aarch64_full_sve_mode): Likewise.
1681 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1683 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1684 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1685 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1686 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1688 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1690 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1691 (TYPES_all_data): Add bf16.
1692 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1693 (register_tuple_type): Increase buffer size.
1694 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1695 (bf16): New type suffix.
1696 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1697 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1698 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1699 Change type from all_data to all_arith.
1700 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1703 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1704 Matthew Malcomson <matthew.malcomson@arm.com>
1705 Richard Sandiford <richard.sandiford@arm.com>
1707 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1708 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1709 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1710 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1711 __ARM_FEATURE_MATMUL_FP64.
1712 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1713 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1714 be disabled at the same time.
1715 (f32mm): New extension.
1716 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1717 (AARCH64_FL_F64MM): Bump to the next bit up.
1718 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1719 (TARGET_SVE_F64MM): New macros.
1720 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1721 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1722 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1723 (UNSPEC_ZIP2Q): New unspeccs.
1724 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1725 (optab, sur, perm_insn): Handle the new unspecs.
1726 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1727 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1728 TARGET_SVE_F64MM instead of separate tests.
1729 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1730 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1731 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1732 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1733 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1734 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1735 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1736 (TYPES_s_signed): New macro.
1737 (TYPES_s_integer): Use it.
1738 (TYPES_d_float): New macro.
1739 (TYPES_d_data): Use it.
1740 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1741 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1742 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1743 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1744 (svmmla): New shape.
1745 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1746 template parameters.
1747 (ternary_resize2_lane_base): Likewise.
1748 (ternary_resize2_base): New class.
1749 (ternary_qq_lane_base): Likewise.
1750 (ternary_intq_uintq_lane_def): Likewise.
1751 (ternary_intq_uintq_lane): New shape.
1752 (ternary_intq_uintq_opt_n_def): New class
1753 (ternary_intq_uintq_opt_n): New shape.
1754 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1755 (ternary_uintq_intq_def): New class.
1756 (ternary_uintq_intq): New shape.
1757 (ternary_uintq_intq_lane_def): New class.
1758 (ternary_uintq_intq_lane): New shape.
1759 (ternary_uintq_intq_opt_n_def): New class.
1760 (ternary_uintq_intq_opt_n): New shape.
1761 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1762 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1763 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1764 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1766 (svdotprod_lane_impl): ...this new class.
1767 (svmmla_impl, svusdot_impl): New classes.
1768 (svdot_lane): Update to use svdotprod_lane_impl.
1769 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1770 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1772 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1773 function, with no types defined.
1774 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1775 AARCH64_FL_I8MM functions.
1776 (svmmla): New AARCH64_FL_F32MM function.
1777 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1778 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1779 AARCH64_FL_F64MM function.
1780 (REQUIRED_EXTENSIONS):
1782 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1784 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1787 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1789 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1790 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1791 (*movti_internal): Do not check for
1792 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1793 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1794 just after check for TARGET_AVX.
1795 (*movdf_internal): Ditto.
1796 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1797 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1798 * config/i386/sse.md (mov<mode>_internal): Only check
1799 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1800 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1801 (<sse>_andnot<mode>3<mask_name>): Move check for
1802 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1803 (<code><mode>3<mask_name>): Ditto.
1804 (*andnot<mode>3): Ditto.
1805 (*andnottf3): Ditto.
1806 (*<code><mode>3): Ditto.
1807 (*<code>tf3): Ditto.
1808 (*andnot<VI:mode>3): Remove
1809 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1810 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1811 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1812 (sse4_1_blendv<ssemodesuffix>): Ditto.
1813 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1814 Explain that tune applies to 128bit instructions only.
1816 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1818 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1819 to definition of hsa_kernel_description. Parse assembly to find SGPR
1820 and VGPR count of kernel and store in hsa_kernel_description.
1822 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1824 PR rtl-optimization/91838
1825 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1826 to truncate if allowed or reject combination.
1828 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1830 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1831 (find_inv_vars_cb): Likewise.
1833 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1835 * calls.c (special_function_p): Split out the check for DECL_NAME
1836 being non-NULL and fndecl being extern at file scope into a
1837 new maybe_special_function_p and call it. Drop check for fndecl
1838 being non-NULL that was after a usage of DECL_NAME (fndecl).
1839 * tree.h (maybe_special_function_p): New inline function.
1841 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1843 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1844 (mask_gather_load<mode>): ... here, and zero-initialize the
1846 (maskload<mode>di): Zero-initialize the destination.
1849 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1852 * doc/analyzer.texi (Limitations): Note that constraints on
1853 floating-point values are currently ignored.
1855 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1858 * symtab.c (symtab_node::noninterposable_alias): If localalias
1859 already exists, but is not usable, append numbers after it until
1860 a unique name is found. Formatting fix.
1863 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1866 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1868 * config/gcn/gcn.c (print_operand): Handle LTGT.
1869 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1871 2020-01-30 Richard Biener <rguenther@suse.de>
1873 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1874 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1876 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1878 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1879 without a DECL in .data.rel.ro.local.
1881 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1884 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1888 * config/i386/sse.md
1889 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1890 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1891 any_extend code iterator instead of always zero_extend.
1892 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1893 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1894 Use any_extend code iterator instead of always zero_extend.
1895 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1896 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1897 Use any_extend code iterator instead of always zero_extend.
1898 (*sse2_pmovmskb_ext): New define_insn.
1899 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1902 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1903 (*popcountsi2_zext_falsedep): New define_insn.
1905 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1907 * config.in: Regenerated.
1908 * configure: Regenerated.
1910 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1913 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1914 LLVM's assembler changed the default in version 9.
1916 2020-01-24 Jeff Law <law@redhat.com>
1918 PR tree-optimization/89689
1919 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1921 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1925 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1927 PR rtl-optimization/87763
1928 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1929 simplification to handle subregs as well as bare regs.
1930 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1932 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1935 * ira.c (ira): Revert use of simplified LRA algorithm.
1937 2020-01-29 Martin Jambor <mjambor@suse.cz>
1939 PR tree-optimization/92706
1940 * tree-sra.c (struct access): Fields first_link, last_link,
1941 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1942 next_rhs_queued and grp_rhs_queued respectively, new fields
1943 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1944 (struct assign_link): Field next renamed to next_rhs, new field
1945 next_lhs. Updated comment.
1946 (work_queue_head): Renamed to rhs_work_queue_head.
1947 (lhs_work_queue_head): New variable.
1948 (add_link_to_lhs): New function.
1949 (relink_to_new_repr): Also relink LHS lists.
1950 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1951 (add_access_to_lhs_work_queue): New function.
1952 (pop_access_from_work_queue): Renamed to
1953 pop_access_from_rhs_work_queue.
1954 (pop_access_from_lhs_work_queue): New function.
1955 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1957 (child_would_conflict_in_lacc): Renamed to
1958 child_would_conflict_in_acc. Adjusted parameter names.
1959 (create_artificial_child_access): New parameter set_grp_read, use it.
1960 (subtree_mark_written_and_enqueue): Renamed to
1961 subtree_mark_written_and_rhs_enqueue.
1962 (propagate_subaccesses_across_link): Renamed to
1963 propagate_subaccesses_from_rhs.
1964 (propagate_subaccesses_from_lhs): New function.
1965 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1968 2020-01-29 Martin Jambor <mjambor@suse.cz>
1970 PR tree-optimization/92706
1971 * tree-sra.c (struct access): Adjust comment of
1972 grp_total_scalarization.
1973 (find_access_in_subtree): Look for single children spanning an entire
1975 (scalarizable_type_p): Allow register accesses, adjust callers.
1976 (completely_scalarize): Remove function.
1977 (scalarize_elem): Likewise.
1978 (create_total_scalarization_access): Likewise.
1979 (sort_and_splice_var_accesses): Do not track total scalarization
1981 (analyze_access_subtree): New parameter totally, adjust to new meaning
1982 of grp_total_scalarization.
1983 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1984 (can_totally_scalarize_forest_p): New function.
1985 (create_total_scalarization_access): Likewise.
1986 (create_total_access_and_reshape): Likewise.
1987 (total_should_skip_creating_access): Likewise.
1988 (totally_scalarize_subtree): Likewise.
1989 (analyze_all_variable_accesses): Perform total scalarization after
1990 subaccess propagation using the new functions above.
1991 (initialize_constant_pool_replacements): Output initializers by
1992 traversing the access tree.
1994 2020-01-29 Martin Jambor <mjambor@suse.cz>
1996 * tree-sra.c (verify_sra_access_forest): New function.
1997 (verify_all_sra_access_forests): Likewise.
1998 (create_artificial_child_access): Set parent.
1999 (analyze_all_variable_accesses): Call the verifier.
2001 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2003 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2004 if called on indirect edge.
2005 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2006 speculative call if needed.
2008 2020-01-29 Richard Biener <rguenther@suse.de>
2010 PR tree-optimization/93428
2011 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2012 permutation when the load node is created.
2013 (vect_analyze_slp_instance): Re-use it here.
2015 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2017 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2019 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2021 PR rtl-optimization/93272
2022 * ira-lives.c (process_out_of_region_eh_regs): New function.
2023 (process_bb_node_lives): Call it.
2025 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2027 * coverage.c (read_counts_file): Make error message lowercase.
2029 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2031 * profile-count.c (profile_quality_display_names): Fix ordering.
2033 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2036 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2037 hash only when edge is first within the sequence.
2038 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2039 (symbol_table::create_edge): Do not set target_prob.
2040 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2042 (cgraph_edge::make_speculative): Drop target_prob parameter.
2043 (cgraph_edge::speculative_call_info): Remove.
2044 (cgraph_edge::first_speculative_call_target): New member function.
2045 (update_call_stmt_hash_for_removing_direct_edge): New function.
2046 (cgraph_edge::resolve_speculation): Rewrite to new API.
2047 (cgraph_edge::speculative_call_for_target): New member function.
2048 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2049 multiple speculation targets.
2050 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2052 (verify_speculative_call): Verify that targets form an interval.
2053 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2054 (cgraph_edge::first_speculative_call_target): New member function.
2055 (cgraph_edge::next_speculative_call_target): New member function.
2056 (cgraph_edge::speculative_call_target_ref): New member function.
2057 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2058 (cgraph_edge): Remove target_prob.
2059 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2060 Fix handling of speculative calls.
2061 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2062 * ipa-fnsummary.c (analyze_function_body): Likewise.
2063 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2064 * ipa-profile.c (dump_histogram): Fix formating.
2065 (ipa_profile_generate_summary): Watch for overflows.
2066 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2067 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2068 (update_indirect_edges_after_inlining): Update to new API.
2069 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2071 * profile-count.h: (profile_probability::adjusted): New.
2072 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2073 updating of profile.
2074 * value-prof.c (gimple_ic_transform): Rename to ...
2075 (dump_ic_profile): ... this one; update dumping.
2076 (stream_in_histogram_value): Fix formating.
2077 (gimple_value_profile_transformations): Update.
2079 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2082 * config/i386/i386.md (*movoi_internal_avx): Remove
2083 TARGET_SSE_TYPELESS_STORES check.
2084 (*movti_internal): Prefer TARGET_AVX over
2085 TARGET_SSE_TYPELESS_STORES.
2086 (*movtf_internal): Likewise.
2087 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2088 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2089 from TARGET_SSE_TYPELESS_STORES.
2091 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2093 * diagnostic-core.h (warning_at): Rename overload to...
2094 (warning_meta): ...this.
2095 (emit_diagnostic_valist): Delete decl of overload taking
2096 diagnostic_metadata.
2097 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2098 (warning_at): Rename overload taking diagnostic_metadata to...
2099 (warning_meta): ...this.
2101 2020-01-28 Richard Biener <rguenther@suse.de>
2103 PR tree-optimization/93439
2104 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2105 * tree-cfg.c (move_sese_region_to_fn): ... here.
2106 (verify_types_in_gimple_reference): Verify used cliques are
2109 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2112 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2113 argument of a pointer to struct gcc_options and pass it to
2114 parse_mtune_ctrl_str.
2115 (ix86_function_specific_restore): Pass opts to
2116 set_ix86_tune_features.
2117 (ix86_option_override_internal): Likewise.
2118 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2119 gcc_options and use it for x_ix86_tune_ctrl_string.
2121 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2123 PR rtl-optimization/87763
2124 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2125 simplification to handle subregs as well as bare regs.
2126 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2128 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2130 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2131 for reduction chains that (now) include a call.
2133 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2135 PR tree-optimization/92822
2136 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2137 out the don't-care elements of a vector whose significant elements
2138 are duplicates, make the don't-care elements duplicates too.
2140 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2142 PR tree-optimization/93434
2143 * tree-predcom.c (split_data_refs_to_components): Record which
2144 components have had aliasing loads removed. Prevent store-store
2145 commoning for all such components.
2147 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2150 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2151 -1 or is_vshift is true, use new_vector with number of elts npatterns
2152 rather than new_unary_operation.
2154 PR tree-optimization/93454
2155 * gimple-fold.c (fold_array_ctor_reference): Perform
2156 elt_size.to_uhwi () just once, instead of calling it in every
2157 iteration. Punt if that value is above size of the temporary
2158 buffer. Decrease third native_encode_expr argument when
2159 bufoff + elt_sz is above size of buf.
2161 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2163 * config/mips/mips.c (mips_declare_object_name)
2164 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2166 2020-01-27 Martin Liska <mliska@suse.cz>
2168 PR gcov-profile/93403
2169 * tree-profile.c (gimple_init_gcov_profiler): Generate
2170 both __gcov_indirect_call_profiler_v4 and
2171 __gcov_indirect_call_profiler_v4_atomic.
2173 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2176 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2178 (@aarch64_split_simd_mov<mode>): Use it.
2179 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2180 Leave the vec_extract patterns to handle 2-element vectors.
2181 (aarch64_simd_mov_from_<mode>high): Likewise.
2182 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2183 (vec_extractv2dfv1df): Likewise.
2185 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2187 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2188 jump conditions for *compare_condjump<GPI:mode>.
2190 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2193 * digraph.cc (test_edge::test_edge): Specify template for base
2196 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2198 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2200 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2202 * config/arc/arc-protos.h (gen_mlo): Remove.
2203 (gen_mhi): Likewise.
2204 * config/arc/arc.c (AUX_MULHI): Define.
2205 (arc_must_save_reister): Special handling for r58/59.
2206 (arc_compute_frame_size): Consider mlo/mhi registers.
2207 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2209 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2210 mlo/mhi name selection.
2211 (arc_restore_callee_saves): Don't early restore blink when ISR.
2212 (arc_expand_prologue): Add mlo/mhi saving.
2213 (arc_expand_epilogue): Add mlo/mhi restoring.
2216 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2217 numbering when MUL64 option is used.
2218 (DWARF2_FRAME_REG_OUT): Define.
2219 * config/arc/arc.md (arc600_stall): New pattern.
2220 (VUNSPEC_ARC_ARC600_STALL): Define.
2221 (mulsi64): Use correct mlo/mhi registers.
2222 (mulsi_600): Clean it up.
2223 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2225 (mhi_operand): Likewise.
2227 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2228 Petro Karashchenko <petro.karashchenko@ring.com>
2230 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2231 attributes if needed.
2232 (prepare_move_operands): Generate special unspec instruction for
2234 (arc_isuncached_mem_p): Propagate uncached attribute to each
2236 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2237 (VUNSPEC_ARC_STDI): Likewise.
2238 (ALLI): New mode iterator.
2239 (mALLI): New mode attribute.
2240 (lddi): New instruction pattern.
2242 (stdidi_split): Split instruction for architectures which are not
2243 supporting ll64 option.
2244 (lddidi_split): Likewise.
2246 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2248 PR rtl-optimization/92989
2249 * lra-lives.c (process_bb_lives): Update the live-in set before
2250 processing additional clobbers.
2252 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2254 PR rtl-optimization/93170
2255 * cselib.c (cselib_invalidate_regno_val): New function, split out
2257 (cselib_invalidate_regno): ...here.
2258 (cselib_invalidated_by_call_p): New function.
2259 (cselib_process_insn): Iterate over all the hard-register entries in
2260 REG_VALUES and invalidate any that cross call-clobbered registers.
2262 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2264 * dojump.c (split_comparison): Use HONOR_NANS rather than
2265 HONOR_SNANS when splitting LTGT.
2267 2020-01-27 Martin Liska <mliska@suse.cz>
2270 * opts.c (print_filtered_help): Exclude language-specific
2271 options from --help=common unless enabled in all FEs.
2273 2020-01-27 Martin Liska <mliska@suse.cz>
2275 * opts.c (print_help): Exclude params from
2276 all except --help=param.
2278 2020-01-27 Martin Liska <mliska@suse.cz>
2281 * config/i386/i386-features.c (make_resolver_func):
2282 Align the code with ppc64 target implementation.
2283 Do not generate a unique name for resolver function.
2285 2020-01-27 Richard Biener <rguenther@suse.de>
2287 PR tree-optimization/93397
2288 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2289 converted reduction chain SLP graph adjustment.
2291 2020-01-26 Marek Polacek <polacek@redhat.com>
2294 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2297 2020-01-26 Jason Merrill <jason@redhat.com>
2300 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2303 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2305 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2306 (rx_setmem): Likewise.
2308 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2311 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2312 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2313 drop <di> from constraint of last operand.
2316 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2317 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2318 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2320 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2323 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2326 2020-01-24 Jeff Law <law@redhat.com>
2328 PR tree-optimization/92788
2329 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2332 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2335 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2336 *avx_vperm_broadcast_<mode>,
2337 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2338 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2339 Move before avx2_perm<mode>/avx512f_perm<mode>.
2342 * simplify-rtx.c (simplify_const_unary_operation,
2343 simplify_const_binary_operation): Punt for mode precision above
2344 MAX_BITSIZE_MODE_ANY_INT.
2346 2020-01-24 Andrew Pinski <apinski@marvell.com>
2348 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2351 2020-01-24 Jeff Law <law@redhat.com>
2354 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2355 for REGs. Call output_operand_lossage to get more reasonable
2358 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2360 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2361 gcn_fp_compare_operator.
2362 (vec_cmpu<mode>di): Use gcn_compare_operator.
2363 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2364 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2365 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2366 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2367 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2368 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2369 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2370 gcn_fp_compare_operator.
2371 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2372 gcn_fp_compare_operator.
2373 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2374 gcn_fp_compare_operator.
2375 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2376 gcn_fp_compare_operator.
2378 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2380 * doc/install.texi (Cross-Compiler-Specific Options): Document
2381 `--with-toolexeclibdir' option.
2383 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2385 * target.def (flags_regnum): Also mention effect on delay slot filling.
2386 * doc/tm.texi: Regenerate.
2388 2020-01-23 Jeff Law <law@redhat.com>
2390 PR translation/90162
2391 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2393 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2396 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2399 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2401 PR rtl-optimization/93402
2402 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2405 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2407 * config.in: Regenerated.
2408 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2409 for TARGET_LIBC_GNUSTACK.
2410 * configure: Regenerated.
2411 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2412 found to be 2.31 or greater.
2414 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2416 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2418 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2419 (mips_asm_file_end): New function. Delegate to
2420 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2421 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2423 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2426 * config/i386/i386-modes.def (POImode): New mode.
2427 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2428 * config/i386/i386.md (DPWI): New mode attribute.
2429 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2431 (QPWI): ... this. Use POI instead of OI for TImode.
2432 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2433 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2436 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2439 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2441 (speculation_tracker_rev): New pattern.
2442 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2443 Use speculation_tracker_rev to track the inverse condition.
2445 2020-01-23 Richard Biener <rguenther@suse.de>
2447 PR tree-optimization/93381
2448 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2449 alias-set of the def as argument and record the first one.
2450 (vn_walk_cb_data::first_set): New member.
2451 (vn_reference_lookup_3): Pass the alias-set of the current def
2452 to push_partial_def. Fix alias-set used in the aggregate copy
2454 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2455 * real.c (clear_significand_below): Fix out-of-bound access.
2457 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2460 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2461 New define_insn patterns.
2463 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2465 * doc/sourcebuild.texi (check-function-bodies): Add an
2466 optional target/xfail selector.
2468 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2470 PR rtl-optimization/93124
2471 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2472 bare USE and CLOBBER insns.
2474 2020-01-22 Andrew Pinski <apinski@marvell.com>
2476 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2478 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2481 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2482 diagnostic_manager into "ana" namespace.
2483 * selftest-run-tests.c (selftest::run_tests): Update for move of
2484 selftest::run_analyzer_selftests to
2485 ana::selftest::run_analyzer_selftests.
2487 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2489 * cfgexpand.c (union_stack_vars): Update the size.
2491 2020-01-22 Richard Biener <rguenther@suse.de>
2493 PR tree-optimization/93381
2494 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2495 throughout, handle all conversions the same.
2497 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2500 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2501 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2502 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2503 Call force_reg on high_in2 unconditionally.
2505 2020-01-22 Martin Liska <mliska@suse.cz>
2507 PR tree-optimization/92924
2508 * profile.c (compute_value_histograms): Divide
2511 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2514 * output.h (assemble_name_resolve): Declare.
2515 * varasm.c (assemble_name_resolve): New function.
2516 (assemble_name): Use it.
2517 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2519 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2521 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2522 update_web_docs_git instead of update_web_docs_svn.
2524 2020-01-21 Andrew Pinski <apinski@marvell.com>
2527 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2528 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2529 (*tlsgd_small_<mode>): Likewise.
2530 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2531 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2532 register. Convert that register back to dest using convert_mode.
2534 2020-01-21 Jim Wilson <jimw@sifive.com>
2536 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2539 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2540 Uros Bizjak <ubizjak@gmail.com>
2543 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2545 (legitimize_tls_address): Do GNU2 TLS address computation in
2546 ptr_mode and zero-extend result to Pmode.
2547 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2548 :P with :PTR and Pmode with ptr_mode.
2549 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2550 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2551 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2553 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2556 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2557 the last two operands are CONST_INT_P before using them as such.
2559 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2561 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2562 to get the integer element types.
2564 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2566 * config/aarch64/aarch64-sve-builtins.h
2567 (function_expander::convert_to_pmode): Declare.
2568 * config/aarch64/aarch64-sve-builtins.cc
2569 (function_expander::convert_to_pmode): New function.
2570 (function_expander::get_contiguous_base): Use it.
2571 (function_expander::prepare_gather_address_operands): Likewise.
2572 * config/aarch64/aarch64-sve-builtins-sve2.cc
2573 (svwhilerw_svwhilewr_impl::expand): Likewise.
2575 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2578 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2579 cfun->machine->label_is_assembled.
2580 (aarch64_print_patchable_function_entry): New.
2581 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2582 * config/aarch64/aarch64.h (struct machine_function): New field,
2585 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2588 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2591 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2594 * cgraph.c (cgraph_edge::resolve_speculation,
2595 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2596 call_stmt_site_hash.
2598 2020-01-21 Martin Liska <mliska@suse.cz>
2600 * config/rs6000/rs6000.c (common_mode_defined): Remove
2603 2020-01-21 Richard Biener <rguenther@suse.de>
2605 PR tree-optimization/92328
2606 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2607 type when value-numbering same-sized store by inserting a
2609 (eliminate_dom_walker::eliminate_stmt): When eliminating
2610 a redundant store handle bit-reinterpretation of the same value.
2612 2020-01-21 Andrew Pinski <apinski@marvel.com>
2615 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2617 (prepare_block_for_update): This. Use a worklist instead of
2620 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2622 * gcc/config/arm/arm.c (clear_operation_p):
2623 Initialise last_regno, skip first iteration
2624 based on the first_set value and use ints instead
2625 of the unnecessary HOST_WIDE_INTs.
2627 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2630 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2631 compare_mode other than SFmode or DFmode.
2633 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2636 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2637 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2638 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2640 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2642 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2644 2020-01-20 Andrew Pinski <apinski@marvell.com>
2647 * targhooks.c (default_print_patchable_function_entry): Use
2648 output_asm_insn to emit the nop instruction.
2650 2020-01-20 Fangrui Song <maskray@google.com>
2653 * targhooks.c (default_print_patchable_function_entry): Align to
2656 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2659 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2660 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2661 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2662 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2663 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2664 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2665 Remove the {q} suffix from lea.
2666 (*tls_dynamic_gnu2_call_64): Renamed to ...
2667 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2668 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2669 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2670 Pass Pmode to gen_tls_dynamic_gnu2_64.
2672 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2674 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2676 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2678 * config/aarch64/aarch64-sve-builtins-base.cc
2679 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2681 2020-01-20 Richard Biener <rguenther@suse.de>
2684 * dwarf2out.c (prune_unused_types): Unconditionally mark
2685 called function DIEs.
2687 2020-01-20 Martin Liska <mliska@suse.cz>
2689 PR tree-optimization/93199
2690 * tree-eh.c (struct leh_state): Add
2691 new field outer_non_cleanup.
2692 (cleanup_is_dead_in): Pass leh_state instead
2693 of eh_region. Add a checking that state->outer_non_cleanup
2694 points to outer non-clean up region.
2695 (lower_try_finally): Record outer_non_cleanup
2697 (lower_catch): Likewise.
2698 (lower_eh_filter): Likewise.
2699 (lower_eh_must_not_throw): Likewise.
2700 (lower_cleanup): Likewise.
2702 2020-01-20 Richard Biener <rguenther@suse.de>
2704 PR tree-optimization/93094
2705 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2706 (vect_transform_loop): Likewise.
2707 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2708 loop_vectorized_call to vect_transform_loop.
2709 * tree-vect-loop.c (vect_transform_loop): Pass down
2710 loop_vectorized_call to vect_loop_versioning.
2711 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2712 the earlier discovered loop_vectorized_call.
2714 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2716 * doc/contribute.texi: Update for SVN -> Git transition.
2717 * doc/install.texi: Likewise.
2719 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2722 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2723 speculative targets.
2724 (verify_speculative_call): New function
2725 (cgraph_node::verify_node): Use it.
2726 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2729 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2732 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2733 (cgraph_edge::make_direct): Remove all indirect targets.
2734 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2735 (cgraph_node::verify_node): Verify that only one call_stmt or
2736 lto_stmt_uid is set.
2737 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2739 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2740 (lto_output_ref): Simplify streaming of stmt.
2741 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2743 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2745 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2746 Mark parameter unused.
2748 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2750 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2752 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2754 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2756 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2758 * Makefile.in: Add coroutine-passes.o.
2759 * builtin-types.def (BT_CONST_SIZE): New.
2760 (BT_FN_BOOL_PTR): New.
2761 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2762 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2763 * coroutine-builtins.def: New file.
2764 * coroutine-passes.cc: New file.
2765 * function.h (struct GTY function): Add a bit to indicate that the
2766 function is a coroutine component.
2767 * internal-fn.c (expand_CO_FRAME): New.
2768 (expand_CO_YIELD): New.
2769 (expand_CO_SUSPN): New.
2770 (expand_CO_ACTOR): New.
2771 * internal-fn.def (CO_ACTOR): New.
2775 * passes.def: Add pass_coroutine_lower_builtins,
2776 pass_coroutine_early_expand_ifns.
2777 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2778 (make_pass_coroutine_early_expand_ifns): New.
2779 * doc/invoke.texi: Document the fcoroutines command line
2782 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2784 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2787 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2788 after checking the argument is a REG. Don't use REGNO (reg)
2789 again to set last_regno, reuse regno variable instead.
2791 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2793 * doc/analyzer.texi (Limitations): Add note about NaN.
2795 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2796 Sudakshina Das <sudi.das@arm.com>
2798 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2799 and valid immediate.
2800 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2801 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2802 * config/arm/constraints.md (Pg): New.
2803 * config/arm/predicates.md (long_shift_imm): New.
2804 (arm_reg_or_long_shift_imm): Likewise.
2805 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2806 (thumb2_lsll): Likewise.
2809 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2810 Sudakshina Das <sudi.das@arm.com>
2812 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2813 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2814 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2815 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2816 * config/arm/thumb2.md (thumb2_asrl): New.
2817 (thumb2_lsll): Likewise.
2819 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2821 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2824 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2826 * gdbinit.in (help-gcc-hooks): New command.
2827 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2828 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2831 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2833 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2834 correct target macro.
2836 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2838 * config/aarch64/aarch64-protos.h
2839 (aarch64_sve_ld1ro_operand_p): New.
2840 * config/aarch64/aarch64-sve-builtins-base.cc
2841 (class load_replicate): New.
2842 (class svld1ro_impl): New.
2843 (class svld1rq_impl): Change to inherit from load_replicate.
2844 (svld1ro): New sve intrinsic function base.
2845 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2846 New DEF_SVE_FUNCTION.
2847 * config/aarch64/aarch64-sve-builtins-base.h
2848 (svld1ro): New decl.
2849 * config/aarch64/aarch64-sve-builtins.cc
2850 (function_expander::add_mem_operand): Modify assert to allow
2852 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2854 * config/aarch64/aarch64.c
2855 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2856 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2857 (aarch64_sve_ld1ro_operand_p): New.
2858 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2859 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2860 * config/aarch64/predicates.md
2861 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2863 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2865 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2866 Introduce this ACLE specified predefined macro.
2867 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2868 (fp): Disabling this disables f64mm.
2869 (simd): Disabling this disables f64mm.
2870 (fp16): Disabling this disables f64mm.
2871 (sve): Disabling this disables f64mm.
2872 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2873 (AARCH64_ISA_F64MM): New.
2874 (TARGET_F64MM): New.
2875 * doc/invoke.texi (f64mm): Document new option.
2877 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2879 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2880 (neoversen1_tunings): Likewise.
2882 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2885 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2886 Add assert to ensure prolog has been emitted.
2887 (aarch64_split_atomic_op): Likewise.
2888 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2889 Use epilogue_completed rather than reload_completed.
2890 (aarch64_atomic_exchange<mode>): Likewise.
2891 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2892 (atomic_nand<mode>): Likewise.
2893 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2894 (atomic_fetch_nand<mode>): Likewise.
2895 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2896 (atomic_nand_fetch<mode>): Likewise.
2898 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2901 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2903 (REVERSE_CONDITION): Delete.
2904 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2905 (CCFP_CCFPE): Likewise.
2906 (e): New mode attribute.
2907 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2908 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2909 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2910 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2911 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2912 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2913 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2914 name of generator from gen_ccmpdi to gen_ccmpccdi.
2915 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2916 the previous comparison but aren't able to, use the new ccmp_rev
2919 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2921 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2922 than testing directly for INTEGER_CST.
2923 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2925 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2927 PR tree-optimization/93292
2928 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2929 get_vectype_for_scalar_type returns NULL.
2931 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2933 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2934 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2936 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2938 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2940 * params.opt: (max-predicted-iterations): Set bounds.
2941 * predict.c (real_almost_one, real_br_prob_base,
2942 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2943 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2944 probabilities; do not truncate to reg_br_prob_bases.
2945 (estimate_loops_at_level): Pass max_cyclic_prob.
2946 (estimate_loops): Compute max_cyclic_prob.
2947 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2949 * profile-count.c (profile_probability::to_sreal): New.
2950 * profile-count.h (class sreal): Move up in file.
2951 (profile_probability::to_sreal): Declare.
2953 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2956 (arm_invalid_conversion): New function for target hook.
2957 (arm_invalid_unary_op): New function for target hook.
2958 (arm_invalid_binary_op): New function for target hook.
2960 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2962 * config.gcc: Add arm_bf16.h.
2963 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2964 (arm_simd_builtin_std_type): Add BFmode.
2965 (arm_init_simd_builtin_types): Define element types for vector types.
2966 (arm_init_bf16_types): New function.
2967 (arm_init_builtins): Add arm_init_bf16_types function call.
2968 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2969 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2970 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2971 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2972 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2973 (arm_mangle_type): Add __bf16.
2974 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2975 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2976 arm_bf16_ptr_type_node.
2977 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2978 define_split between ARM registers.
2979 * config/arm/arm_bf16.h: New file.
2980 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2981 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2983 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2984 * config/arm/vfp.md: Add BFmode to movhf patterns.
2986 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2987 Andre Vieira <andre.simoesdiasvieira@arm.com>
2989 * config/arm/arm-cpus.in (mve, mve_float): New features.
2990 (dsp, mve, mve.fp): New options.
2991 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2992 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2993 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2995 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2996 Thomas Preud'homme <thomas.preudhomme@arm.com>
2998 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3000 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3001 error for using -mcmse when targeting Armv8.1-M Mainline.
3003 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3004 Thomas Preud'homme <thomas.preudhomme@arm.com>
3006 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3007 address in r4 when targeting Armv8.1-M Mainline.
3008 (nonsecure_call_value_internal): Likewise.
3009 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3010 a register match_operand again. Emit BLXNS when targeting
3012 (nonsecure_call_value_reg_thumb2): Likewise.
3014 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3015 Thomas Preud'homme <thomas.preudhomme@arm.com>
3017 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3018 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3019 variable as true when floating-point ABI is not hard. Replace
3020 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3021 Generate VLSTM and VLLDM instruction respectively before and
3022 after a function call to cmse_nonsecure_call function.
3023 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3024 (VUNSPEC_VLLDM): Likewise.
3025 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3026 (lazy_load_multiple_insn): Likewise.
3028 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3029 Thomas Preud'homme <thomas.preudhomme@arm.com>
3031 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3032 (arm_emit_vfp_multi_reg_pop): Likewise.
3033 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3034 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3035 restore callee-saved VFP registers.
3037 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3038 Thomas Preud'homme <thomas.preudhomme@arm.com>
3040 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3041 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3042 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3043 callee-saved GPRs as well as clear ip register before doing a nonsecure
3044 call then restore callee-saved GPRs after it when targeting
3046 (arm_reorg): Adapt to function rename.
3048 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3049 Thomas Preud'homme <thomas.preudhomme@arm.com>
3051 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3052 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3053 clear_vfp_multiple pattern based on a new vfp parameter.
3054 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3055 targeting Armv8.1-M Mainline.
3056 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3057 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3058 whether VFP registers are available before looking call_used_regs for a
3060 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3061 of prototype of clear_operation_p.
3062 (clear_vfp_multiple_operation): New predicate.
3063 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3064 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3066 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3067 Thomas Preud'homme <thomas.preudhomme@arm.com>
3069 * config/arm/arm-protos.h (clear_operation_p): Declare.
3070 * config/arm/arm.c (clear_operation_p): New function.
3071 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3072 targeting Armv8.1-M Mainline or successor.
3073 (output_return_instruction): Only output APSR register clearing if
3074 Armv8.1-M Mainline instructions not available.
3075 (thumb_exit): Likewise.
3076 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3077 * config/arm/thumb2.md (clear_apsr): New define_insn.
3078 (clear_multiple): Likewise.
3079 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3081 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3082 Thomas Preud'homme <thomas.preudhomme@arm.com>
3084 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3085 (use_return_insn): Also return false for Armv8.1-M Mainline.
3086 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3087 Mainline instructions are available.
3088 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3089 when targeting Armv8.1-M Mainline Security Extensions.
3090 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3091 Mainline entry function.
3092 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3093 targeting Armv8.1-M Mainline or successor.
3094 (arm_expand_epilogue): Fix indentation of caller-saved register
3095 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3097 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3098 (FP_SYSREGS): Likewise.
3099 (enum vfp_sysregs_encoding): Define enum.
3100 (fp_sysreg_names): Declare.
3101 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3102 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3103 (pop_fpsysreg_insn): Likewise.
3105 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3106 Thomas Preud'homme <thomas.preudhomme@arm.com>
3108 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3109 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3110 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3111 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3112 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3113 (ARMv8_1m_main): New feature group.
3114 (armv8.1-m.main): New architecture.
3115 * config/arm/arm-tables.opt: Regenerate.
3116 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3117 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3118 (arm_options_perform_arch_sanity_checks): Error out when targeting
3119 Armv8.1-M Mainline Security Extensions.
3120 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3122 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3124 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3125 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3126 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3127 aarch64_bfdot_laneq): New.
3128 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3129 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3130 vbfdotq_laneq_f32): New.
3131 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3132 VBFMLA_W, VBF): New.
3133 (isquadop): Add V4BF, V8BF.
3135 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3137 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3138 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3139 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3140 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3141 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3142 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3143 usdot_laneq, sudot_lane,sudot_laneq): New.
3144 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3145 (aarch64_<sur>dot_lane): New.
3146 * config/aarch64/arm_neon.h (vusdot_s32): New.
3148 (vusdot_lane_s32): New.
3149 (vsudot_lane_s32): New.
3150 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3151 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3153 2020-01-16 Martin Liska <mliska@suse.cz>
3155 * value-prof.c (dump_histogram_value): Fix
3156 obvious spacing issue.
3158 2020-01-16 Andrew Pinski <apinski@marvell.com>
3160 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3161 !storage_order_barrier_p.
3163 2020-01-16 Andrew Pinski <apinski@marvell.com>
3165 * sched-int.h (_dep): Add unused bit-field field for the padding.
3166 * sched-deps.c (init_dep_1): Init unused field.
3168 2020-01-16 Andrew Pinski <apinski@marvell.com>
3170 * optabs.h (create_expand_operand): Initialize target field also.
3172 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3174 PR tree-optimization/92429
3175 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3176 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3178 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3181 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3183 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3184 aarch64_sve_int_mode to each mode.
3186 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3188 * doc/analyzer.texi (Overview): Add note about
3189 -fdump-ipa-analyzer.
3191 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3193 PR tree-optimization/93231
3194 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3195 input_type is unsigned. Use tree_to_shwi for shift constant.
3196 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3197 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3198 inputs more efficiently.
3200 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3202 * config/i386/i386.md (*movsf_internal): Do not require
3203 SSE2 ISA for alternatives 14 and 15.
3205 2020-01-15 Richard Biener <rguenther@suse.de>
3208 * tree-eh.c (sink_clobbers): If we already visited the destination
3209 block do not defer insertion.
3210 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3211 the purpose of defered insertion.
3213 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3215 * BASE-VER: Bump to 10.0.1.
3217 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3219 PR tree-optimization/93247
3220 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3221 type of the stmt that we're going to vectorize.
3223 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3225 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3226 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3229 2020-01-15 Martin Liska <mliska@suse.cz>
3231 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3232 2 calls of streamer_read_hwi in a function call.
3234 2020-01-15 Richard Biener <rguenther@suse.de>
3236 * alias.c (record_alias_subset): Avoid redundant work when
3237 subset is already recorded.
3239 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3241 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3242 the analyzer options provide CWE identifiers.
3244 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3246 * tree-diagnostic-path.cc (path_summary::event_range::print):
3247 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3248 using get_pure_location.
3250 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3252 PR tree-optimization/93262
3253 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3254 perform head trimming only if the last argument is constant,
3255 either all ones, or larger or equal to head trim, in the latter
3256 case decrease the last argument by head_trim.
3258 PR tree-optimization/93249
3259 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3260 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3261 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3262 perform head trim unless we can prove there are no '\0' chars
3263 from the source among the first head_trim chars.
3265 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3267 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3269 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3272 * config/i386/sse.md
3273 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3274 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3275 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3276 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3277 just a single alternative instead of two, make operands 1 and 2
3280 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3283 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3286 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3288 * Makefile.in (lang_opt_files): Add analyzer.opt.
3289 (ANALYZER_OBJS): New.
3290 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3291 tristate.o and ANALYZER_OBJS.
3292 (TEXI_GCCINT_FILES): Add analyzer.texi.
3293 * common.opt (-fanalyzer): New driver option.
3294 * config.in: Regenerate.
3295 * configure: Regenerate.
3296 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3297 (gccdepdir): Also create depdir for "analyzer" subdir.
3298 * digraph.cc: New file.
3299 * digraph.h: New file.
3300 * doc/analyzer.texi: New file.
3301 * doc/gccint.texi ("Static Analyzer") New menu item.
3302 (analyzer.texi): Include it.
3303 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3304 ("Warning Options"): Add static analysis warnings to the list.
3305 (-Wno-analyzer-double-fclose): New option.
3306 (-Wno-analyzer-double-free): New option.
3307 (-Wno-analyzer-exposure-through-output-file): New option.
3308 (-Wno-analyzer-file-leak): New option.
3309 (-Wno-analyzer-free-of-non-heap): New option.
3310 (-Wno-analyzer-malloc-leak): New option.
3311 (-Wno-analyzer-possible-null-argument): New option.
3312 (-Wno-analyzer-possible-null-dereference): New option.
3313 (-Wno-analyzer-null-argument): New option.
3314 (-Wno-analyzer-null-dereference): New option.
3315 (-Wno-analyzer-stale-setjmp-buffer): New option.
3316 (-Wno-analyzer-tainted-array-index): New option.
3317 (-Wno-analyzer-use-after-free): New option.
3318 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3319 (-Wno-analyzer-use-of-uninitialized-value): New option.
3320 (-Wanalyzer-too-complex): New option.
3321 (-fanalyzer-call-summaries): New warning.
3322 (-fanalyzer-checker=): New warning.
3323 (-fanalyzer-fine-grained): New warning.
3324 (-fno-analyzer-state-merge): New warning.
3325 (-fno-analyzer-state-purge): New warning.
3326 (-fanalyzer-transitivity): New warning.
3327 (-fanalyzer-verbose-edges): New warning.
3328 (-fanalyzer-verbose-state-changes): New warning.
3329 (-fanalyzer-verbosity=): New warning.
3330 (-fdump-analyzer): New warning.
3331 (-fdump-analyzer-callgraph): New warning.
3332 (-fdump-analyzer-exploded-graph): New warning.
3333 (-fdump-analyzer-exploded-nodes): New warning.
3334 (-fdump-analyzer-exploded-nodes-2): New warning.
3335 (-fdump-analyzer-exploded-nodes-3): New warning.
3336 (-fdump-analyzer-supergraph): New warning.
3337 * doc/sourcebuild.texi (dg-require-dot): New.
3338 (dg-check-dot): New.
3339 * gdbinit.in (break-on-saved-diagnostic): New command.
3340 * graphviz.cc: New file.
3341 * graphviz.h: New file.
3342 * ordered-hash-map-tests.cc: New file.
3343 * ordered-hash-map.h: New file.
3344 * passes.def (pass_analyzer): Add before
3345 pass_ipa_whole_program_visibility.
3346 * selftest-run-tests.c (selftest::run_tests): Call
3347 selftest::ordered_hash_map_tests_cc_tests.
3348 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3350 * shortest-paths.h: New file.
3351 * timevar.def (TV_ANALYZER): New timevar.
3352 (TV_ANALYZER_SUPERGRAPH): Likewise.
3353 (TV_ANALYZER_STATE_PURGE): Likewise.
3354 (TV_ANALYZER_PLAN): Likewise.
3355 (TV_ANALYZER_SCC): Likewise.
3356 (TV_ANALYZER_WORKLIST): Likewise.
3357 (TV_ANALYZER_DUMP): Likewise.
3358 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3359 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3360 * tree-pass.h (make_pass_analyzer): New decl.
3361 * tristate.cc: New file.
3362 * tristate.h: New file.
3364 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3367 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3368 alternatives 9 and 10.
3370 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3372 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3373 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3374 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3375 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3376 (selftest::hash_map_tests_c_tests): Call it.
3377 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3378 New static constant, using the value of = H::empty_zero_p.
3379 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3380 from default_hash_traits <Value>.
3381 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3383 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3384 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3385 calls to mark_empty with !Descriptor::empty_zero_p.
3386 (hash_table::empty_slow): Conditionalize the memset call with a
3387 check that Descriptor::empty_zero_p; otherwise, loop through the
3388 entries calling mark_empty on them.
3389 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3390 (pointer_hash::empty_zero_p): Likewise.
3391 (pair_hash::empty_zero_p): Likewise.
3392 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3394 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3395 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3396 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3397 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3398 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3399 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3400 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3401 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3403 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3406 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3408 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3409 fix typo on return value.
3411 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3414 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3416 (cgraph_edge::make_speculative): Add param for setting speculative_id
3418 (cgraph_edge::speculative_call_info): Update comments and find reference
3419 by speculative_id for multiple indirect targets.
3420 (cgraph_edge::resolve_speculation): Decrease the speculations
3421 for indirect edge, drop it's speculative if not direct target
3422 left. Update comments.
3423 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3424 (cgraph_node::dump): Print num_speculative_call_targets.
3425 (cgraph_node::verify_node): Don't report error if speculative
3426 edge not include statement.
3427 (cgraph_edge::num_speculative_call_targets_p): New function.
3428 * cgraph.h (int common_target_id): Remove.
3429 (int common_target_probability): Remove.
3430 (num_speculative_call_targets): New variable.
3431 (make_speculative): Add param for setting speculative_id.
3432 (cgraph_edge::num_speculative_call_targets_p): New declare.
3433 (target_prob): New variable.
3434 (speculative_id): New variable.
3435 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3436 call summaries for multiple speculative call targets.
3437 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3438 * ipa-profile.c (struct speculative_call_target): New struct.
3439 (class speculative_call_summary): New class.
3440 (class speculative_call_summaries): New class.
3441 (call_sums): New variable.
3442 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3443 (ipa_profile_write_edge_summary): New function.
3444 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3445 (ipa_profile_dump_all_summaries): New function.
3446 (ipa_profile_read_edge_summary): New function.
3447 (ipa_profile_read_summary_section): New function.
3448 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3449 (ipa_profile): Generate num_speculative_call_targets from
3451 * ipa-ref.h (speculative_id): New variable.
3452 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3453 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3454 common_target_probability. Stream out speculative_id and
3455 num_speculative_call_targets.
3456 (input_edge): Likewise.
3457 * predict.c (dump_prediction): Remove edges count assert to be
3459 * symtab.c (symtab_node::create_reference): Init speculative_id.
3460 (symtab_node::clone_references): Clone speculative_id.
3461 (symtab_node::clone_referring): Clone speculative_id.
3462 (symtab_node::clone_reference): Clone speculative_id.
3463 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3464 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3465 if indirect call contains multiple speculative targets.
3466 * value-prof.h (check_ic_target): Remove.
3467 * value-prof.c (gimple_value_profile_transformations):
3468 Use void function gimple_ic_transform.
3469 * value-prof.c (gimple_ic_transform): Handle topn case.
3470 Fix comment typos. Change it to a void function.
3472 2020-01-13 Andrew Pinski <apinski@marvell.com>
3474 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3475 (octeontx2t98): New define.
3476 (octeontx2t96): New define.
3477 (octeontx2t93): New define.
3478 (octeontx2f95): New define.
3479 (octeontx2f95n): New define.
3480 (octeontx2f95mm): New define.
3481 * config/aarch64/aarch64-tune.md: Regenerate.
3482 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3484 2020-01-13 Jason Merrill <jason@redhat.com>
3486 PR c++/33799 - destroy return value if local cleanup throws.
3487 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3489 2020-01-13 Martin Liska <mliska@suse.cz>
3491 * ipa-cp.c (get_max_overall_size): Use newly
3492 renamed param param_ipa_cp_unit_growth.
3493 * params.opt: Remove legacy param name.
3495 2020-01-13 Martin Sebor <msebor@redhat.com>
3497 PR tree-optimization/93213
3498 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3499 stores to be eliminated.
3501 2020-01-13 Martin Liska <mliska@suse.cz>
3503 * opts.c (print_help): Do not print CL_PARAM
3504 and CL_WARNING for CL_OPTIMIZATION.
3506 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3509 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3510 depending on optimization settings.
3512 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3514 PR tree-optimization/90838
3515 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3516 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3517 argument rather than to initialize temporary for targets that
3518 don't use the mode argument at all. Initialize ctzval to avoid
3521 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3523 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3524 * tree-core.h: Document it.
3525 * gimplify.c (gimplify_omp_workshare): Set it.
3526 * omp-low.c (lower_omp_target): Use it.
3527 * tree-pretty-print.c (dump_omp_clause): Print it.
3529 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3530 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3532 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3534 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3535 * common.opt (fdiagnostics-path-format=): New option.
3536 (diagnostic_path_format): New enum.
3537 (fdiagnostics-show-path-depths): New option.
3538 * coretypes.h (diagnostic_event_id_t): New forward decl.
3539 * diagnostic-color.c (color_dict): Add "path".
3540 * diagnostic-event-id.h: New file.
3541 * diagnostic-format-json.cc (json_from_expanded_location): Make
3543 (json_end_diagnostic): Call context->make_json_for_path if it
3544 exists and the diagnostic has a path.
3545 (diagnostic_output_format_init): Clear context->print_path.
3546 * diagnostic-path.h: New file.
3547 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3548 when printing a run of events in a diagnostic_path so that they
3549 all get the same color.
3550 (layout::m_diagnostic_path_p): New field.
3551 (layout::layout): Initialize it.
3552 (layout::print_any_labels): Don't colorize the label text for an
3553 event in a diagnostic_path.
3554 (gcc_rich_location::add_location_if_nearby): Add
3555 "restrict_to_current_line_spans" and "label" params. Pass the
3556 former to layout.maybe_add_location_range; pass the latter
3557 when calling add_range.
3558 * diagnostic.c: Include "diagnostic-path.h".
3559 (diagnostic_initialize): Initialize context->path_format and
3560 context->show_path_depths.
3561 (diagnostic_show_any_path): New function.
3562 (diagnostic_path::interprocedural_p): New function.
3563 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3564 (simple_diagnostic_path::num_events): New function.
3565 (simple_diagnostic_path::get_event): New function.
3566 (simple_diagnostic_path::add_event): New function.
3567 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3568 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3569 (debug): New overload taking a diagnostic_path *.
3570 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3571 * diagnostic.h (enum diagnostic_path_format): New enum.
3572 (json::value): New forward decl.
3573 (diagnostic_context::path_format): New field.
3574 (diagnostic_context::show_path_depths): New field.
3575 (diagnostic_context::print_path): New callback field.
3576 (diagnostic_context::make_json_for_path): New callback field.
3577 (diagnostic_show_any_path): New decl.
3578 (json_from_expanded_location): New decl.
3579 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3580 (-fdiagnostics-show-path-depths): New option.
3581 (-fdiagnostics-color): Add "path" to description of default
3582 GCC_COLORS; describe it.
3583 (-fdiagnostics-format=json): Document how diagnostic paths are
3584 represented in the JSON output format.
3585 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3586 Add optional params "restrict_to_current_line_spans" and "label".
3587 * opts.c (common_handle_option): Handle
3588 OPT_fdiagnostics_path_format_ and
3589 OPT_fdiagnostics_show_path_depths.
3590 * pretty-print.c: Include "diagnostic-event-id.h".
3591 (pp_format): Implement "%@" format code for printing
3592 diagnostic_event_id_t *.
3593 (selftest::test_pp_format): Add tests for "%@".
3594 * selftest-run-tests.c (selftest::run_tests): Call
3595 selftest::tree_diagnostic_path_cc_tests.
3596 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3597 * toplev.c (general_init): Initialize global_dc->path_format and
3598 global_dc->show_path_depths.
3599 * tree-diagnostic-path.cc: New file.
3600 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3601 non-static. Drop "diagnostic" param in favor of storing the
3602 original value of "where" and re-using it.
3603 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3604 maybe_unwind_expanded_macro_loc.
3605 (tree_diagnostics_defaults): Initialize context->print_path and
3606 context->make_json_for_path.
3607 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3609 (default_tree_make_json_for_path): New decl.
3610 (maybe_unwind_expanded_macro_loc): New decl.
3612 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3614 PR tree-optimization/93210
3615 * fold-const.h (native_encode_initializer,
3616 can_native_interpret_type_p): Declare.
3617 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3619 (native_encode_initializer): New function, moved from dwarf2out.c.
3620 Adjust to native_encode_expr compatible arguments, including dry-run
3621 and partial extraction modes. Don't handle STRING_CST.
3622 (can_native_interpret_type_p): No longer static.
3623 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3624 offset / BITS_PER_UNIT fits into int and don't call it if
3625 can_native_interpret_type_p fails. If suboff is NULL and for
3626 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3627 native_encode_initializer.
3628 (fold_const_aggregate_ref_1): Formatting fix.
3629 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3630 (tree_add_const_value_attribute): Adjust caller.
3632 PR tree-optimization/90838
3633 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3634 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3635 CTZ_DEFINED_VALUE_AT_ZERO.
3637 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3640 * lra-constraints.c (match_reload): Permit input operands have the
3641 same mode as output while other input operands have a different
3644 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3646 PR tree-optimization/90838
3647 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3648 (check_ctz_string): Likewise.
3649 (optimize_count_trailing_zeroes): Likewise.
3650 (simplify_count_trailing_zeroes): Likewise.
3651 (pass_forwprop::execute): Try ctz simplification.
3652 * match.pd: Add matching for ctz idioms.
3654 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3656 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3658 (aarch64_invalid_unary_op): New function for target hook.
3659 (aarch64_invalid_binary_op): New function for target hook.
3661 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3663 * config.gcc: Add arm_bf16.h.
3664 * config/aarch64/aarch64-builtins.c
3665 (aarch64_simd_builtin_std_type): Add BFmode.
3666 (aarch64_init_simd_builtin_types): Define element types for vector
3668 (aarch64_init_bf16_types): New function.
3669 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3670 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3672 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3673 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3675 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3676 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3677 * config/aarch64/aarch64.c
3678 (aarch64_classify_vector_mode): Add support for BF types.
3679 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3680 (aarch64_vq_mode): Add support for BF types.
3681 (aarch64_simd_container_mode): Add support for BF types.
3682 (aarch64_mangle_type): Add support for BF scalar type.
3683 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3684 * config/aarch64/arm_bf16.h: New file.
3685 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3686 * config/aarch64/iterators.md: Add BF types to mode attributes.
3687 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3689 2020-01-10 Jason Merrill <jason@redhat.com>
3691 PR c++/93173 - incorrect tree sharing.
3692 * gimplify.c (copy_if_shared): No longer static.
3693 * gimplify.h: Declare it.
3695 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3697 * doc/invoke.texi (-msve-vector-bits=): Document that
3698 -msve-vector-bits=128 now generates VL-specific code for
3699 little-endian targets.
3700 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3701 build_vector_type_for_mode to construct the data vector types.
3702 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3703 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3704 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3705 for 128-bit vectors.
3707 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3709 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3712 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3714 * config/aarch64/aarch64-builtins.c
3715 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3716 rather than checking the number of elements and the element mode.
3718 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3720 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3721 get_related_vectype_for_scalar_type rather than build_vector_type
3722 to create the index type for a conditional reduction.
3724 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3726 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3727 for any type of gather or scatter, including strided accesses.
3729 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3731 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3734 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3736 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3738 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3739 parameter and its use to reset DR_OFFSET's.
3740 (vect_transform_loop): Remove orig_drs_init argument.
3741 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3742 member of dr_vec_info rather than the offset of the associated
3743 data_reference's innermost_loop_behavior.
3744 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3745 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3746 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3747 get_dr_vinfo_offset.
3748 (vectorizable_store): Likewise.
3749 (vectorizable_load): Likewise.
3751 2020-01-10 Richard Biener <rguenther@suse.de>
3753 * gimple-ssa-store-merging
3754 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3756 2020-01-10 Martin Liska <mliska@suse.cz>
3759 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3760 encapsulation that was there before r280040.
3762 2020-01-10 Richard Biener <rguenther@suse.de>
3765 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3766 sequences to avoid walking them again for secondary opportunities.
3767 (pass_lower_eh_dispatch::execute): Instead actually insert
3770 2020-01-10 Richard Biener <rguenther@suse.de>
3773 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3774 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3775 avoid quadraticness.
3777 2020-01-10 Martin Jambor <mjambor@suse.cz>
3779 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3780 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3781 to get param_ipa_sra_max_replacements.
3782 (param_splitting_across_edge): Pass the caller to
3783 pull_accesses_from_callee.
3785 2020-01-10 Martin Jambor <mjambor@suse.cz>
3787 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3788 * ipa-cp.c (max_new_size): Removed.
3789 (orig_overall_size): New variable.
3790 (get_max_overall_size): New function.
3791 (estimate_local_effects): Use it. Adjust dump.
3792 (decide_about_value): Likewise.
3793 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3794 orig_overall_size. Adjust dump.
3795 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3797 2020-01-10 Martin Jambor <mjambor@suse.cz>
3799 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3800 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3801 instead of param_ipa_max_agg_items.
3802 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3803 optimization info for the callee.
3805 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3807 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3808 markers if debug_inline_points is false.
3810 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3812 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3814 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3815 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3816 aarch64-sve-builtins-sve2.h.
3817 (aarch64-sve-builtins-sve2.o): New rule.
3818 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3819 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3820 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3821 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3822 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3823 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3825 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3826 instructions that are handled here.
3827 (@cond_asrd<mode>): Generalize to...
3828 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3829 (*cond_asrd<mode>_2): Generalize to...
3830 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3831 (*cond_asrd<mode>_z): Generalize to...
3832 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3833 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3834 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3835 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3836 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3838 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3839 (@aarch64_scatter_stnt<mode>): Likewise.
3840 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3841 (@aarch64_mul_lane_<mode>): Likewise.
3842 (@aarch64_sve_suqadd<mode>_const): Likewise.
3843 (*<sur>h<addsub><mode>): Generalize to...
3844 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3846 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3847 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3848 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3849 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3850 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3851 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3852 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3853 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3854 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3855 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3856 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3857 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3858 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3859 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3860 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3861 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3862 (@aarch64_sve2_xar<mode>): Likewise.
3863 (@aarch64_sve2_bcax<mode>): Likewise.
3864 (*aarch64_sve2_eor3<mode>): Rename to...
3865 (@aarch64_sve2_eor3<mode>): ...this.
3866 (@aarch64_sve2_bsl<mode>): New expander.
3867 (@aarch64_sve2_nbsl<mode>): Likewise.
3868 (@aarch64_sve2_bsl1n<mode>): Likewise.
3869 (@aarch64_sve2_bsl2n<mode>): Likewise.
3870 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3871 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3872 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3873 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3874 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3875 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3876 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3877 (<su>mull<bt><Vwide>): Generalize to...
3878 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3880 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3881 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3882 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3883 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3884 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3885 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3886 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3887 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3888 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3889 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3890 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3891 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3892 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3893 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3894 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3895 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3896 (<SHRNB:r>shrnb<mode>): Generalize to...
3897 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3899 (<SHRNT:r>shrnt<mode>): Generalize to...
3900 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3902 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3903 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3904 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3905 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3906 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3907 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3908 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3909 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3910 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3911 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3912 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3913 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3914 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3915 (@aarch64_sve2_cvtnt<mode>): Likewise.
3916 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3917 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3918 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3919 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3920 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3921 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3922 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3923 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3924 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3925 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3926 (@aarch64_sve2_pmul<mode>): Likewise.
3927 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3928 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3929 (@aarch64_sve2_tbl2<mode>): Likewise.
3930 (@aarch64_sve2_tbx<mode>): Likewise.
3931 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3932 (@aarch64_sve2_histcnt<mode>): Likewise.
3933 (@aarch64_sve2_histseg<mode>): Likewise.
3934 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3935 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3936 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3937 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3938 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3939 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3940 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3941 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3942 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3943 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3944 (SVE2_PMULL_PAIR_I): New mode iterators.
3945 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3946 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3947 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3948 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3949 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3950 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3951 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3952 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3953 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3954 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3955 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3956 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3957 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3958 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3959 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3960 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3961 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3962 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3963 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3964 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3965 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3966 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3967 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3968 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3969 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3970 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3971 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3972 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3973 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3974 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3975 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3977 (VNARROW, Ventype): New mode attributes.
3978 (Vewtype): Handle VNx2DI. Fix typo in comment.
3979 (VDOUBLE): New mode attribute.
3980 (sve_lane_con): Handle VNx8HI.
3981 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3982 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3983 (sve_int_op, sve_int_op_rev): Handle the above codes.
3984 (sve_pred_int_rhs2_operand): Likewise.
3985 (MULLBT, SHRNB, SHRNT): Delete.
3986 (SVE_INT_SHIFT_IMM): New int iterator.
3987 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3988 and UNSPEC_WHILEHS for TARGET_SVE2.
3989 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3990 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3991 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3992 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3993 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3994 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3995 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3996 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3997 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3998 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3999 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4000 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4001 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4002 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4003 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4004 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4005 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4006 (optab): Handle the new unspecs.
4007 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4009 (lr): Handle the new unspecs.
4011 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4012 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4013 (sve_int_qsub_op): New int attributes.
4014 (sve_fp_op, rot): Handle the new unspecs.
4015 * config/aarch64/aarch64-sve-builtins.h
4016 (function_resolver::require_matching_pointer_type): Declare.
4017 (function_resolver::resolve_unary): Add an optional boolean argument.
4018 (function_resolver::finish_opt_n_resolution): Add an optional
4019 type_suffix_index argument.
4020 (gimple_folder::redirect_call): Declare.
4021 (gimple_expander::prepare_gather_address_operands): Add an optional
4023 * config/aarch64/aarch64-sve-builtins.cc: Include
4024 aarch64-sve-builtins-sve2.h.
4025 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4026 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4027 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4028 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4029 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4030 (TYPES_s_unsigned): Likewise.
4031 (TYPES_s_integer): Use TYPES_s_unsigned.
4032 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4033 (TYPES_sd_integer): Use them.
4034 (TYPES_d_unsigned): New macro.
4035 (TYPES_d_integer): Use it.
4036 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4037 (TYPES_cvt_narrow): Likewise.
4038 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4039 (preds_mx): New variable.
4040 (function_builder::add_overloaded_function): Allow the new feature
4041 set to be more restrictive than the original one.
4042 (function_resolver::infer_pointer_type): Remove qualifiers from
4043 the pointer type before printing it.
4044 (function_resolver::require_matching_pointer_type): New function.
4045 (function_resolver::resolve_sv_displacement): Handle functions
4046 that don't support 32-bit vector indices or svint32_t vector offsets.
4047 (function_resolver::finish_opt_n_resolution): Take the inferred type
4048 as a separate argument.
4049 (function_resolver::resolve_unary): Optionally treat all forms in
4050 the same way as normal merging functions.
4051 (gimple_folder::redirect_call): New function.
4052 (function_expander::prepare_gather_address_operands): Add an argument
4053 that says whether scaled forms are available. If they aren't,
4054 handle scaling of vector indices and don't add the extension and
4056 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4057 fall back to using cond_* instead.
4058 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4059 Split out the member variables into...
4060 (rtx_code_function_base): ...this new base class.
4061 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4062 (unspec_based_function): Split out the member variables into...
4063 (unspec_based_function_base): ...this new base class.
4064 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4065 (unspec_based_function_exact_insn): New class.
4066 (unspec_based_add_function, unspec_based_add_lane_function)
4067 (unspec_based_lane_function, unspec_based_pred_function)
4068 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4069 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4070 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4072 (unspec_based_fused_function): New class.
4073 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4074 (unspec_based_fused_lane_function): New class.
4075 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4077 (CODE_FOR_MODE1): New macro.
4078 (fixed_insn_function): New class.
4079 (while_comparison): Likewise.
4080 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4081 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4082 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4083 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4084 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4085 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4086 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4087 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4088 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4089 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4090 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4091 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4092 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4093 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4094 Also add an initial argument for unary_convert_narrowt, regardless
4095 of the predication type.
4096 (build_32_64): Allow loads and stores to specify MODE_none.
4097 (build_sv_index64, build_sv_uint_offset): New functions.
4098 (long_type_suffix): New function.
4099 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4100 (binary_imm_long_base, load_gather_sv_base): Likewise.
4101 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4102 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4103 (unary_narrowb_base, unary_narrowt_base): Likewise.
4104 (binary_long_lane_def, binary_long_lane): New shape.
4105 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4106 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4107 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4108 (binary_to_uint_def, binary_to_uint): Likewise.
4109 (binary_wide_def, binary_wide): Likewise.
4110 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4111 (compare_def, compare): Likewise.
4112 (compare_ptr_def, compare_ptr): Likewise.
4113 (load_ext_gather_index_restricted_def,
4114 load_ext_gather_index_restricted): Likewise.
4115 (load_ext_gather_offset_restricted_def,
4116 load_ext_gather_offset_restricted): Likewise.
4117 (load_gather_sv_def): Inherit from load_gather_sv_base.
4118 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4119 (shift_left_imm_def, shift_left_imm): Likewise.
4120 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4121 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4122 (store_scatter_index_restricted_def,
4123 store_scatter_index_restricted): Likewise.
4124 (store_scatter_offset_restricted_def,
4125 store_scatter_offset_restricted): Likewise.
4126 (tbl_tuple_def, tbl_tuple): Likewise.
4127 (ternary_long_lane_def, ternary_long_lane): Likewise.
4128 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4129 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4130 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4131 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4132 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4133 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4134 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4135 (ternary_uint_def, ternary_uint): Likewise.
4136 (unary_convert): Fix typo in comment.
4137 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4138 (unary_long_def, unary_long): Likewise.
4139 (unary_narrowb_def, unary_narrowb): Likewise.
4140 (unary_narrowt_def, unary_narrowt): Likewise.
4141 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4142 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4143 (unary_to_int_def, unary_to_int): Likewise.
4144 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4145 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4146 (svasrd_impl): Delete.
4147 (svcadd_impl::expand): Handle integer operations too.
4148 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4149 new functions to derive the unspec numbers.
4150 (svmla_svmls_lane_impl): Replace with...
4151 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4152 integer operations too.
4153 (svwhile_impl): Rename to...
4154 (svwhilelx_impl): ...this and inherit from while_comparison.
4155 (svasrd): Use unspec_based_function.
4156 (svmla_lane): Use svmla_lane_impl.
4157 (svmls_lane): Use svmls_lane_impl.
4158 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4159 (svwhilele, svwhilelt): Use svwhilelx_impl.
4160 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4161 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4162 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4163 * config/aarch64/aarch64-sve-builtins.def: Include
4164 aarch64-sve-builtins-sve2.def.
4166 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4168 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4169 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4170 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4171 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4172 immediates as well as vector ones.
4173 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4174 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4175 (aarch64_sve_qsub_immediate): Update calls accordingly.
4177 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4179 * config/aarch64/aarch64-sve2.md: Add banner comments.
4180 (<su>mulh<r>s<mode>3): Move further up file.
4181 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4182 (*aarch64_sve2_sra<mode>): Move further down file.
4183 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4185 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4187 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4189 (while_optab_cmp): Handle them.
4190 * config/aarch64/aarch64-sve.md
4191 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4192 and add a "@" marker.
4193 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4194 instead of gen_aarch64_sve2_while_ptest.
4195 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4197 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4199 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4200 (UNSPEC_WHILELE): ...this.
4201 (UNSPEC_WHILE_LO): Rename to...
4202 (UNSPEC_WHILELO): ...this.
4203 (UNSPEC_WHILE_LS): Rename to...
4204 (UNSPEC_WHILELS): ...this.
4205 (UNSPEC_WHILE_LT): Rename to...
4206 (UNSPEC_WHILELT): ...this.
4207 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4208 (cmp_op, while_optab_cmp): Likewise.
4209 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4210 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4211 (svwhilelt): Likewise.
4213 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4215 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4216 (unary_to_uint): Define.
4217 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4218 (unary_count): Rename to...
4219 (unary_to_uint_def, unary_to_uint): ...this.
4220 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4222 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4224 * config/aarch64/aarch64-sve-builtins-functions.h
4225 (code_for_mode_function): New class.
4226 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4227 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4228 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4229 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4230 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4232 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4234 * config/aarch64/iterators.md (addsub): New code attribute.
4235 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4237 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4238 in the asm string and attributes. Fix indentation.
4239 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4241 (@aarch64_sve_<optab><mode>): ...this.
4242 * config/aarch64/aarch64-sve-builtins.h
4243 (function_expander::expand_signed_unpred_op): Delete.
4244 * config/aarch64/aarch64-sve-builtins.cc
4245 (function_expander::expand_signed_unpred_op): Likewise.
4246 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4247 try using code_for_aarch64_sve instead.
4248 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4249 (svqsub_impl): Likewise.
4250 (svqadd, svqsub): Use rtx_code_function instead.
4252 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4254 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4255 (HADDSUB, sur, addsub): Remove them.
4257 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4259 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4260 functions the same way, rather than singling out those that
4261 aren't mapped directly to optabs.
4263 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4265 * target.def (compatible_vector_types_p): New target hook.
4266 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4267 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4268 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4269 * doc/tm.texi: Regenerate.
4270 * gimple-expr.c: Include target.h.
4271 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4272 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4274 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4275 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4276 Use the original predicate if it already has a suitable type.
4278 2020-01-09 Martin Jambor <mjambor@suse.cz>
4280 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4281 resolve_speculation and redirect_call_stmt_to_callee static. Change
4282 return type of set_call_stmt to cgraph_edge *.
4283 * auto-profile.c (afdo_indirect_call): Adjust call to
4284 redirect_call_stmt_to_callee.
4285 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4286 make the this pointer explicit, adjust self-recursive calls and the
4287 call top make_direct. Return the resulting edge.
4288 (cgraph_edge::remove): Make this pointer explicit.
4289 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4290 (cgraph_edge::make_direct): Likewise, adjust call to
4291 resolve_speculation.
4292 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4293 call to set_call_stmt.
4294 (cgraph_update_edges_for_call_stmt_node): Update call to
4295 set_call_stmt and remove.
4296 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4297 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4298 (cgraph_node::create_edge_including_clones): Moved "first" definition
4299 of edge to the block where it was used. Adjusted calls to
4301 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4302 cgraph_edge::remove.
4303 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4304 make_direct and redirect_call_stmt_to_callee.
4305 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4306 resolve_speculation and make_direct.
4307 * ipa-inline-transform.c (inline_transform): Adjust call to
4308 redirect_call_stmt_to_callee.
4309 (check_speculations_1):: Adjust call to resolve_speculation.
4310 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4311 resolve-speculation.
4312 (inline_small_functions): Adjust call to resolve_speculation.
4313 (ipa_inline): Likewise.
4314 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4316 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4317 safe with regards to edge removal, adjust calls to
4318 redirect_call_stmt_to_callee.
4319 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4320 and redirect_call_stmt_to_callee.
4321 * multiple_target.c (create_dispatcher_calls): Adjust call to
4322 redirect_call_stmt_to_callee
4323 (redirect_to_specific_clone): Likewise.
4324 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4325 Adjust calls to cgraph_edge::remove.
4326 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4327 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4328 (expand_call_inline): Adjust call to cgraph_edge::remove.
4330 2020-01-09 Martin Liska <mliska@suse.cz>
4332 * params.opt: Set Optimization for
4333 param_max_speculative_devirt_maydefs.
4335 2020-01-09 Martin Sebor <msebor@redhat.com>
4339 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4341 2020-01-09 Martin Liska <mliska@suse.cz>
4343 * auto-profile.c (auto_profile): Use opt_for_fn
4345 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4346 (propagate_vals_across_arith_jfunc): Likewise.
4347 (hint_time_bonus): Likewise.
4348 (incorporate_penalties): Likewise.
4349 (good_cloning_opportunity_p): Likewise.
4350 (perform_estimation_of_a_value): Likewise.
4351 (estimate_local_effects): Likewise.
4352 (ipcp_propagate_stage): Likewise.
4353 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4354 (set_switch_stmt_execution_predicate): Likewise.
4355 (analyze_function_body): Likewise.
4356 * ipa-inline-analysis.c (offline_size): Likewise.
4357 * ipa-inline.c (early_inliner): Likewise.
4358 * ipa-prop.c (ipa_analyze_node): Likewise.
4359 (ipcp_transform_function): Likewise.
4360 * ipa-sra.c (process_scan_results): Likewise.
4361 (ipa_sra_summarize_function): Likewise.
4362 * params.opt: Rename ipcp-unit-growth to
4363 ipa-cp-unit-growth. Add Optimization for various
4364 IPA-related parameters.
4366 2020-01-09 Richard Biener <rguenther@suse.de>
4369 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4371 2020-01-09 Richard Biener <rguenther@suse.de>
4373 PR tree-optimization/93040
4374 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4376 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4378 * common/config/avr/avr-common.c (avr_option_optimization_table)
4379 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4381 2020-01-09 Martin Liska <mliska@suse.cz>
4383 * cgraphclones.c (symbol_table::materialize_all_clones):
4384 Use cgraph_node::dump_name.
4386 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4389 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4390 output_operand_lossage instead of gcc_unreachable.
4391 * doc/md.texi (riscv f constraint): Fix typo.
4394 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4395 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4396 CONST_SCALAR_INT_P instead of CONST_INT_P.
4397 (*subv<mode>4_1): Rename to ...
4398 (subv<mode>4_1): ... this.
4399 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4400 define_insn_and_split patterns.
4401 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4404 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4406 * vec.c (class selftest::count_dtor): New class.
4407 (selftest::test_auto_delete_vec): New test.
4408 (selftest::vec_c_tests): Call it.
4409 * vec.h (class auto_delete_vec): New class template.
4410 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4412 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4414 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4416 2020-01-08 Jim Wilson <jimw@sifive.com>
4418 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4419 use of TLS_MODEL_LOCAL_EXEC when not pic.
4421 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4423 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4426 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4429 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4430 *stack_protect_set_3 peephole2): Also check that the second
4431 insns source is general_operand.
4434 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4435 predicate for output operand instead of register_operand.
4436 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4437 memory destination and non-memory operands[2].
4439 2020-01-08 Martin Liska <mliska@suse.cz>
4441 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4442 ::dump_asm_name instead of (::name or ::asm_name).
4443 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4444 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4445 (analyze_functions): Likewise.
4446 (expand_all_functions): Likewise.
4447 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4448 (propagate_bits_across_jump_function): Likewise.
4449 (dump_profile_updates): Likewise.
4450 (ipcp_store_bits_results): Likewise.
4451 (ipcp_store_vr_results): Likewise.
4452 * ipa-devirt.c (dump_targets): Likewise.
4453 * ipa-fnsummary.c (analyze_function_body): Likewise.
4454 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4455 (process_hsa_functions): Likewise.
4456 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4457 (set_alias_uids): Likewise.
4458 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4459 * ipa-inline.c (recursive_inlining): Likewise.
4460 (inline_to_all_callers_1): Likewise.
4461 (ipa_inline): Likewise.
4462 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4463 (ipa_propagate_frequency): Likewise.
4464 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4465 (remove_described_reference): Likewise.
4466 * ipa-pure-const.c (worse_state): Likewise.
4467 (check_retval_uses): Likewise.
4468 (analyze_function): Likewise.
4469 (propagate_pure_const): Likewise.
4470 (propagate_nothrow): Likewise.
4471 (dump_malloc_lattice): Likewise.
4472 (propagate_malloc): Likewise.
4473 (pass_local_pure_const::execute): Likewise.
4474 * ipa-visibility.c (optimize_weakref): Likewise.
4475 (function_and_variable_visibility): Likewise.
4476 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4477 (ipa_discover_variable_flags): Likewise.
4478 * lto-streamer-out.c (output_function): Likewise.
4479 (output_constructor): Likewise.
4480 * tree-inline.c (copy_bb): Likewise.
4481 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4482 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4484 2020-01-08 Richard Biener <rguenther@suse.de>
4487 * tree-eh.c (sink_clobbers): Update virtual operands for
4488 the first and last stmt only. Add a dry-run capability.
4489 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4490 after CFG manipulations and in RPO order to catch all
4491 secondary opportunities reliably.
4493 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4496 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4498 2019-01-08 Richard Biener <rguenther@suse.de>
4501 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4502 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4503 virtual operand, also updating SSA use.
4504 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4505 Update stmt after resetting virtual operand.
4506 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4507 * gimple-iterator.c (gsi_remove): When not removing the stmt
4508 permanently do not delink immediate uses or mark the stmt modified.
4510 2020-01-08 Martin Liska <mliska@suse.cz>
4512 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4513 (ipa_call_context::estimate_size_and_time): Likewise.
4514 (inline_analyze_function): Likewise.
4516 2020-01-08 Martin Liska <mliska@suse.cz>
4518 * cgraph.c (cgraph_node::dump): Use systematically
4521 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4523 Add -nodevicespecs option for avr.
4526 * config/avr/avr.opt (-nodevicespecs): New driver option.
4527 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4528 "-specs=device-specs/..." if that option is not set.
4529 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4531 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4533 Implement 64-bit double functions for avr.
4536 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4537 --with-double-comparison.
4538 * doc/install.texi: Document them.
4539 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4540 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4541 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4542 * doc/invoke.texi (AVR Built-in Macros): Document them.
4543 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4544 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4545 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4547 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4550 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4551 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4552 when only building rm-profile multilibs.
4554 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4557 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4558 lattice for a value to check.
4559 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4560 finite propagation in self-recursive scc.
4562 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4564 * ipa-inline.c (caller_growth_limits): Restore the AND.
4566 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4568 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4569 (VEC_ALLREG_ALT): New iterator.
4570 (VEC_ALLREG_INT_MODE): New iterator.
4571 (VCMP_MODE): New iterator.
4572 (VCMP_MODE_INT): New iterator.
4573 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4574 (vec_cmp<u>v64qidi): New define_expand.
4575 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4576 (vec_cmpu<mode>di_exec): New define_expand.
4577 (vec_cmp<u>v64qidi_exec): New define_expand.
4578 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4579 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4580 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4581 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4582 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4583 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4584 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4585 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4586 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4587 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4589 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4590 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4592 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4594 * config/gcn/constraints.md (DA): Update description and match.
4596 (Db): New constraint.
4597 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4599 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4600 Implement 'Db' mixed immediate type.
4601 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4602 (addcv64si3_dup<exec_vcc>): Delete.
4603 (subcv64si3<exec_vcc>): Rework constraints.
4604 (addv64di3): Rework constraints.
4605 (addv64di3_exec): Rework constraints.
4606 (subv64di3): Rework constraints.
4607 (addv64di3_dup): Delete.
4608 (addv64di3_dup_exec): Delete.
4609 (addv64di3_zext): Rework constraints.
4610 (addv64di3_zext_exec): Rework constraints.
4611 (addv64di3_zext_dup): Rework constraints.
4612 (addv64di3_zext_dup_exec): Rework constraints.
4613 (addv64di3_zext_dup2): Rework constraints.
4614 (addv64di3_zext_dup2_exec): Rework constraints.
4615 (addv64di3_sext_dup2): Rework constraints.
4616 (addv64di3_sext_dup2_exec): Rework constraints.
4618 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4620 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4621 existing target checks.
4623 2020-01-07 Richard Biener <rguenther@suse.de>
4625 * doc/install.texi: Bump minimal supported MPC version.
4627 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4629 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4630 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4631 * langhooks.c: Include stor-layout.h.
4632 (lhd_simulate_enum_decl): New function.
4633 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4634 handle_arm_sve_h for the LTO frontend.
4635 (register_vector_type): Cope with null returns from pushdecl.
4637 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4639 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4640 (aarch64_sve::nvectors_if_data_type): Replace with...
4641 (aarch64_sve::builtin_type_p): ...this.
4642 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4643 (find_vector_type): Delete.
4644 (add_sve_type_attribute): New function.
4645 (lookup_sve_type_attribute): Likewise.
4646 (register_builtin_types): Add an "SVE type" attribute to each type.
4647 (register_tuple_type): Likewise.
4648 (svbool_type_p, nvectors_if_data_type): Delete.
4649 (mangle_builtin_type): Use lookup_sve_type_attribute.
4650 (builtin_type_p): Likewise. Add an overload that returns the
4651 number of constituent vector and predicate registers.
4652 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4653 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4654 instead of aarch64_sve_argument_p.
4655 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4656 (aarch64_pass_by_reference): Likewise.
4657 (aarch64_function_value_1): Likewise.
4658 (aarch64_return_in_memory): Likewise.
4659 (aarch64_layout_arg): Likewise.
4661 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4663 PR tree-optimization/93156
4664 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4665 least significant bit is always clear.
4667 PR tree-optimization/93118
4668 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4669 simplifier with two intermediate conversions.
4671 2020-01-07 Martin Liska <mliska@suse.cz>
4673 * params.opt: Add Optimization for various parameters.
4675 2020-01-07 Martin Liska <mliska@suse.cz>
4678 * doc/extend.texi: Explain cloning for target_clone
4681 2020-01-07 Martin Liska <mliska@suse.cz>
4683 PR tree-optimization/92860
4684 * common.opt: Make in Optimization option
4685 as it is affected by -O0, which is an Optimization
4687 * tree-inline.c (tree_inlinable_function_p):
4688 Use opt_for_fn for warn_inline.
4689 (expand_call_inline): Likewise.
4691 2020-01-07 Martin Liska <mliska@suse.cz>
4693 PR tree-optimization/92860
4694 * common.opt: Make flag_ree as optimization
4697 2020-01-07 Martin Liska <mliska@suse.cz>
4699 PR optimization/92860
4700 * params.opt: Mark param_min_crossjump_insns with Optimization
4703 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4705 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4706 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4708 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4710 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4711 helper function to return the valid addressing formats for a given
4712 hard register and mode.
4713 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4715 * config/rs6000/constraints.md (Q constraint): Update
4717 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4720 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4721 Use 'Q' for doing vector extract from memory.
4722 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4724 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4725 doing vector extract from memory.
4726 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4727 extract from memory.
4729 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4730 for the offset being 34-bits when -mcpu=future is used.
4732 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4734 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4735 instead of cmpib_comparison_operator in cmpib patterns.
4736 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4737 of cmpib_comparison_operator. Revise comment.
4739 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4741 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4742 in an IFN_DIV_POW2 node to be equal.
4744 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4746 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4747 (vect_check_scalar_mask): ...this.
4748 (vectorizable_store, vectorizable_load): Update call accordingly.
4749 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4750 argument in calls to conditional internal functions.
4752 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4754 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4755 '0' matching inputs.
4756 (subv64di3_exec): Likewise.
4758 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4760 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4761 * doc/md.texi (movstr): Likewise.
4763 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4765 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4768 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4770 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4772 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4773 to a temporary file and use move-if-change to update the real
4774 file where necessary.
4776 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4778 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4779 rather than Upa for CPY /M.
4781 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4783 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4786 2020-01-06 Martin Liska <mliska@suse.cz>
4788 PR tree-optimization/92860
4789 * params.opt: Mark param_max_combine_insns with Optimization
4792 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4795 * config/i386/i386.md (SWIDWI): New mode iterator.
4796 (DWI, dwi): Add TImode variants.
4797 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4798 <general_hilo_operand> instead of <general_operand>. Use
4799 CONST_SCALAR_INT_P instead of CONST_INT_P.
4800 (*addv<mode>4_1): Rename to ...
4801 (addv<mode>4_1): ... this.
4802 (QWI): New mode attribute.
4803 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4804 define_insn_and_split patterns.
4805 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4807 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4808 <general_hilo_operand> instead of <general_operand>.
4809 (*addcarry<mode>_1): New define_insn.
4810 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4812 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4814 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4815 Use "call" instead of "set".
4817 2020-01-03 Martin Jambor <mjambor@suse.cz>
4820 * ipa-cp.c (print_all_lattices): Skip functions without info.
4822 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4825 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4826 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4827 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4828 for 'e' simd clones.
4831 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4833 (mprefer-vector-width=): Add Save.
4834 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4835 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4836 (ix86_debug_options, ix86_function_specific_print): Adjust
4837 ix86_target_string callers.
4838 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4839 (ix86_valid_target_attribute_tree): Likewise.
4840 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4841 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4842 ix86_target_string caller.
4845 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4846 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4847 instead of gen_int_shift_amount + convert_modes.
4849 PR rtl-optimization/93088
4850 * loop-iv.c (find_single_def_src): Punt after looking through
4851 128 reg copies for regs with single definitions. Move definitions
4854 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4856 * config/arm/arm-c.c (arm_cpu_builtins): Define
4857 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4858 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4859 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4860 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4861 * config/arm/arm-tables.opt: Regenerated.
4862 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4863 arm_arch_i8mm and arm_arch_bf16 when enabled.
4864 * config/arm/arm.h (TARGET_I8MM): New macro.
4865 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4866 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4867 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4868 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4869 (v8_6_a_simd_variants): New.
4870 (v8_*_a_simd_variants): Add i8mm and bf16.
4871 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4873 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4876 * predict.c (compute_function_frequency): Don't call
4877 warn_function_cold on functions that already have cold attribute.
4879 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4882 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4883 COMDAT group function labels in .data.rel.ro.local section.
4884 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4887 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4888 comparison_operator in B and S integer comparisons. Likewise, use
4889 ordered_comparison_operator instead of cmpib_comparison_operator in
4891 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4893 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4895 Update copyright years.
4897 * gcc.c (process_command): Update copyright notice dates.
4898 * gcov-dump.c (print_version): Ditto.
4899 * gcov.c (print_version): Ditto.
4900 * gcov-tool.c (print_version): Ditto.
4901 * gengtype.c (create_file): Ditto.
4902 * doc/cpp.texi: Bump @copying's copyright year.
4903 * doc/cppinternals.texi: Ditto.
4904 * doc/gcc.texi: Ditto.
4905 * doc/gccint.texi: Ditto.
4906 * doc/gcov.texi: Ditto.
4907 * doc/install.texi: Ditto.
4908 * doc/invoke.texi: Ditto.
4910 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4912 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4915 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4917 PR tree-optimization/93098
4918 * match.pd (popcount): For shift amounts, use integer_onep
4919 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4920 tests. Make sure that precision is power of two larger than or equal
4921 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4922 instead of ULL suffixed constants. Formatting fixes.
4924 Copyright (C) 2020 Free Software Foundation, Inc.
4926 Copying and distribution of this file, with or without modification,
4927 are permitted in any medium without royalty provided the copyright
4928 notice and this notice are preserved.