1 2020-03-03 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/94002
4 * explow.c (plus_constant): Punt if cst has VOIDmode and
5 get_pool_mode is different from mode.
7 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
9 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
10 address has an offset which fits the scalling constraint for a
12 (legitimate_scaled_address_p): Update use
13 leigitimate_small_data_address_p.
14 (arc_print_operand): Likewise.
15 (arc_legitimate_address_p): Likewise.
16 (legitimate_small_data_address_p): Likewise.
18 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
20 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
21 (fnmasf4_fpu): Likewise.
23 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
25 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
28 (adddi3_i): Remove pattern.
31 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
33 * config/arc/arc.md (eh_return): Add length info.
35 2020-03-02 David Malcolm <dmalcolm@redhat.com>
37 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
39 2020-03-02 David Malcolm <dmalcolm@redhat.com>
41 * doc/invoke.texi (Static Analyzer Options): Add
42 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
45 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
48 * config/i386/i386.md (movstrict<mode>): Allow only
49 registers with VALID_INT_MODE_P modes.
51 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
53 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
54 (reduc_insn): Use 'U' and 'B' operand codes.
55 (reduc_<reduc_op>_scal_<mode>): Allow all types.
56 (reduc_<reduc_op>_scal_v64di): Delete.
57 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
58 (*plus_carry_dpp_shr_v64si): Change to ...
59 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
60 (mov_from_lane63_v64di): Change to ...
61 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
62 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
63 Support UNSPEC_MOV_DPP_SHR output formats.
64 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
65 Add "use_extends" reductions.
66 (print_operand_address): Add 'I' and 'U' codes.
67 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
69 2020-03-02 Martin Liska <mliska@suse.cz>
71 * lto-wrapper.c: Fix typo in comment about
74 2020-03-01 Martin Sebor <msebor@redhat.com>
77 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
79 2020-03-01 Martin Sebor <msebor@redhat.com>
82 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
83 of a pointer in the outermost ADDR_EXPRs.
85 2020-02-28 Jeff Law <law@redhat.com>
87 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
88 * config/v850/v850.c (v850_asm_trampoline_template): Update
91 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
94 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
97 2020-02-28 Martin Liska <mliska@suse.cz>
100 * configure.ac: Improve detection of ld_date by requiring
101 either two dashes or none.
102 * configure: Regenerate.
104 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
106 PR rtl-optimization/93564
107 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
108 do not honor reg alloc order.
110 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
113 * config/aarch64/aarch64.c (aarch64_override_options): Fix
114 misleading warning string.
116 2020-02-27 Martin Sebor <msebor@redhat.com>
118 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
120 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
123 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
124 Split the insn into two parts. This insn only does variable
125 extract from a register.
126 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
127 variable extract from memory.
128 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
129 only does variable extract from a register.
130 (vsx_extract_v4sf_var_load): New insn, do variable extract from
132 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
133 into two parts. This insn only does variable extract from a
135 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
136 do variable extract from memory.
138 2020-02-27 Martin Jambor <mjambor@suse.cz>
139 Feng Xue <fxue@os.amperecomputing.com>
142 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
143 new function calls_same_node_or_its_all_contexts_clone_p.
144 (cgraph_edge_brings_value_p): Use it.
145 (cgraph_edge_brings_value_p): Likewise.
146 (self_recursive_pass_through_p): Return false if caller is a clone.
147 (self_recursive_agg_pass_through_p): Likewise.
149 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
152 * alias.c (ends_tbaa_access_path_p): Break out from ...
153 (component_uses_parent_alias_set_from): ... here.
154 * alias.h (ends_tbaa_access_path_p): Declare.
155 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
156 handle trailing arrays past end of tbaa access path.
157 (aliasing_component_refs_p): ... here; likewise.
158 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
159 path; disambiguate also past end of it.
160 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
163 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
165 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
166 beginning of the file.
167 (vcreate_bf16, vcombine_bf16): New.
168 (vdup_n_bf16, vdupq_n_bf16): New.
169 (vdup_lane_bf16, vdup_laneq_bf16): New.
170 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
171 (vduph_lane_bf16, vduph_laneq_bf16): New.
172 (vset_lane_bf16, vsetq_lane_bf16): New.
173 (vget_lane_bf16, vgetq_lane_bf16): New.
174 (vget_high_bf16, vget_low_bf16): New.
175 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
176 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
177 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
178 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
179 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
180 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
181 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
182 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
183 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
184 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
185 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
186 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
187 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
188 (vreinterpretq_bf16_p128): New.
189 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
190 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
191 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
192 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
193 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
194 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
195 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
196 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
197 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
198 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
199 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
200 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
201 (vreinterpretq_p128_bf16): New.
202 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
204 (V_elem_l): Likewise.
207 (V_DOUBLE): Likewise.
208 (VDQX): Add V4BF and V8BF.
209 (V_two_elem, V_three_elem, V_four_elem): Likewise.
212 (V_double_vector_mode): Likewise.
213 (V_cmp_result): Likewise.
214 (V_uf_sclr): Likewise.
215 (V_sz_elem): Likewise.
216 (Is_d_reg): Likewise.
217 (V_mode_nunits): Likewise.
218 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
220 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
222 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
223 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
224 (<expander><mode>3<exec>): Likewise.
225 (<expander><mode>3): New.
226 (v<expander><mode>3): New.
227 (<expander><mode>3): New.
228 (<expander><mode>3<exec>): Rename to ...
229 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
230 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
232 2020-02-27 Alexandre Oliva <oliva@adacore.com>
234 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
237 2020-02-27 Richard Biener <rguenther@suse.de>
239 PR tree-optimization/93508
240 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
241 non-_CHK variants. Valueize their length arguments.
243 2020-02-27 Richard Biener <rguenther@suse.de>
245 PR tree-optimization/93953
246 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
247 to the hash-map entry.
249 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
251 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
253 2020-02-27 Mark Williams <mwilliams@fb.com>
255 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
256 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
257 -ffile-prefix-map and -fmacro-prefix-map.
258 * lto-streamer-out.c: Include file-prefix-map.h.
259 (lto_output_location): Remap the file part of locations.
261 2020-02-27 Jakub Jelinek <jakub@redhat.com>
264 * gimplify.c (gimplify_init_constructor): Don't promote readonly
265 DECL_REGISTER variables to TREE_STATIC.
267 PR tree-optimization/93582
268 PR tree-optimization/93945
269 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
270 non-zero INTEGER_CST second argument and ref->offset or ref->size
271 not a multiple of BITS_PER_UNIT.
273 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
275 * doc/install.texi (Binaries): Update description of BullFreeware.
277 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
281 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
282 C++ Language Options, Warning Options, and Static Analyzer
283 Options lists. Document negative form of options enabled by
284 default. Move some things around to more accurately sort
285 warnings by category.
286 (C++ Dialect Options, Warning Options, Static Analyzer
287 Options): Document negative form of options when enabled by
288 default. Move some things around to more accurately sort
289 warnings by category. Add some missing index entries.
292 2020-02-26 Carl Love <cel@us.ibm.com>
295 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
296 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
297 for the vector unsigned short arguments. It is also listed as the
298 name of the built-in for arguments vector unsigned short,
299 vector unsigned int and vector unsigned long long built-ins. The
300 name of the builtins for these arguments should be:
301 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
302 __builtin_crypto_vpmsumd respectively.
304 2020-02-26 Richard Biener <rguenther@suse.de>
306 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
307 and load permutation.
309 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
312 * optabs-tree.c (supportable_convert_operation): Reject types with
315 2020-02-26 David Malcolm <dmalcolm@redhat.com>
317 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
319 2020-02-26 Jakub Jelinek <jakub@redhat.com>
321 PR tree-optimization/93820
322 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
323 argument to ALL_INTEGER_CST_P boolean.
324 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
325 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
326 adjacent INTEGER_CST store into merged_store->only_constants like
329 2020-02-25 Jakub Jelinek <jakub@redhat.com>
332 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
334 * cfghooks.c (verify_flow_info): Likewise.
335 * predict.c (combine_predictions_for_bb): Likewise.
336 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
337 sucessor -> successor.
338 (find_traces_1_round): Fix comment typo, destinarion -> destination.
339 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
341 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
342 message typo, sucessors -> successors.
344 2020-02-25 Martin Sebor <msebor@redhat.com>
346 * doc/extend.texi (attribute access): Correct an example.
348 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
350 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
352 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
354 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
355 (VD): Enable for V4BF.
357 (VQ): Enable for V8BF.
360 (VDBL, Vdbl): Add V4BF.
361 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
362 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
363 (bfloat16x8x2_t): Likewise.
364 (bfloat16x4x3_t): Likewise.
365 (bfloat16x8x3_t): Likewise.
366 (bfloat16x4x4_t): Likewise.
367 (bfloat16x8x4_t): Likewise.
368 (vcombine_bf16): New.
369 (vld1_bf16, vld1_bf16_x2): New.
370 (vld1_bf16_x3, vld1_bf16_x4): New.
371 (vld1q_bf16, vld1q_bf16_x2): New.
372 (vld1q_bf16_x3, vld1q_bf16_x4): New.
373 (vld1_lane_bf16): New.
374 (vld1q_lane_bf16): New.
375 (vld1_dup_bf16): New.
376 (vld1q_dup_bf16): New.
379 (vld2_dup_bf16): New.
380 (vld2q_dup_bf16): New.
383 (vld3_dup_bf16): New.
384 (vld3q_dup_bf16): New.
387 (vld4_dup_bf16): New.
388 (vld4q_dup_bf16): New.
389 (vst1_bf16, vst1_bf16_x2): New.
390 (vst1_bf16_x3, vst1_bf16_x4): New.
391 (vst1q_bf16, vst1q_bf16_x2): New.
392 (vst1q_bf16_x3, vst1q_bf16_x4): New.
393 (vst1_lane_bf16): New.
394 (vst1q_lane_bf16): New.
402 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
404 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
405 (VALL_F16): Likewise.
406 (VALLDI_F16): Likewise.
409 (vswap_width_name): Likewise.
410 (VSWAP_WIDTH): Likewise.
414 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
415 (vget_lane_bf16, vgetq_lane_bf16): New.
417 (vdup_n_bf16, vdupq_n_bf16): New.
418 (vdup_lane_bf16, vdup_laneq_bf16): New.
419 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
420 (vduph_lane_bf16, vduph_laneq_bf16): New.
421 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
422 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
423 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
424 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
425 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
426 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
427 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
428 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
429 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
430 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
431 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
432 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
433 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
434 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
435 (vreinterpretq_bf16_p128): New.
436 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
437 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
438 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
439 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
440 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
441 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
442 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
443 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
444 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
445 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
446 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
447 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
448 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
449 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
450 (vreinterpretq_p128_bf16): New.
452 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
454 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
455 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
456 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
457 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
458 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
459 * config/arm/iterators.md (VSF2BF): New attribute.
460 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
461 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
462 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
464 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
466 * config/arm/arm.md (required_for_purecode): New attribute.
467 (enabled): Handle required_for_purecode.
468 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
469 work with -mpure-code.
471 2020-02-25 Jakub Jelinek <jakub@redhat.com>
473 PR rtl-optimization/93908
474 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
477 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
479 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
481 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
483 * doc/install.texi (--enable-checking): Adjust wording.
485 2020-02-25 Richard Biener <rguenther@suse.de>
487 PR tree-optimization/93868
488 * tree-vect-slp.c (slp_copy_subtree): New function.
489 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
490 re-arranging stmts in it.
492 2020-02-25 Jakub Jelinek <jakub@redhat.com>
495 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
496 dummy function and remove it at the end.
499 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
500 paramter -> parameter.
501 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
502 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
504 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
506 * doc/install.texi (--enable-checking): Properly document current
508 (--enable-stage1-checking): Minor clarification about bootstrap.
510 2020-02-24 David Malcolm <dmalcolm@redhat.com>
513 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
514 -fanalyzer-checker=taint is also required.
515 (-fanalyzer-checker=): Note that providing this option enables the
516 given checker, and doing so may be required for checkers that are
519 2020-02-24 David Malcolm <dmalcolm@redhat.com>
521 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
522 significant control flow events; add a "3" which shows all
523 control flow events; the old "3" becomes "4".
525 2020-02-24 Jakub Jelinek <jakub@redhat.com>
527 PR tree-optimization/93582
528 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
529 pd.offset and pd.size to be counted in bits rather than bytes, add
530 support for maxsizei that is not a multiple of BITS_PER_UNIT and
531 handle bitfield stores and loads.
532 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
533 uncomparable quantities - bytes vs. bits. Allow push_partial_def
534 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
535 pd.offset/pd.size to be counted in bits rather than bytes.
536 Formatting fix. Rename shadowed len variable to buflen.
538 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
539 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
542 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
543 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
544 * opts-common.c (parse_options_from_collect_gcc_options): New function.
545 (prepend_xassembler_to_collect_as_options): Likewise.
546 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
547 (prepend_xassembler_to_collect_as_options): Likewise.
548 * lto-opts.c (lto_write_options): Stream assembler options
549 in COLLECT_AS_OPTIONS.
550 * lto-wrapper.c (xassembler_options_error): New static variable.
551 (get_options_from_collect_gcc_options): Move parsing options code to
552 parse_options_from_collect_gcc_options and call it.
553 (merge_and_complain): Validate -Xassembler options.
554 (append_compiler_options): Handle OPT_Xassembler.
555 (run_gcc): Append command line -Xassembler options to
557 * doc/invoke.texi: Add documentation about using Xassembler
560 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
562 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
564 (riscv_rtx_costs): Update cost model for LTGT.
566 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
568 PR rtl-optimization/93564
569 * ira-color.c (struct update_cost_queue_elem): New member start.
570 (queue_update_cost, get_next_update_cost): Add new arg start.
571 (allocnos_conflict_p): New function.
572 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
573 Add checking conflicts with allocnos_conflict_p.
574 (update_costs_from_prefs, restore_costs_from_copies): Adjust
575 update_costs_from_allocno calls.
576 (update_conflict_hard_regno_costs): Add checking conflicts with
577 allocnos_conflict_p. Adjust calls of queue_update_cost and
578 get_next_update_cost.
579 (assign_hard_reg): Adjust calls of queue_update_cost. Add
581 (bucket_allocno_compare_func): Restore previous version.
583 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
585 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
586 double-word size when handling aggregate return values.
587 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
588 that homogeneous SFmode and DFmode aggregates are passed and returned
589 in general registers.
591 2020-02-21 Jakub Jelinek <jakub@redhat.com>
594 * opts.c (print_filtered_help): Translate help before appending
595 messages to it rather than after that.
597 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
599 PR rtl-optimization/PR92989
600 * lra-lives.c (process_bb_lives): Restore the original order
601 of the bb liveness update. Call make_hard_regno_dead for each
602 register clobbered at the start of an EH receiver.
604 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
607 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
608 self-recursively generated.
610 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
613 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
616 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
618 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
619 Document new target supports option.
621 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
623 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
624 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
625 * config/arm/iterators.md (MATMUL): New iterator.
626 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
627 (mmla_sfx): New attribute.
628 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
629 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
630 (UNSPEC_MATMUL_US): New.
632 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
634 * config/arm/arm.md: Prevent scalar shifts from being used when big
637 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
638 Richard Biener <rguenther@suse.de>
640 PR tree-optimization/93586
641 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
642 after mismatched array refs; do not sure type size information to
643 recover from unmatched referneces with !flag_strict_aliasing_p.
645 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
647 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
648 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
649 (scatter_store<mode>): Rename to ...
650 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
651 (scatter<mode>_exec): Delete. Move contents ...
652 (mask_scatter_store<mode>): ... here, and rename that to ...
653 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
654 Remove mode conversion.
655 (mask_gather_load<mode>): Rename to ...
656 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
657 Remove mode conversion.
658 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
660 2020-02-21 Martin Jambor <mjambor@suse.cz>
662 PR tree-optimization/93845
663 * tree-sra.c (verify_sra_access_forest): Only test access size of
666 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
668 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
669 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
670 (addv64di3_exec): Likewise.
671 (subv64di3): Likewise.
672 (subv64di3_exec): Likewise.
673 (addv64di3_zext): Likewise.
674 (addv64di3_zext_exec): Likewise.
675 (addv64di3_zext_dup): Likewise.
676 (addv64di3_zext_dup_exec): Likewise.
677 (addv64di3_zext_dup2): Likewise.
678 (addv64di3_zext_dup2_exec): Likewise.
679 (addv64di3_sext_dup2): Likewise.
680 (addv64di3_sext_dup2_exec): Likewise.
681 (<expander>v64di3): Likewise.
682 (<expander>v64di3_exec): Likewise.
683 (*<reduc_op>_dpp_shr_v64di): Likewise.
684 (*plus_carry_dpp_shr_v64di): Likewise.
685 * config/gcn/gcn.md (adddi3): Likewise.
686 (addptrdi3): Likewise.
687 (<expander>di3): Likewise.
689 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
691 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
693 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
695 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
696 support. Use aarch64_emit_mult instead of emitting multiplication
697 instructions directly.
698 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
699 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
701 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
703 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
704 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
705 instead of emitting multiplication instructions directly.
706 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
707 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
708 (@aarch64_frecps<mode>): New expanders.
710 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
712 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
713 on and produce uint64_ts rather than ints.
714 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
715 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
717 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
719 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
720 an unused xmsk register when handling approximate rsqrt.
722 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
724 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
725 flag_finite_math_only condition.
727 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
730 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
731 to destination operand for shufps alternative.
732 (*vec_extractv2si_1): Ditto.
734 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
737 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
740 2020-02-20 Martin Liska <mliska@suse.cz>
743 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
745 2020-02-20 Martin Liska <mliska@suse.cz>
748 * common/config/avr/avr-common.c: Remote trailing "|".
750 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
752 * collect2.c (maybe_run_lto_and_relink): Fix typo in
755 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
757 PR tree-optimization/93767
758 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
759 access-size bias from the offset calculations for negative strides.
761 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
763 * collect2.c (c_file, o_file): Make const again.
764 (ldout,lderrout, dump_ld_file): Remove.
765 (tool_cleanup): Avoid calling not signal-safe functions.
766 (maybe_run_lto_and_relink): Avoid possible signal handler
767 access to unintialzed memory (lto_o_files).
768 (main): Avoid leaking temp files in $TMPDIR.
769 Initialize c_file/o_file with concat, which avoids exposing
770 uninitialized memory to signal handler, which calls unlink(!).
771 Avoid calling maybe_unlink when the main function returns,
772 since the atexit handler is already doing this.
773 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
775 2020-02-19 Martin Jambor <mjambor@suse.cz>
777 PR tree-optimization/93776
778 * tree-sra.c (create_access): Do not create zero size accesses.
779 (get_access_for_expr): Do not search for zero sized accesses.
781 2020-02-19 Martin Jambor <mjambor@suse.cz>
783 PR tree-optimization/93667
784 * tree-sra.c (scalarizable_type_p): Return false if record fields
785 do not follow wach other.
787 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
789 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
790 rather than fmv.x.s/fmv.s.x.
792 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
794 * config/aarch64/aarch64-simd-builtins.def
795 (intrinsic_vec_smult_lo_): New.
796 (intrinsic_vec_umult_lo_): Likewise.
797 (vec_widen_smult_hi_): Likewise.
798 (vec_widen_umult_hi_): Likewise.
799 * config/aarch64/aarch64-simd.md
800 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
801 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
802 (vmull_high_s16): Likewise.
803 (vmull_high_s32): Likewise.
804 (vmull_high_u8): Likewise.
805 (vmull_high_u16): Likewise.
806 (vmull_high_u32): Likewise.
807 (vmull_s8): Likewise.
808 (vmull_s16): Likewise.
809 (vmull_s32): Likewise.
810 (vmull_u8): Likewise.
811 (vmull_u16): Likewise.
812 (vmull_u32): Likewise.
814 2020-02-18 Martin Liska <mliska@suse.cz>
816 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
817 bootstrap by missing removal of invalid sanity check.
819 2020-02-18 Martin Liska <mliska@suse.cz>
822 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
823 Always compare LHS of gimple_assign.
825 2020-02-18 Martin Liska <mliska@suse.cz>
828 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
829 and return type of functions.
830 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
831 Drop MALLOC attribute for void functions.
832 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
833 malloc_state for a new VOID clone.
835 2020-02-18 Martin Liska <mliska@suse.cz>
838 * common.opt: Add -fprofile-reproducibility.
839 * doc/invoke.texi: Document it.
840 * value-prof.c (dump_histogram_value):
841 Document and support behavior for counters[0]
842 being a negative value.
843 (get_nth_most_common_value): Handle negative
844 counters[0] in respect to flag_profile_reproducible.
846 2020-02-18 Jakub Jelinek <jakub@redhat.com>
849 * cgraph.c (verify_speculative_call): Use speculative_id instead of
850 speculative_uid in messages. Remove trailing whitespace from error
851 message. Use num_speculative_call_targets instead of
852 num_speculative_targets in a message.
853 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
854 edge messages and stmt instead of cal_stmt in reference message.
856 PR tree-optimization/93780
857 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
858 before calling build_vector_type.
859 (execute_update_addresses_taken): Likewise.
862 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
863 typo, functoin -> function.
864 * tree.c (free_lang_data_in_decl): Fix comment typo,
865 functoin -> function.
866 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
868 2020-02-17 David Malcolm <dmalcolm@redhat.com>
870 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
872 (print_option_information): Don't call get_option_url if URLs
875 2020-02-17 Alexandre Oliva <oliva@adacore.com>
877 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
878 handling of register_common-less targets.
880 2020-02-17 Martin Liska <mliska@suse.cz>
883 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
885 2020-02-17 Martin Liska <mliska@suse.cz>
888 * config/rs6000/rs6000.c (rs6000_option_override_internal):
891 2020-02-17 Martin Liska <mliska@suse.cz>
894 * config/rx/elf.opt: Fix typo.
896 2020-02-17 Richard Biener <rguenther@suse.de>
899 * opts-global.c (print_ignored_options): Use inform and
902 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
905 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
907 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
910 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
911 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
913 2020-02-15 Jason Merrill <jason@redhat.com>
915 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
917 2020-02-15 Jakub Jelinek <jakub@redhat.com>
919 PR tree-optimization/93744
920 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
921 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
922 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
923 sure @2 in the first and @1 in the other patterns has no side-effects.
925 2020-02-15 David Malcolm <dmalcolm@redhat.com>
926 Bernd Edlinger <bernd.edlinger@hotmail.de>
930 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
931 * configure.ac (--with-diagnostics-urls): New configuration
932 option, based on --with-diagnostics-color.
933 (DIAGNOSTICS_URLS_DEFAULT): New define.
934 * config.h: Regenerate.
935 * configure: Regenerate.
936 * diagnostic.c (diagnostic_urls_init): Handle -1 for
937 DIAGNOSTICS_URLS_DEFAULT from configure-time
938 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
939 and TERM_URLS environment variable.
940 * diagnostic-url.h (diagnostic_url_format): New enum type.
941 (diagnostic_urls_enabled_p): rename to...
942 (determine_url_format): ... this, and change return type.
943 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
944 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
945 the linux console, and mingw.
946 (diagnostic_urls_enabled_p): rename to...
947 (determine_url_format): ... this, and adjust.
948 * pretty-print.h (pretty_printer::show_urls): rename to...
949 (pretty_printer::url_format): ... this, and change to enum.
950 * pretty-print.c (pretty_printer::pretty_printer,
951 pp_begin_url, pp_end_url, test_urls): Adjust.
952 * doc/install.texi (--with-diagnostics-urls): Document the new
953 configuration option.
954 (--with-diagnostics-color): Document the existing interaction
955 with GCC_COLORS better.
956 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
957 vindex reference. Update description of defaults based on the above.
958 (-fdiagnostics-color): Update description of how -fdiagnostics-color
959 interacts with GCC_COLORS.
961 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
964 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
965 conjunction with TARGET_GNU_TLS in early return.
967 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
969 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
970 the mode is not wider than UNITS_PER_WORD.
972 2020-02-14 Martin Jambor <mjambor@suse.cz>
974 PR tree-optimization/93516
975 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
976 access of the same type as the parent.
977 (propagate_subaccesses_from_lhs): Likewise.
979 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
982 * config/i386/avx512vbmi2intrin.h
983 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
984 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
985 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
986 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
987 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
988 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
989 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
990 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
991 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
992 of lacking a closing parenthesis.
993 * config/i386/avx512vbmi2vlintrin.h
994 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
995 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
996 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
997 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
998 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
999 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1000 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1001 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1002 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1003 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1004 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1005 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1006 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1007 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1008 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1009 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1010 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1011 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1013 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1016 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1017 the target function entry.
1019 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1021 * common/config/arc/arc-common.c (arc_option_optimization_table):
1022 Disable if-conversion step when optimized for size.
1024 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1026 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1027 R12-R15 are always in ARCOMPACT16_REGS register class.
1028 * config/arc/arc.opt (mq-class): Deprecate.
1029 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1031 * doc/invoke.texi (mq-class): Update text.
1032 * common/config/arc/arc-common.c (arc_option_optimization_table):
1035 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1037 * config/arc/arc.c (arc_insn_cost): New function.
1038 (TARGET_INSN_COST): Define.
1039 * config/arc/arc.md (cost): New attribute.
1040 (add_n): Use arc_nonmemory_operand.
1041 (ashlsi3_insn): Likewise, also update constraints.
1042 (ashrsi3_insn): Likewise.
1043 (rotrsi3): Likewise.
1044 (add_shift): Likewise.
1045 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1047 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1049 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1051 (umulsidi_600): Likewise.
1053 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1056 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1057 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1058 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1059 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1060 pass __A to the builtin followed by __W instead of __A followed by
1062 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1063 _mm512_mask_popcnt_epi64): Likewise.
1064 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1065 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1066 _mm256_mask_popcnt_epi64): Likewise.
1068 PR tree-optimization/93582
1069 * fold-const.h (shift_bytes_in_array_left,
1070 shift_bytes_in_array_right): Declare.
1071 * fold-const.c (shift_bytes_in_array_left,
1072 shift_bytes_in_array_right): New function, moved from
1073 gimple-ssa-store-merging.c, no longer static.
1074 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1075 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1076 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1077 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1078 shift_bytes_in_array.
1079 (verify_shift_bytes_in_array): Rename to ...
1080 (verify_shift_bytes_in_array_left): ... this. Use
1081 shift_bytes_in_array_left instead of shift_bytes_in_array.
1082 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1083 instead of verify_shift_bytes_in_array.
1084 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1085 / native_interpret_expr where the store covers all needed bits,
1086 punt on PDP-endian, otherwise allow all involved offsets and sizes
1087 not to be byte-aligned.
1090 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1091 use const_0_to_255_operand predicate instead of immediate_operand.
1092 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1093 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1094 vgf2p8affineinvqb_<mode><mask_name>,
1095 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1096 const_0_to_255_operand predicated operands.
1098 2020-02-12 Jeff Law <law@redhat.com>
1100 * config/h8300/h8300.md (comparison shortening peepholes): Use
1101 a mode iterator to merge the HImode and SImode peepholes.
1103 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1106 * real.c (is_even): Make static. Function comment fix.
1107 (is_halfway_below): Make static, don't assert R is not inf/nan,
1108 instead return false for those. Small formatting fixes.
1110 2020-02-12 Martin Sebor <msebor@redhat.com>
1113 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1114 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1115 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1116 (strlen_check_and_optimize_call): Adjust callee name.
1118 2020-02-12 Jeff Law <law@redhat.com>
1120 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1121 (and (xor)) variant. Combine other two into single peephole.
1123 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1125 PR rtl-optimization/93565
1126 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1128 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1130 * config/aarch64/aarch64-simd.md
1131 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1132 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1133 generating separate ADDV and zero_extend patterns.
1134 * config/aarch64/iterators.md (VDQV_E): New iterator.
1136 2020-02-12 Jeff Law <law@redhat.com>
1138 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1139 expanders, splits, etc.
1140 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1141 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1142 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1143 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1144 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1145 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1147 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1149 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1152 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1153 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1154 TARGET_AVX512DQ from condition.
1155 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1156 instead of <mask_mode512bit_condition> in condition. If
1157 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1159 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1162 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1165 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1167 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1169 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1170 where strlen is more legible.
1171 (rs6000_builtin_vectorized_libmass): Ditto.
1172 (rs6000_print_options_internal): Ditto.
1174 2020-02-11 Martin Sebor <msebor@redhat.com>
1176 PR tree-optimization/93683
1177 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1179 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1181 * config/rs6000/predicates.md (cint34_operand): Rename the
1182 -mprefixed-addr option to be -mprefixed.
1183 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1184 the -mprefixed-addr option to be -mprefixed.
1185 (OTHER_FUTURE_MASKS): Likewise.
1186 (POWERPC_MASKS): Likewise.
1187 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1188 the -mprefixed-addr option to be -mprefixed. Change error
1189 messages to refer to -mprefixed.
1190 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1192 (rs6000_legitimate_offset_address_p): Likewise.
1193 (rs6000_mode_dependent_address): Likewise.
1194 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1195 "-mprefixed" for target attributes and pragmas.
1196 (address_to_insn_form): Rename the -mprefixed-addr option to be
1198 (rs6000_adjust_insn_length): Likewise.
1199 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1200 -mprefixed-addr option to be -mprefixed.
1201 (ASM_OUTPUT_OPCODE): Likewise.
1202 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1203 -mprefixed-addr option to be -mprefixed.
1204 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1205 -mprefixed-addr option to be prefixed. Change the option from
1206 being undocumented to being documented.
1207 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1208 -mprefixed option. Update the -mpcrel documentation to mention
1211 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1213 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1214 including FIRST_PSEUDO_REGISTER - 1.
1215 * ira-color.c (print_hard_reg_set): Ditto.
1217 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1219 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1220 (USTERNOP_QUALIFIERS): New define.
1221 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1222 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1223 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1224 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1225 * config/arm/arm_neon.h (vusdot_s32): New.
1226 (vusdot_lane_s32): New.
1227 (vusdotq_lane_s32): New.
1228 (vsudot_lane_s32): New.
1229 (vsudotq_lane_s32): New.
1230 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1231 * config/arm/iterators.md (DOTPROD_I8MM): New.
1232 (sup, opsuffix): Add <us/su>.
1233 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1234 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1236 2020-02-11 Richard Biener <rguenther@suse.de>
1238 PR tree-optimization/93661
1239 PR tree-optimization/93662
1240 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1242 * tree-sra.c (get_access_for_expr): Likewise.
1244 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1247 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1248 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1249 Change condition from TARGET_AVX2 to TARGET_AVX.
1251 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1254 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1255 argument of strncmp.
1257 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1259 Try to generate zero-based comparisons.
1260 * config/cris/cris.c (cris_reduce_compare): New function.
1261 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1262 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1263 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1265 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1268 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1269 in Thumb state and also as a destination in Arm state. Add T16
1272 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1274 * md.texi (Define Subst): Match closing paren in example.
1276 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1280 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1281 arguments of strncmp.
1283 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1286 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1287 but different source value.
1288 (adjust_callers_for_value_intersection): New function.
1289 (gather_edges_for_value): Adjust order of callers to let a
1290 non-self-recursive caller be the first element.
1291 (self_recursive_pass_through_p): Add a new parameter "simple", and
1292 check generalized self-recursive pass-through jump function.
1293 (self_recursive_agg_pass_through_p): Likewise.
1294 (find_more_scalar_values_for_callers_subset): Compute value from
1295 pass-through jump function for self-recursive.
1296 (intersect_with_plats): Cleanup previous implementation code for value
1297 itersection with self-recursive call edge.
1298 (intersect_with_agg_replacements): Likewise.
1299 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1300 function for self-recursive call edge. Cleanup previous implementation
1301 code for value intersection with self-recursive call edge.
1302 (decide_whether_version_node): Remove dead callers and adjust order
1303 to let a non-self-recursive caller be the first element.
1305 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1307 * recog.c: Move pass_split_before_sched2 code in front of
1308 pass_split_before_regstack.
1309 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1310 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1311 (rest_of_handle_split_before_sched2): Remove.
1312 (pass_split_before_sched2::execute): Unconditionally call
1314 (enable_split_before_sched2): New function.
1315 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1316 (pass_split_before_regstack::gate): Ditto.
1317 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1318 Update name check for renamed split4 pass.
1319 * config/sh/sh.c (register_sh_passes): Update pass insertion
1320 point for renamed split4 pass.
1322 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1324 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1325 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1326 copying them around between host and target.
1328 2020-02-08 Andrew Pinski <apinski@marvell.com>
1331 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1332 STRICT_ALIGNMENT also.
1334 2020-02-08 Jim Wilson <jimw@sifive.com>
1337 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1339 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1340 Jakub Jelinek <jakub@redhat.com>
1343 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1344 xmm16-xmm31 call-used even in 64-bit ms-abi.
1346 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1348 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1349 (simd_ummla, simd_usmmla): Likewise.
1350 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1351 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1352 (vusmmlaq_s32): New.
1354 2020-02-07 Richard Biener <rguenther@suse.de>
1357 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1358 skipping unreachable regions.
1359 (optimize_inline_calls): Skip folding stmts when we didn't
1362 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1365 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1366 Don't return aggregates with only SFmode and DFmode in SSE
1368 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1370 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1373 * config/rs6000/rs6000-logue.c
1374 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1375 if it fails, move rs into end_addr and retry. Add
1376 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1377 the insn pattern doesn't describe well what exactly happens to
1381 * config/i386/predicates.md (avx_identity_operand): Remove.
1382 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1383 (avx_<castmode><avxsizesuffix>_<castmode>,
1384 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1385 a VEC_CONCAT of the operand and UNSPEC_CAST.
1386 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1387 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1391 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1392 recog_data.insn if distance_non_agu_define changed it.
1394 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1397 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1398 we only had X-FORM (reg+reg) addressing for vectors. Also before
1399 ISA 3.0, we only had X-FORM addressing for scalars in the
1400 traditional Altivec registers.
1402 2020-02-06 <zhongyunde@huawei.com>
1403 Vladimir Makarov <vmakarov@redhat.com>
1405 PR rtl-optimization/93561
1406 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1407 hard register range.
1409 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1411 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1414 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1416 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1417 where the low and the high 32 bits are equal to each other specially,
1418 with an rldimi instruction.
1420 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1422 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1424 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1426 * config/arm/arm-tables.opt: Regenerate.
1428 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1431 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1432 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1433 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1435 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1437 PR rtl-optimization/87763
1438 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1440 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1442 * config/aarch64/aarch64-simd-builtins.def
1443 (bfmlaq): New built-in function.
1444 (bfmlalb): New built-in function.
1445 (bfmlalt): New built-in function.
1446 (bfmlalb_lane): New built-in function.
1447 (bfmlalt_lane): New built-in function.
1448 * config/aarch64/aarch64-simd.md
1449 (aarch64_bfmmlaqv4sf): New pattern.
1450 (aarch64_bfmlal<bt>v4sf): New pattern.
1451 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1452 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1453 (vbfmlalbq_f32): New intrinsic.
1454 (vbfmlaltq_f32): New intrinsic.
1455 (vbfmlalbq_lane_f32): New intrinsic.
1456 (vbfmlaltq_lane_f32): New intrinsic.
1457 (vbfmlalbq_laneq_f32): New intrinsic.
1458 (vbfmlaltq_laneq_f32): New intrinsic.
1459 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1460 (bt): New int attribute.
1462 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1464 * config/i386/i386.md (*pushtf): Emit "#" instead of
1465 calling gcc_unreachable in insn output.
1468 (*pushsf_rex64): Ditto for alternatives other than 1.
1469 (*pushsf): Ditto for alternatives other than 1.
1471 2020-02-06 Martin Liska <mliska@suse.cz>
1473 PR gcov-profile/91971
1474 PR gcov-profile/93466
1475 * coverage.c (coverage_init): Revert mangling of
1476 path into filename. It can lead to huge filename length.
1477 Creation of subfolders seem more natural.
1479 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1482 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1483 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1484 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1486 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1489 * config/i386/predicates.md (avx_identity_operand): New predicate.
1490 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1491 define_insn_and_split.
1494 * omp-low.c (use_pointer_for_field): For nested constructs, also
1495 look for map clauses on target construct.
1496 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1497 taskreg_nesting_level.
1500 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1501 shared clause, call omp_notice_variable on outer context if any.
1503 2020-02-05 Jason Merrill <jason@redhat.com>
1506 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1507 non-zero address even if weak and not yet defined.
1509 2020-02-05 Martin Sebor <msebor@redhat.com>
1511 PR tree-optimization/92765
1512 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1513 * tree-ssa-strlen.c (compute_string_length): Remove.
1514 (determine_min_objsize): Remove.
1515 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1516 Avoid using type size as the upper bound on string length.
1517 (handle_builtin_string_cmp): Add an argument. Adjust.
1518 (strlen_check_and_optimize_call): Pass additional argument to
1519 handle_builtin_string_cmp.
1521 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1523 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1524 (*pushdi2_rex64 peephole2): Unconditionally split after
1526 (*ashl<mode>3_doubleword): Ditto.
1527 (*<shift_insn><mode>3_doubleword): Ditto.
1529 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1532 * config/rs6000/rs6000.c (get_vector_offset): Fix
1534 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1536 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1538 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1541 (Special Functions for Debugging the Analyzer): Update description
1542 of __analyzer_dump_exploded_nodes.
1544 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1547 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1548 include sets and not clobbers in the vzeroupper pattern.
1549 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1550 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1551 (*avx_vzeroupper_1): New define_insn_and_split.
1554 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1555 don't run when !optimize.
1556 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1559 2020-02-05 Richard Biener <rguenther@suse.de>
1562 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1563 checks before matching calls.
1565 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1567 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1568 function comment typo.
1571 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1572 simd_clone_create failed when i == 0, adjust clone->nargs by
1575 2020-02-05 Martin Liska <mliska@suse.cz>
1578 * doc/invoke.texi: Document that one should
1579 not combine ASLR and -fpch.
1581 2020-02-04 Richard Biener <rguenther@suse.de>
1583 PR tree-optimization/93538
1584 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1586 2020-02-04 Richard Biener <rguenther@suse.de>
1588 PR tree-optimization/91123
1589 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1590 (vn_walk_cb_data::last_vuse): New member.
1591 (vn_walk_cb_data::saved_operands): Likewsie.
1592 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1593 (vn_walk_cb_data::push_partial_def): Use finish.
1594 (vn_reference_lookup_2): Update last_vuse and use finish if
1595 we've saved operands.
1596 (vn_reference_lookup_3): Use finish and update calls to
1597 push_partial_defs everywhere. When translating through
1598 memcpy or aggregate copies save off operands and alias-set.
1599 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1600 operation for redundant store removal.
1602 2020-02-04 Richard Biener <rguenther@suse.de>
1604 PR tree-optimization/92819
1605 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1606 generating more stmts than before.
1608 2020-02-04 Martin Liska <mliska@suse.cz>
1610 * config/arm/arm.c (arm_gen_far_branch): Move the function
1611 outside of selftests.
1613 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1615 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1616 function to adjust PC-relative vector addresses.
1617 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1618 handle vectors with PC-relative addresses.
1620 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1622 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1624 (hard_reg_and_mode_to_addr_mask): Delete.
1625 (rs6000_adjust_vec_address): If the original vector address
1626 was REG+REG or REG+OFFSET and the element is not zero, do the add
1627 of the elements in the original address before adding the offset
1628 for the vector element. Use address_to_insn_form to validate the
1629 address using the register being loaded, rather than guessing
1630 whether the address is a DS-FORM or DQ-FORM address.
1632 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1634 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1635 to calculate the offset in memory from the start of a vector of a
1636 particular element. Add code to keep the element number in
1637 bounds if the element number is variable.
1638 (rs6000_adjust_vec_address): Move calculation of offset of the
1639 vector element to get_vector_offset.
1640 (rs6000_split_vec_extract_var): Do not do the initial AND of
1641 element here, move the code to get_vector_offset.
1643 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1645 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1648 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1650 * config/rs6000/constraints.md: Improve documentation.
1652 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1655 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1656 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1658 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1660 * config.gcc: Remove "carrizo" support.
1661 * config/gcn/gcn-opts.h (processor_type): Likewise.
1662 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1663 * config/gcn/gcn.opt (gpu_type): Likewise.
1664 * config/gcn/t-omp-device: Likewise.
1666 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1669 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1670 * config/arm/arm.c (arm_gen_far_branch): New function
1672 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1674 2020-02-03 Julian Brown <julian@codesourcery.com>
1675 Tobias Burnus <tobias@codesourcery.com>
1677 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1679 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1682 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1683 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1686 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1688 PR rtl-optimization/91333
1689 * ira-color.c (struct allocno_color_data): Add member
1691 (init_allocno_threads): Set the member up.
1692 (bucket_allocno_compare_func): Add compare hard reg
1695 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1697 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1699 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1700 * config.in: Regenerated.
1701 * configure: Regenerated.
1702 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1703 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1704 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1706 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1708 * configure: Regenerate.
1710 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1712 PR rtl-optimization/91333
1713 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1714 reg preferences comparison up.
1716 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1718 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1719 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1720 aarch64-sve-builtins-base.h.
1721 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1722 aarch64-sve-builtins-base.cc.
1723 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1724 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1726 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1727 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1728 (svcvtnt): New functions.
1729 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1730 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1731 (svcvtnt): New functions.
1732 (svcvt): Add a form that converts f32 to bf16.
1733 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1734 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1736 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1737 Treat B as bfloat16_t.
1738 (ternary_bfloat_lane_base): New class.
1739 (ternary_bfloat_def): Likewise.
1740 (ternary_bfloat): New shape.
1741 (ternary_bfloat_lane_def): New class.
1742 (ternary_bfloat_lane): New shape.
1743 (ternary_bfloat_lanex2_def): New class.
1744 (ternary_bfloat_lanex2): New shape.
1745 (ternary_bfloat_opt_n_def): New class.
1746 (ternary_bfloat_opt_n): New shape.
1747 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1748 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1749 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1750 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1751 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1752 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1753 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1754 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1755 the pattern off the narrow mode instead of the wider one.
1756 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1757 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1758 (sve_fp_op): Handle them.
1759 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1760 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1762 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1764 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1765 * config/aarch64/aarch64-modes.def (BF): Move definition before
1766 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1767 (SVE_MODES): Handle BF modes.
1768 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1770 (aarch64_full_sve_mode): Likewise.
1771 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1773 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1774 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1775 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1776 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1778 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1780 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1781 (TYPES_all_data): Add bf16.
1782 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1783 (register_tuple_type): Increase buffer size.
1784 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1785 (bf16): New type suffix.
1786 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1787 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1788 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1789 Change type from all_data to all_arith.
1790 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1793 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1794 Matthew Malcomson <matthew.malcomson@arm.com>
1795 Richard Sandiford <richard.sandiford@arm.com>
1797 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1798 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1799 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1800 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1801 __ARM_FEATURE_MATMUL_FP64.
1802 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1803 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1804 be disabled at the same time.
1805 (f32mm): New extension.
1806 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1807 (AARCH64_FL_F64MM): Bump to the next bit up.
1808 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1809 (TARGET_SVE_F64MM): New macros.
1810 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1811 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1812 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1813 (UNSPEC_ZIP2Q): New unspeccs.
1814 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1815 (optab, sur, perm_insn): Handle the new unspecs.
1816 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1817 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1818 TARGET_SVE_F64MM instead of separate tests.
1819 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1820 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1821 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1822 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1823 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1824 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1825 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1826 (TYPES_s_signed): New macro.
1827 (TYPES_s_integer): Use it.
1828 (TYPES_d_float): New macro.
1829 (TYPES_d_data): Use it.
1830 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1831 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1832 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1833 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1834 (svmmla): New shape.
1835 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1836 template parameters.
1837 (ternary_resize2_lane_base): Likewise.
1838 (ternary_resize2_base): New class.
1839 (ternary_qq_lane_base): Likewise.
1840 (ternary_intq_uintq_lane_def): Likewise.
1841 (ternary_intq_uintq_lane): New shape.
1842 (ternary_intq_uintq_opt_n_def): New class
1843 (ternary_intq_uintq_opt_n): New shape.
1844 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1845 (ternary_uintq_intq_def): New class.
1846 (ternary_uintq_intq): New shape.
1847 (ternary_uintq_intq_lane_def): New class.
1848 (ternary_uintq_intq_lane): New shape.
1849 (ternary_uintq_intq_opt_n_def): New class.
1850 (ternary_uintq_intq_opt_n): New shape.
1851 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1852 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1853 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1854 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1856 (svdotprod_lane_impl): ...this new class.
1857 (svmmla_impl, svusdot_impl): New classes.
1858 (svdot_lane): Update to use svdotprod_lane_impl.
1859 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1860 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1862 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1863 function, with no types defined.
1864 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1865 AARCH64_FL_I8MM functions.
1866 (svmmla): New AARCH64_FL_F32MM function.
1867 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1868 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1869 AARCH64_FL_F64MM function.
1870 (REQUIRED_EXTENSIONS):
1872 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1874 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1877 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1879 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1880 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1881 (*movti_internal): Do not check for
1882 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1883 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1884 just after check for TARGET_AVX.
1885 (*movdf_internal): Ditto.
1886 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1887 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1888 * config/i386/sse.md (mov<mode>_internal): Only check
1889 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1890 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1891 (<sse>_andnot<mode>3<mask_name>): Move check for
1892 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1893 (<code><mode>3<mask_name>): Ditto.
1894 (*andnot<mode>3): Ditto.
1895 (*andnottf3): Ditto.
1896 (*<code><mode>3): Ditto.
1897 (*<code>tf3): Ditto.
1898 (*andnot<VI:mode>3): Remove
1899 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1900 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1901 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1902 (sse4_1_blendv<ssemodesuffix>): Ditto.
1903 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1904 Explain that tune applies to 128bit instructions only.
1906 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1908 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1909 to definition of hsa_kernel_description. Parse assembly to find SGPR
1910 and VGPR count of kernel and store in hsa_kernel_description.
1912 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1914 PR rtl-optimization/91838
1915 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1916 to truncate if allowed or reject combination.
1918 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1920 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1921 (find_inv_vars_cb): Likewise.
1923 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1925 * calls.c (special_function_p): Split out the check for DECL_NAME
1926 being non-NULL and fndecl being extern at file scope into a
1927 new maybe_special_function_p and call it. Drop check for fndecl
1928 being non-NULL that was after a usage of DECL_NAME (fndecl).
1929 * tree.h (maybe_special_function_p): New inline function.
1931 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1933 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1934 (mask_gather_load<mode>): ... here, and zero-initialize the
1936 (maskload<mode>di): Zero-initialize the destination.
1939 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1942 * doc/analyzer.texi (Limitations): Note that constraints on
1943 floating-point values are currently ignored.
1945 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1948 * symtab.c (symtab_node::noninterposable_alias): If localalias
1949 already exists, but is not usable, append numbers after it until
1950 a unique name is found. Formatting fix.
1953 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1956 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1958 * config/gcn/gcn.c (print_operand): Handle LTGT.
1959 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1961 2020-01-30 Richard Biener <rguenther@suse.de>
1963 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1964 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1966 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1968 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1969 without a DECL in .data.rel.ro.local.
1971 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1974 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1978 * config/i386/sse.md
1979 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1980 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1981 any_extend code iterator instead of always zero_extend.
1982 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1983 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1984 Use any_extend code iterator instead of always zero_extend.
1985 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1986 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1987 Use any_extend code iterator instead of always zero_extend.
1988 (*sse2_pmovmskb_ext): New define_insn.
1989 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1992 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1993 (*popcountsi2_zext_falsedep): New define_insn.
1995 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1997 * config.in: Regenerated.
1998 * configure: Regenerated.
2000 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2003 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2004 LLVM's assembler changed the default in version 9.
2006 2020-01-24 Jeff Law <law@redhat.com>
2008 PR tree-optimization/89689
2009 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2011 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2015 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2017 PR rtl-optimization/87763
2018 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2019 simplification to handle subregs as well as bare regs.
2020 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2022 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2025 * ira.c (ira): Revert use of simplified LRA algorithm.
2027 2020-01-29 Martin Jambor <mjambor@suse.cz>
2029 PR tree-optimization/92706
2030 * tree-sra.c (struct access): Fields first_link, last_link,
2031 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2032 next_rhs_queued and grp_rhs_queued respectively, new fields
2033 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2034 (struct assign_link): Field next renamed to next_rhs, new field
2035 next_lhs. Updated comment.
2036 (work_queue_head): Renamed to rhs_work_queue_head.
2037 (lhs_work_queue_head): New variable.
2038 (add_link_to_lhs): New function.
2039 (relink_to_new_repr): Also relink LHS lists.
2040 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2041 (add_access_to_lhs_work_queue): New function.
2042 (pop_access_from_work_queue): Renamed to
2043 pop_access_from_rhs_work_queue.
2044 (pop_access_from_lhs_work_queue): New function.
2045 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2047 (child_would_conflict_in_lacc): Renamed to
2048 child_would_conflict_in_acc. Adjusted parameter names.
2049 (create_artificial_child_access): New parameter set_grp_read, use it.
2050 (subtree_mark_written_and_enqueue): Renamed to
2051 subtree_mark_written_and_rhs_enqueue.
2052 (propagate_subaccesses_across_link): Renamed to
2053 propagate_subaccesses_from_rhs.
2054 (propagate_subaccesses_from_lhs): New function.
2055 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2058 2020-01-29 Martin Jambor <mjambor@suse.cz>
2060 PR tree-optimization/92706
2061 * tree-sra.c (struct access): Adjust comment of
2062 grp_total_scalarization.
2063 (find_access_in_subtree): Look for single children spanning an entire
2065 (scalarizable_type_p): Allow register accesses, adjust callers.
2066 (completely_scalarize): Remove function.
2067 (scalarize_elem): Likewise.
2068 (create_total_scalarization_access): Likewise.
2069 (sort_and_splice_var_accesses): Do not track total scalarization
2071 (analyze_access_subtree): New parameter totally, adjust to new meaning
2072 of grp_total_scalarization.
2073 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2074 (can_totally_scalarize_forest_p): New function.
2075 (create_total_scalarization_access): Likewise.
2076 (create_total_access_and_reshape): Likewise.
2077 (total_should_skip_creating_access): Likewise.
2078 (totally_scalarize_subtree): Likewise.
2079 (analyze_all_variable_accesses): Perform total scalarization after
2080 subaccess propagation using the new functions above.
2081 (initialize_constant_pool_replacements): Output initializers by
2082 traversing the access tree.
2084 2020-01-29 Martin Jambor <mjambor@suse.cz>
2086 * tree-sra.c (verify_sra_access_forest): New function.
2087 (verify_all_sra_access_forests): Likewise.
2088 (create_artificial_child_access): Set parent.
2089 (analyze_all_variable_accesses): Call the verifier.
2091 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2093 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2094 if called on indirect edge.
2095 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2096 speculative call if needed.
2098 2020-01-29 Richard Biener <rguenther@suse.de>
2100 PR tree-optimization/93428
2101 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2102 permutation when the load node is created.
2103 (vect_analyze_slp_instance): Re-use it here.
2105 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2107 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2109 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2111 PR rtl-optimization/93272
2112 * ira-lives.c (process_out_of_region_eh_regs): New function.
2113 (process_bb_node_lives): Call it.
2115 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2117 * coverage.c (read_counts_file): Make error message lowercase.
2119 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2121 * profile-count.c (profile_quality_display_names): Fix ordering.
2123 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2126 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2127 hash only when edge is first within the sequence.
2128 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2129 (symbol_table::create_edge): Do not set target_prob.
2130 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2132 (cgraph_edge::make_speculative): Drop target_prob parameter.
2133 (cgraph_edge::speculative_call_info): Remove.
2134 (cgraph_edge::first_speculative_call_target): New member function.
2135 (update_call_stmt_hash_for_removing_direct_edge): New function.
2136 (cgraph_edge::resolve_speculation): Rewrite to new API.
2137 (cgraph_edge::speculative_call_for_target): New member function.
2138 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2139 multiple speculation targets.
2140 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2142 (verify_speculative_call): Verify that targets form an interval.
2143 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2144 (cgraph_edge::first_speculative_call_target): New member function.
2145 (cgraph_edge::next_speculative_call_target): New member function.
2146 (cgraph_edge::speculative_call_target_ref): New member function.
2147 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2148 (cgraph_edge): Remove target_prob.
2149 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2150 Fix handling of speculative calls.
2151 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2152 * ipa-fnsummary.c (analyze_function_body): Likewise.
2153 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2154 * ipa-profile.c (dump_histogram): Fix formating.
2155 (ipa_profile_generate_summary): Watch for overflows.
2156 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2157 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2158 (update_indirect_edges_after_inlining): Update to new API.
2159 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2161 * profile-count.h: (profile_probability::adjusted): New.
2162 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2163 updating of profile.
2164 * value-prof.c (gimple_ic_transform): Rename to ...
2165 (dump_ic_profile): ... this one; update dumping.
2166 (stream_in_histogram_value): Fix formating.
2167 (gimple_value_profile_transformations): Update.
2169 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2172 * config/i386/i386.md (*movoi_internal_avx): Remove
2173 TARGET_SSE_TYPELESS_STORES check.
2174 (*movti_internal): Prefer TARGET_AVX over
2175 TARGET_SSE_TYPELESS_STORES.
2176 (*movtf_internal): Likewise.
2177 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2178 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2179 from TARGET_SSE_TYPELESS_STORES.
2181 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2183 * diagnostic-core.h (warning_at): Rename overload to...
2184 (warning_meta): ...this.
2185 (emit_diagnostic_valist): Delete decl of overload taking
2186 diagnostic_metadata.
2187 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2188 (warning_at): Rename overload taking diagnostic_metadata to...
2189 (warning_meta): ...this.
2191 2020-01-28 Richard Biener <rguenther@suse.de>
2193 PR tree-optimization/93439
2194 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2195 * tree-cfg.c (move_sese_region_to_fn): ... here.
2196 (verify_types_in_gimple_reference): Verify used cliques are
2199 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2202 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2203 argument of a pointer to struct gcc_options and pass it to
2204 parse_mtune_ctrl_str.
2205 (ix86_function_specific_restore): Pass opts to
2206 set_ix86_tune_features.
2207 (ix86_option_override_internal): Likewise.
2208 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2209 gcc_options and use it for x_ix86_tune_ctrl_string.
2211 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2213 PR rtl-optimization/87763
2214 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2215 simplification to handle subregs as well as bare regs.
2216 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2218 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2220 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2221 for reduction chains that (now) include a call.
2223 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2225 PR tree-optimization/92822
2226 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2227 out the don't-care elements of a vector whose significant elements
2228 are duplicates, make the don't-care elements duplicates too.
2230 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2232 PR tree-optimization/93434
2233 * tree-predcom.c (split_data_refs_to_components): Record which
2234 components have had aliasing loads removed. Prevent store-store
2235 commoning for all such components.
2237 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2240 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2241 -1 or is_vshift is true, use new_vector with number of elts npatterns
2242 rather than new_unary_operation.
2244 PR tree-optimization/93454
2245 * gimple-fold.c (fold_array_ctor_reference): Perform
2246 elt_size.to_uhwi () just once, instead of calling it in every
2247 iteration. Punt if that value is above size of the temporary
2248 buffer. Decrease third native_encode_expr argument when
2249 bufoff + elt_sz is above size of buf.
2251 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2253 * config/mips/mips.c (mips_declare_object_name)
2254 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2256 2020-01-27 Martin Liska <mliska@suse.cz>
2258 PR gcov-profile/93403
2259 * tree-profile.c (gimple_init_gcov_profiler): Generate
2260 both __gcov_indirect_call_profiler_v4 and
2261 __gcov_indirect_call_profiler_v4_atomic.
2263 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2266 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2268 (@aarch64_split_simd_mov<mode>): Use it.
2269 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2270 Leave the vec_extract patterns to handle 2-element vectors.
2271 (aarch64_simd_mov_from_<mode>high): Likewise.
2272 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2273 (vec_extractv2dfv1df): Likewise.
2275 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2277 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2278 jump conditions for *compare_condjump<GPI:mode>.
2280 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2283 * digraph.cc (test_edge::test_edge): Specify template for base
2286 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2288 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2290 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2292 * config/arc/arc-protos.h (gen_mlo): Remove.
2293 (gen_mhi): Likewise.
2294 * config/arc/arc.c (AUX_MULHI): Define.
2295 (arc_must_save_reister): Special handling for r58/59.
2296 (arc_compute_frame_size): Consider mlo/mhi registers.
2297 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2299 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2300 mlo/mhi name selection.
2301 (arc_restore_callee_saves): Don't early restore blink when ISR.
2302 (arc_expand_prologue): Add mlo/mhi saving.
2303 (arc_expand_epilogue): Add mlo/mhi restoring.
2306 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2307 numbering when MUL64 option is used.
2308 (DWARF2_FRAME_REG_OUT): Define.
2309 * config/arc/arc.md (arc600_stall): New pattern.
2310 (VUNSPEC_ARC_ARC600_STALL): Define.
2311 (mulsi64): Use correct mlo/mhi registers.
2312 (mulsi_600): Clean it up.
2313 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2315 (mhi_operand): Likewise.
2317 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2318 Petro Karashchenko <petro.karashchenko@ring.com>
2320 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2321 attributes if needed.
2322 (prepare_move_operands): Generate special unspec instruction for
2324 (arc_isuncached_mem_p): Propagate uncached attribute to each
2326 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2327 (VUNSPEC_ARC_STDI): Likewise.
2328 (ALLI): New mode iterator.
2329 (mALLI): New mode attribute.
2330 (lddi): New instruction pattern.
2332 (stdidi_split): Split instruction for architectures which are not
2333 supporting ll64 option.
2334 (lddidi_split): Likewise.
2336 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2338 PR rtl-optimization/92989
2339 * lra-lives.c (process_bb_lives): Update the live-in set before
2340 processing additional clobbers.
2342 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2344 PR rtl-optimization/93170
2345 * cselib.c (cselib_invalidate_regno_val): New function, split out
2347 (cselib_invalidate_regno): ...here.
2348 (cselib_invalidated_by_call_p): New function.
2349 (cselib_process_insn): Iterate over all the hard-register entries in
2350 REG_VALUES and invalidate any that cross call-clobbered registers.
2352 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2354 * dojump.c (split_comparison): Use HONOR_NANS rather than
2355 HONOR_SNANS when splitting LTGT.
2357 2020-01-27 Martin Liska <mliska@suse.cz>
2360 * opts.c (print_filtered_help): Exclude language-specific
2361 options from --help=common unless enabled in all FEs.
2363 2020-01-27 Martin Liska <mliska@suse.cz>
2365 * opts.c (print_help): Exclude params from
2366 all except --help=param.
2368 2020-01-27 Martin Liska <mliska@suse.cz>
2371 * config/i386/i386-features.c (make_resolver_func):
2372 Align the code with ppc64 target implementation.
2373 Do not generate a unique name for resolver function.
2375 2020-01-27 Richard Biener <rguenther@suse.de>
2377 PR tree-optimization/93397
2378 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2379 converted reduction chain SLP graph adjustment.
2381 2020-01-26 Marek Polacek <polacek@redhat.com>
2384 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2387 2020-01-26 Jason Merrill <jason@redhat.com>
2390 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2393 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2395 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2396 (rx_setmem): Likewise.
2398 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2401 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2402 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2403 drop <di> from constraint of last operand.
2406 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2407 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2408 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2410 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2413 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2416 2020-01-24 Jeff Law <law@redhat.com>
2418 PR tree-optimization/92788
2419 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2422 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2425 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2426 *avx_vperm_broadcast_<mode>,
2427 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2428 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2429 Move before avx2_perm<mode>/avx512f_perm<mode>.
2432 * simplify-rtx.c (simplify_const_unary_operation,
2433 simplify_const_binary_operation): Punt for mode precision above
2434 MAX_BITSIZE_MODE_ANY_INT.
2436 2020-01-24 Andrew Pinski <apinski@marvell.com>
2438 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2441 2020-01-24 Jeff Law <law@redhat.com>
2444 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2445 for REGs. Call output_operand_lossage to get more reasonable
2448 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2450 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2451 gcn_fp_compare_operator.
2452 (vec_cmpu<mode>di): Use gcn_compare_operator.
2453 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2454 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2455 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2456 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2457 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2458 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2459 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2460 gcn_fp_compare_operator.
2461 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2462 gcn_fp_compare_operator.
2463 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2464 gcn_fp_compare_operator.
2465 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2466 gcn_fp_compare_operator.
2468 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2470 * doc/install.texi (Cross-Compiler-Specific Options): Document
2471 `--with-toolexeclibdir' option.
2473 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2475 * target.def (flags_regnum): Also mention effect on delay slot filling.
2476 * doc/tm.texi: Regenerate.
2478 2020-01-23 Jeff Law <law@redhat.com>
2480 PR translation/90162
2481 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2483 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2486 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2489 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2491 PR rtl-optimization/93402
2492 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2495 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2497 * config.in: Regenerated.
2498 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2499 for TARGET_LIBC_GNUSTACK.
2500 * configure: Regenerated.
2501 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2502 found to be 2.31 or greater.
2504 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2506 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2508 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2509 (mips_asm_file_end): New function. Delegate to
2510 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2511 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2513 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2516 * config/i386/i386-modes.def (POImode): New mode.
2517 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2518 * config/i386/i386.md (DPWI): New mode attribute.
2519 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2521 (QPWI): ... this. Use POI instead of OI for TImode.
2522 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2523 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2526 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2529 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2531 (speculation_tracker_rev): New pattern.
2532 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2533 Use speculation_tracker_rev to track the inverse condition.
2535 2020-01-23 Richard Biener <rguenther@suse.de>
2537 PR tree-optimization/93381
2538 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2539 alias-set of the def as argument and record the first one.
2540 (vn_walk_cb_data::first_set): New member.
2541 (vn_reference_lookup_3): Pass the alias-set of the current def
2542 to push_partial_def. Fix alias-set used in the aggregate copy
2544 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2545 * real.c (clear_significand_below): Fix out-of-bound access.
2547 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2550 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2551 New define_insn patterns.
2553 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2555 * doc/sourcebuild.texi (check-function-bodies): Add an
2556 optional target/xfail selector.
2558 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2560 PR rtl-optimization/93124
2561 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2562 bare USE and CLOBBER insns.
2564 2020-01-22 Andrew Pinski <apinski@marvell.com>
2566 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2568 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2571 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2572 diagnostic_manager into "ana" namespace.
2573 * selftest-run-tests.c (selftest::run_tests): Update for move of
2574 selftest::run_analyzer_selftests to
2575 ana::selftest::run_analyzer_selftests.
2577 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2579 * cfgexpand.c (union_stack_vars): Update the size.
2581 2020-01-22 Richard Biener <rguenther@suse.de>
2583 PR tree-optimization/93381
2584 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2585 throughout, handle all conversions the same.
2587 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2590 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2591 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2592 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2593 Call force_reg on high_in2 unconditionally.
2595 2020-01-22 Martin Liska <mliska@suse.cz>
2597 PR tree-optimization/92924
2598 * profile.c (compute_value_histograms): Divide
2601 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2604 * output.h (assemble_name_resolve): Declare.
2605 * varasm.c (assemble_name_resolve): New function.
2606 (assemble_name): Use it.
2607 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2609 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2611 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2612 update_web_docs_git instead of update_web_docs_svn.
2614 2020-01-21 Andrew Pinski <apinski@marvell.com>
2617 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2618 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2619 (*tlsgd_small_<mode>): Likewise.
2620 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2621 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2622 register. Convert that register back to dest using convert_mode.
2624 2020-01-21 Jim Wilson <jimw@sifive.com>
2626 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2629 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2630 Uros Bizjak <ubizjak@gmail.com>
2633 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2635 (legitimize_tls_address): Do GNU2 TLS address computation in
2636 ptr_mode and zero-extend result to Pmode.
2637 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2638 :P with :PTR and Pmode with ptr_mode.
2639 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2640 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2641 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2643 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2646 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2647 the last two operands are CONST_INT_P before using them as such.
2649 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2651 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2652 to get the integer element types.
2654 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2656 * config/aarch64/aarch64-sve-builtins.h
2657 (function_expander::convert_to_pmode): Declare.
2658 * config/aarch64/aarch64-sve-builtins.cc
2659 (function_expander::convert_to_pmode): New function.
2660 (function_expander::get_contiguous_base): Use it.
2661 (function_expander::prepare_gather_address_operands): Likewise.
2662 * config/aarch64/aarch64-sve-builtins-sve2.cc
2663 (svwhilerw_svwhilewr_impl::expand): Likewise.
2665 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2668 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2669 cfun->machine->label_is_assembled.
2670 (aarch64_print_patchable_function_entry): New.
2671 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2672 * config/aarch64/aarch64.h (struct machine_function): New field,
2675 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2678 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2681 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2684 * cgraph.c (cgraph_edge::resolve_speculation,
2685 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2686 call_stmt_site_hash.
2688 2020-01-21 Martin Liska <mliska@suse.cz>
2690 * config/rs6000/rs6000.c (common_mode_defined): Remove
2693 2020-01-21 Richard Biener <rguenther@suse.de>
2695 PR tree-optimization/92328
2696 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2697 type when value-numbering same-sized store by inserting a
2699 (eliminate_dom_walker::eliminate_stmt): When eliminating
2700 a redundant store handle bit-reinterpretation of the same value.
2702 2020-01-21 Andrew Pinski <apinski@marvel.com>
2705 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2707 (prepare_block_for_update): This. Use a worklist instead of
2710 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2712 * gcc/config/arm/arm.c (clear_operation_p):
2713 Initialise last_regno, skip first iteration
2714 based on the first_set value and use ints instead
2715 of the unnecessary HOST_WIDE_INTs.
2717 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2720 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2721 compare_mode other than SFmode or DFmode.
2723 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2726 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2727 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2728 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2730 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2732 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2734 2020-01-20 Andrew Pinski <apinski@marvell.com>
2737 * targhooks.c (default_print_patchable_function_entry): Use
2738 output_asm_insn to emit the nop instruction.
2740 2020-01-20 Fangrui Song <maskray@google.com>
2743 * targhooks.c (default_print_patchable_function_entry): Align to
2746 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2749 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2750 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2751 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2752 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2753 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2754 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2755 Remove the {q} suffix from lea.
2756 (*tls_dynamic_gnu2_call_64): Renamed to ...
2757 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2758 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2759 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2760 Pass Pmode to gen_tls_dynamic_gnu2_64.
2762 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2764 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2766 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2768 * config/aarch64/aarch64-sve-builtins-base.cc
2769 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2771 2020-01-20 Richard Biener <rguenther@suse.de>
2774 * dwarf2out.c (prune_unused_types): Unconditionally mark
2775 called function DIEs.
2777 2020-01-20 Martin Liska <mliska@suse.cz>
2779 PR tree-optimization/93199
2780 * tree-eh.c (struct leh_state): Add
2781 new field outer_non_cleanup.
2782 (cleanup_is_dead_in): Pass leh_state instead
2783 of eh_region. Add a checking that state->outer_non_cleanup
2784 points to outer non-clean up region.
2785 (lower_try_finally): Record outer_non_cleanup
2787 (lower_catch): Likewise.
2788 (lower_eh_filter): Likewise.
2789 (lower_eh_must_not_throw): Likewise.
2790 (lower_cleanup): Likewise.
2792 2020-01-20 Richard Biener <rguenther@suse.de>
2794 PR tree-optimization/93094
2795 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2796 (vect_transform_loop): Likewise.
2797 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2798 loop_vectorized_call to vect_transform_loop.
2799 * tree-vect-loop.c (vect_transform_loop): Pass down
2800 loop_vectorized_call to vect_loop_versioning.
2801 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2802 the earlier discovered loop_vectorized_call.
2804 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2806 * doc/contribute.texi: Update for SVN -> Git transition.
2807 * doc/install.texi: Likewise.
2809 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2812 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2813 speculative targets.
2814 (verify_speculative_call): New function
2815 (cgraph_node::verify_node): Use it.
2816 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2819 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2822 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2823 (cgraph_edge::make_direct): Remove all indirect targets.
2824 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2825 (cgraph_node::verify_node): Verify that only one call_stmt or
2826 lto_stmt_uid is set.
2827 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2829 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2830 (lto_output_ref): Simplify streaming of stmt.
2831 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2833 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2835 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2836 Mark parameter unused.
2838 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2840 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2842 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2844 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2846 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2848 * Makefile.in: Add coroutine-passes.o.
2849 * builtin-types.def (BT_CONST_SIZE): New.
2850 (BT_FN_BOOL_PTR): New.
2851 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2852 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2853 * coroutine-builtins.def: New file.
2854 * coroutine-passes.cc: New file.
2855 * function.h (struct GTY function): Add a bit to indicate that the
2856 function is a coroutine component.
2857 * internal-fn.c (expand_CO_FRAME): New.
2858 (expand_CO_YIELD): New.
2859 (expand_CO_SUSPN): New.
2860 (expand_CO_ACTOR): New.
2861 * internal-fn.def (CO_ACTOR): New.
2865 * passes.def: Add pass_coroutine_lower_builtins,
2866 pass_coroutine_early_expand_ifns.
2867 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2868 (make_pass_coroutine_early_expand_ifns): New.
2869 * doc/invoke.texi: Document the fcoroutines command line
2872 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2874 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2877 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2878 after checking the argument is a REG. Don't use REGNO (reg)
2879 again to set last_regno, reuse regno variable instead.
2881 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2883 * doc/analyzer.texi (Limitations): Add note about NaN.
2885 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2886 Sudakshina Das <sudi.das@arm.com>
2888 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2889 and valid immediate.
2890 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2891 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2892 * config/arm/constraints.md (Pg): New.
2893 * config/arm/predicates.md (long_shift_imm): New.
2894 (arm_reg_or_long_shift_imm): Likewise.
2895 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2896 (thumb2_lsll): Likewise.
2899 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2900 Sudakshina Das <sudi.das@arm.com>
2902 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2903 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2904 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2905 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2906 * config/arm/thumb2.md (thumb2_asrl): New.
2907 (thumb2_lsll): Likewise.
2909 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2911 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2914 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2916 * gdbinit.in (help-gcc-hooks): New command.
2917 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2918 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2921 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2923 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2924 correct target macro.
2926 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2928 * config/aarch64/aarch64-protos.h
2929 (aarch64_sve_ld1ro_operand_p): New.
2930 * config/aarch64/aarch64-sve-builtins-base.cc
2931 (class load_replicate): New.
2932 (class svld1ro_impl): New.
2933 (class svld1rq_impl): Change to inherit from load_replicate.
2934 (svld1ro): New sve intrinsic function base.
2935 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2936 New DEF_SVE_FUNCTION.
2937 * config/aarch64/aarch64-sve-builtins-base.h
2938 (svld1ro): New decl.
2939 * config/aarch64/aarch64-sve-builtins.cc
2940 (function_expander::add_mem_operand): Modify assert to allow
2942 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2944 * config/aarch64/aarch64.c
2945 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2946 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2947 (aarch64_sve_ld1ro_operand_p): New.
2948 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2949 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2950 * config/aarch64/predicates.md
2951 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2953 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2955 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2956 Introduce this ACLE specified predefined macro.
2957 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2958 (fp): Disabling this disables f64mm.
2959 (simd): Disabling this disables f64mm.
2960 (fp16): Disabling this disables f64mm.
2961 (sve): Disabling this disables f64mm.
2962 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2963 (AARCH64_ISA_F64MM): New.
2964 (TARGET_F64MM): New.
2965 * doc/invoke.texi (f64mm): Document new option.
2967 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2969 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2970 (neoversen1_tunings): Likewise.
2972 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2975 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2976 Add assert to ensure prolog has been emitted.
2977 (aarch64_split_atomic_op): Likewise.
2978 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2979 Use epilogue_completed rather than reload_completed.
2980 (aarch64_atomic_exchange<mode>): Likewise.
2981 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2982 (atomic_nand<mode>): Likewise.
2983 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2984 (atomic_fetch_nand<mode>): Likewise.
2985 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2986 (atomic_nand_fetch<mode>): Likewise.
2988 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2991 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2993 (REVERSE_CONDITION): Delete.
2994 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2995 (CCFP_CCFPE): Likewise.
2996 (e): New mode attribute.
2997 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2998 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2999 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3000 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3001 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3002 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3003 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3004 name of generator from gen_ccmpdi to gen_ccmpccdi.
3005 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3006 the previous comparison but aren't able to, use the new ccmp_rev
3009 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3011 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3012 than testing directly for INTEGER_CST.
3013 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3015 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3017 PR tree-optimization/93292
3018 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3019 get_vectype_for_scalar_type returns NULL.
3021 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3023 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3024 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3026 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3028 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3030 * params.opt: (max-predicted-iterations): Set bounds.
3031 * predict.c (real_almost_one, real_br_prob_base,
3032 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3033 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3034 probabilities; do not truncate to reg_br_prob_bases.
3035 (estimate_loops_at_level): Pass max_cyclic_prob.
3036 (estimate_loops): Compute max_cyclic_prob.
3037 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3039 * profile-count.c (profile_probability::to_sreal): New.
3040 * profile-count.h (class sreal): Move up in file.
3041 (profile_probability::to_sreal): Declare.
3043 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3046 (arm_invalid_conversion): New function for target hook.
3047 (arm_invalid_unary_op): New function for target hook.
3048 (arm_invalid_binary_op): New function for target hook.
3050 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3052 * config.gcc: Add arm_bf16.h.
3053 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3054 (arm_simd_builtin_std_type): Add BFmode.
3055 (arm_init_simd_builtin_types): Define element types for vector types.
3056 (arm_init_bf16_types): New function.
3057 (arm_init_builtins): Add arm_init_bf16_types function call.
3058 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3059 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3060 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3061 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3062 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3063 (arm_mangle_type): Add __bf16.
3064 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3065 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3066 arm_bf16_ptr_type_node.
3067 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3068 define_split between ARM registers.
3069 * config/arm/arm_bf16.h: New file.
3070 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3071 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3073 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3074 * config/arm/vfp.md: Add BFmode to movhf patterns.
3076 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3077 Andre Vieira <andre.simoesdiasvieira@arm.com>
3079 * config/arm/arm-cpus.in (mve, mve_float): New features.
3080 (dsp, mve, mve.fp): New options.
3081 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3082 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3083 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3085 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3086 Thomas Preud'homme <thomas.preudhomme@arm.com>
3088 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3090 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3091 error for using -mcmse when targeting Armv8.1-M Mainline.
3093 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3094 Thomas Preud'homme <thomas.preudhomme@arm.com>
3096 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3097 address in r4 when targeting Armv8.1-M Mainline.
3098 (nonsecure_call_value_internal): Likewise.
3099 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3100 a register match_operand again. Emit BLXNS when targeting
3102 (nonsecure_call_value_reg_thumb2): Likewise.
3104 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3105 Thomas Preud'homme <thomas.preudhomme@arm.com>
3107 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3108 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3109 variable as true when floating-point ABI is not hard. Replace
3110 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3111 Generate VLSTM and VLLDM instruction respectively before and
3112 after a function call to cmse_nonsecure_call function.
3113 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3114 (VUNSPEC_VLLDM): Likewise.
3115 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3116 (lazy_load_multiple_insn): Likewise.
3118 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3119 Thomas Preud'homme <thomas.preudhomme@arm.com>
3121 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3122 (arm_emit_vfp_multi_reg_pop): Likewise.
3123 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3124 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3125 restore callee-saved VFP registers.
3127 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3128 Thomas Preud'homme <thomas.preudhomme@arm.com>
3130 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3131 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3132 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3133 callee-saved GPRs as well as clear ip register before doing a nonsecure
3134 call then restore callee-saved GPRs after it when targeting
3136 (arm_reorg): Adapt to function rename.
3138 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3139 Thomas Preud'homme <thomas.preudhomme@arm.com>
3141 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3142 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3143 clear_vfp_multiple pattern based on a new vfp parameter.
3144 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3145 targeting Armv8.1-M Mainline.
3146 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3147 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3148 whether VFP registers are available before looking call_used_regs for a
3150 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3151 of prototype of clear_operation_p.
3152 (clear_vfp_multiple_operation): New predicate.
3153 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3154 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3156 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3157 Thomas Preud'homme <thomas.preudhomme@arm.com>
3159 * config/arm/arm-protos.h (clear_operation_p): Declare.
3160 * config/arm/arm.c (clear_operation_p): New function.
3161 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3162 targeting Armv8.1-M Mainline or successor.
3163 (output_return_instruction): Only output APSR register clearing if
3164 Armv8.1-M Mainline instructions not available.
3165 (thumb_exit): Likewise.
3166 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3167 * config/arm/thumb2.md (clear_apsr): New define_insn.
3168 (clear_multiple): Likewise.
3169 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3171 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3172 Thomas Preud'homme <thomas.preudhomme@arm.com>
3174 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3175 (use_return_insn): Also return false for Armv8.1-M Mainline.
3176 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3177 Mainline instructions are available.
3178 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3179 when targeting Armv8.1-M Mainline Security Extensions.
3180 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3181 Mainline entry function.
3182 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3183 targeting Armv8.1-M Mainline or successor.
3184 (arm_expand_epilogue): Fix indentation of caller-saved register
3185 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3187 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3188 (FP_SYSREGS): Likewise.
3189 (enum vfp_sysregs_encoding): Define enum.
3190 (fp_sysreg_names): Declare.
3191 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3192 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3193 (pop_fpsysreg_insn): Likewise.
3195 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3196 Thomas Preud'homme <thomas.preudhomme@arm.com>
3198 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3199 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3200 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3201 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3202 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3203 (ARMv8_1m_main): New feature group.
3204 (armv8.1-m.main): New architecture.
3205 * config/arm/arm-tables.opt: Regenerate.
3206 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3207 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3208 (arm_options_perform_arch_sanity_checks): Error out when targeting
3209 Armv8.1-M Mainline Security Extensions.
3210 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3212 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3214 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3215 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3216 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3217 aarch64_bfdot_laneq): New.
3218 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3219 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3220 vbfdotq_laneq_f32): New.
3221 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3222 VBFMLA_W, VBF): New.
3223 (isquadop): Add V4BF, V8BF.
3225 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3227 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3228 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3229 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3230 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3231 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3232 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3233 usdot_laneq, sudot_lane,sudot_laneq): New.
3234 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3235 (aarch64_<sur>dot_lane): New.
3236 * config/aarch64/arm_neon.h (vusdot_s32): New.
3238 (vusdot_lane_s32): New.
3239 (vsudot_lane_s32): New.
3240 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3241 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3243 2020-01-16 Martin Liska <mliska@suse.cz>
3245 * value-prof.c (dump_histogram_value): Fix
3246 obvious spacing issue.
3248 2020-01-16 Andrew Pinski <apinski@marvell.com>
3250 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3251 !storage_order_barrier_p.
3253 2020-01-16 Andrew Pinski <apinski@marvell.com>
3255 * sched-int.h (_dep): Add unused bit-field field for the padding.
3256 * sched-deps.c (init_dep_1): Init unused field.
3258 2020-01-16 Andrew Pinski <apinski@marvell.com>
3260 * optabs.h (create_expand_operand): Initialize target field also.
3262 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3264 PR tree-optimization/92429
3265 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3266 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3268 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3271 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3273 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3274 aarch64_sve_int_mode to each mode.
3276 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3278 * doc/analyzer.texi (Overview): Add note about
3279 -fdump-ipa-analyzer.
3281 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3283 PR tree-optimization/93231
3284 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3285 input_type is unsigned. Use tree_to_shwi for shift constant.
3286 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3287 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3288 inputs more efficiently.
3290 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3292 * config/i386/i386.md (*movsf_internal): Do not require
3293 SSE2 ISA for alternatives 14 and 15.
3295 2020-01-15 Richard Biener <rguenther@suse.de>
3298 * tree-eh.c (sink_clobbers): If we already visited the destination
3299 block do not defer insertion.
3300 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3301 the purpose of defered insertion.
3303 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3305 * BASE-VER: Bump to 10.0.1.
3307 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3309 PR tree-optimization/93247
3310 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3311 type of the stmt that we're going to vectorize.
3313 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3315 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3316 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3319 2020-01-15 Martin Liska <mliska@suse.cz>
3321 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3322 2 calls of streamer_read_hwi in a function call.
3324 2020-01-15 Richard Biener <rguenther@suse.de>
3326 * alias.c (record_alias_subset): Avoid redundant work when
3327 subset is already recorded.
3329 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3331 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3332 the analyzer options provide CWE identifiers.
3334 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3336 * tree-diagnostic-path.cc (path_summary::event_range::print):
3337 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3338 using get_pure_location.
3340 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3342 PR tree-optimization/93262
3343 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3344 perform head trimming only if the last argument is constant,
3345 either all ones, or larger or equal to head trim, in the latter
3346 case decrease the last argument by head_trim.
3348 PR tree-optimization/93249
3349 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3350 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3351 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3352 perform head trim unless we can prove there are no '\0' chars
3353 from the source among the first head_trim chars.
3355 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3357 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3359 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3362 * config/i386/sse.md
3363 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3364 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3365 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3366 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3367 just a single alternative instead of two, make operands 1 and 2
3370 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3373 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3376 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3378 * Makefile.in (lang_opt_files): Add analyzer.opt.
3379 (ANALYZER_OBJS): New.
3380 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3381 tristate.o and ANALYZER_OBJS.
3382 (TEXI_GCCINT_FILES): Add analyzer.texi.
3383 * common.opt (-fanalyzer): New driver option.
3384 * config.in: Regenerate.
3385 * configure: Regenerate.
3386 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3387 (gccdepdir): Also create depdir for "analyzer" subdir.
3388 * digraph.cc: New file.
3389 * digraph.h: New file.
3390 * doc/analyzer.texi: New file.
3391 * doc/gccint.texi ("Static Analyzer") New menu item.
3392 (analyzer.texi): Include it.
3393 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3394 ("Warning Options"): Add static analysis warnings to the list.
3395 (-Wno-analyzer-double-fclose): New option.
3396 (-Wno-analyzer-double-free): New option.
3397 (-Wno-analyzer-exposure-through-output-file): New option.
3398 (-Wno-analyzer-file-leak): New option.
3399 (-Wno-analyzer-free-of-non-heap): New option.
3400 (-Wno-analyzer-malloc-leak): New option.
3401 (-Wno-analyzer-possible-null-argument): New option.
3402 (-Wno-analyzer-possible-null-dereference): New option.
3403 (-Wno-analyzer-null-argument): New option.
3404 (-Wno-analyzer-null-dereference): New option.
3405 (-Wno-analyzer-stale-setjmp-buffer): New option.
3406 (-Wno-analyzer-tainted-array-index): New option.
3407 (-Wno-analyzer-use-after-free): New option.
3408 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3409 (-Wno-analyzer-use-of-uninitialized-value): New option.
3410 (-Wanalyzer-too-complex): New option.
3411 (-fanalyzer-call-summaries): New warning.
3412 (-fanalyzer-checker=): New warning.
3413 (-fanalyzer-fine-grained): New warning.
3414 (-fno-analyzer-state-merge): New warning.
3415 (-fno-analyzer-state-purge): New warning.
3416 (-fanalyzer-transitivity): New warning.
3417 (-fanalyzer-verbose-edges): New warning.
3418 (-fanalyzer-verbose-state-changes): New warning.
3419 (-fanalyzer-verbosity=): New warning.
3420 (-fdump-analyzer): New warning.
3421 (-fdump-analyzer-callgraph): New warning.
3422 (-fdump-analyzer-exploded-graph): New warning.
3423 (-fdump-analyzer-exploded-nodes): New warning.
3424 (-fdump-analyzer-exploded-nodes-2): New warning.
3425 (-fdump-analyzer-exploded-nodes-3): New warning.
3426 (-fdump-analyzer-supergraph): New warning.
3427 * doc/sourcebuild.texi (dg-require-dot): New.
3428 (dg-check-dot): New.
3429 * gdbinit.in (break-on-saved-diagnostic): New command.
3430 * graphviz.cc: New file.
3431 * graphviz.h: New file.
3432 * ordered-hash-map-tests.cc: New file.
3433 * ordered-hash-map.h: New file.
3434 * passes.def (pass_analyzer): Add before
3435 pass_ipa_whole_program_visibility.
3436 * selftest-run-tests.c (selftest::run_tests): Call
3437 selftest::ordered_hash_map_tests_cc_tests.
3438 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3440 * shortest-paths.h: New file.
3441 * timevar.def (TV_ANALYZER): New timevar.
3442 (TV_ANALYZER_SUPERGRAPH): Likewise.
3443 (TV_ANALYZER_STATE_PURGE): Likewise.
3444 (TV_ANALYZER_PLAN): Likewise.
3445 (TV_ANALYZER_SCC): Likewise.
3446 (TV_ANALYZER_WORKLIST): Likewise.
3447 (TV_ANALYZER_DUMP): Likewise.
3448 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3449 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3450 * tree-pass.h (make_pass_analyzer): New decl.
3451 * tristate.cc: New file.
3452 * tristate.h: New file.
3454 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3457 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3458 alternatives 9 and 10.
3460 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3462 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3463 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3464 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3465 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3466 (selftest::hash_map_tests_c_tests): Call it.
3467 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3468 New static constant, using the value of = H::empty_zero_p.
3469 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3470 from default_hash_traits <Value>.
3471 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3473 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3474 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3475 calls to mark_empty with !Descriptor::empty_zero_p.
3476 (hash_table::empty_slow): Conditionalize the memset call with a
3477 check that Descriptor::empty_zero_p; otherwise, loop through the
3478 entries calling mark_empty on them.
3479 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3480 (pointer_hash::empty_zero_p): Likewise.
3481 (pair_hash::empty_zero_p): Likewise.
3482 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3484 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3485 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3486 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3487 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3488 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3489 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3490 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3491 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3493 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3496 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3498 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3499 fix typo on return value.
3501 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3504 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3506 (cgraph_edge::make_speculative): Add param for setting speculative_id
3508 (cgraph_edge::speculative_call_info): Update comments and find reference
3509 by speculative_id for multiple indirect targets.
3510 (cgraph_edge::resolve_speculation): Decrease the speculations
3511 for indirect edge, drop it's speculative if not direct target
3512 left. Update comments.
3513 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3514 (cgraph_node::dump): Print num_speculative_call_targets.
3515 (cgraph_node::verify_node): Don't report error if speculative
3516 edge not include statement.
3517 (cgraph_edge::num_speculative_call_targets_p): New function.
3518 * cgraph.h (int common_target_id): Remove.
3519 (int common_target_probability): Remove.
3520 (num_speculative_call_targets): New variable.
3521 (make_speculative): Add param for setting speculative_id.
3522 (cgraph_edge::num_speculative_call_targets_p): New declare.
3523 (target_prob): New variable.
3524 (speculative_id): New variable.
3525 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3526 call summaries for multiple speculative call targets.
3527 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3528 * ipa-profile.c (struct speculative_call_target): New struct.
3529 (class speculative_call_summary): New class.
3530 (class speculative_call_summaries): New class.
3531 (call_sums): New variable.
3532 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3533 (ipa_profile_write_edge_summary): New function.
3534 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3535 (ipa_profile_dump_all_summaries): New function.
3536 (ipa_profile_read_edge_summary): New function.
3537 (ipa_profile_read_summary_section): New function.
3538 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3539 (ipa_profile): Generate num_speculative_call_targets from
3541 * ipa-ref.h (speculative_id): New variable.
3542 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3543 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3544 common_target_probability. Stream out speculative_id and
3545 num_speculative_call_targets.
3546 (input_edge): Likewise.
3547 * predict.c (dump_prediction): Remove edges count assert to be
3549 * symtab.c (symtab_node::create_reference): Init speculative_id.
3550 (symtab_node::clone_references): Clone speculative_id.
3551 (symtab_node::clone_referring): Clone speculative_id.
3552 (symtab_node::clone_reference): Clone speculative_id.
3553 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3554 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3555 if indirect call contains multiple speculative targets.
3556 * value-prof.h (check_ic_target): Remove.
3557 * value-prof.c (gimple_value_profile_transformations):
3558 Use void function gimple_ic_transform.
3559 * value-prof.c (gimple_ic_transform): Handle topn case.
3560 Fix comment typos. Change it to a void function.
3562 2020-01-13 Andrew Pinski <apinski@marvell.com>
3564 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3565 (octeontx2t98): New define.
3566 (octeontx2t96): New define.
3567 (octeontx2t93): New define.
3568 (octeontx2f95): New define.
3569 (octeontx2f95n): New define.
3570 (octeontx2f95mm): New define.
3571 * config/aarch64/aarch64-tune.md: Regenerate.
3572 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3574 2020-01-13 Jason Merrill <jason@redhat.com>
3576 PR c++/33799 - destroy return value if local cleanup throws.
3577 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3579 2020-01-13 Martin Liska <mliska@suse.cz>
3581 * ipa-cp.c (get_max_overall_size): Use newly
3582 renamed param param_ipa_cp_unit_growth.
3583 * params.opt: Remove legacy param name.
3585 2020-01-13 Martin Sebor <msebor@redhat.com>
3587 PR tree-optimization/93213
3588 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3589 stores to be eliminated.
3591 2020-01-13 Martin Liska <mliska@suse.cz>
3593 * opts.c (print_help): Do not print CL_PARAM
3594 and CL_WARNING for CL_OPTIMIZATION.
3596 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3599 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3600 depending on optimization settings.
3602 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3604 PR tree-optimization/90838
3605 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3606 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3607 argument rather than to initialize temporary for targets that
3608 don't use the mode argument at all. Initialize ctzval to avoid
3611 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3613 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3614 * tree-core.h: Document it.
3615 * gimplify.c (gimplify_omp_workshare): Set it.
3616 * omp-low.c (lower_omp_target): Use it.
3617 * tree-pretty-print.c (dump_omp_clause): Print it.
3619 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3620 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3622 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3624 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3625 * common.opt (fdiagnostics-path-format=): New option.
3626 (diagnostic_path_format): New enum.
3627 (fdiagnostics-show-path-depths): New option.
3628 * coretypes.h (diagnostic_event_id_t): New forward decl.
3629 * diagnostic-color.c (color_dict): Add "path".
3630 * diagnostic-event-id.h: New file.
3631 * diagnostic-format-json.cc (json_from_expanded_location): Make
3633 (json_end_diagnostic): Call context->make_json_for_path if it
3634 exists and the diagnostic has a path.
3635 (diagnostic_output_format_init): Clear context->print_path.
3636 * diagnostic-path.h: New file.
3637 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3638 when printing a run of events in a diagnostic_path so that they
3639 all get the same color.
3640 (layout::m_diagnostic_path_p): New field.
3641 (layout::layout): Initialize it.
3642 (layout::print_any_labels): Don't colorize the label text for an
3643 event in a diagnostic_path.
3644 (gcc_rich_location::add_location_if_nearby): Add
3645 "restrict_to_current_line_spans" and "label" params. Pass the
3646 former to layout.maybe_add_location_range; pass the latter
3647 when calling add_range.
3648 * diagnostic.c: Include "diagnostic-path.h".
3649 (diagnostic_initialize): Initialize context->path_format and
3650 context->show_path_depths.
3651 (diagnostic_show_any_path): New function.
3652 (diagnostic_path::interprocedural_p): New function.
3653 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3654 (simple_diagnostic_path::num_events): New function.
3655 (simple_diagnostic_path::get_event): New function.
3656 (simple_diagnostic_path::add_event): New function.
3657 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3658 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3659 (debug): New overload taking a diagnostic_path *.
3660 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3661 * diagnostic.h (enum diagnostic_path_format): New enum.
3662 (json::value): New forward decl.
3663 (diagnostic_context::path_format): New field.
3664 (diagnostic_context::show_path_depths): New field.
3665 (diagnostic_context::print_path): New callback field.
3666 (diagnostic_context::make_json_for_path): New callback field.
3667 (diagnostic_show_any_path): New decl.
3668 (json_from_expanded_location): New decl.
3669 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3670 (-fdiagnostics-show-path-depths): New option.
3671 (-fdiagnostics-color): Add "path" to description of default
3672 GCC_COLORS; describe it.
3673 (-fdiagnostics-format=json): Document how diagnostic paths are
3674 represented in the JSON output format.
3675 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3676 Add optional params "restrict_to_current_line_spans" and "label".
3677 * opts.c (common_handle_option): Handle
3678 OPT_fdiagnostics_path_format_ and
3679 OPT_fdiagnostics_show_path_depths.
3680 * pretty-print.c: Include "diagnostic-event-id.h".
3681 (pp_format): Implement "%@" format code for printing
3682 diagnostic_event_id_t *.
3683 (selftest::test_pp_format): Add tests for "%@".
3684 * selftest-run-tests.c (selftest::run_tests): Call
3685 selftest::tree_diagnostic_path_cc_tests.
3686 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3687 * toplev.c (general_init): Initialize global_dc->path_format and
3688 global_dc->show_path_depths.
3689 * tree-diagnostic-path.cc: New file.
3690 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3691 non-static. Drop "diagnostic" param in favor of storing the
3692 original value of "where" and re-using it.
3693 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3694 maybe_unwind_expanded_macro_loc.
3695 (tree_diagnostics_defaults): Initialize context->print_path and
3696 context->make_json_for_path.
3697 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3699 (default_tree_make_json_for_path): New decl.
3700 (maybe_unwind_expanded_macro_loc): New decl.
3702 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3704 PR tree-optimization/93210
3705 * fold-const.h (native_encode_initializer,
3706 can_native_interpret_type_p): Declare.
3707 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3709 (native_encode_initializer): New function, moved from dwarf2out.c.
3710 Adjust to native_encode_expr compatible arguments, including dry-run
3711 and partial extraction modes. Don't handle STRING_CST.
3712 (can_native_interpret_type_p): No longer static.
3713 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3714 offset / BITS_PER_UNIT fits into int and don't call it if
3715 can_native_interpret_type_p fails. If suboff is NULL and for
3716 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3717 native_encode_initializer.
3718 (fold_const_aggregate_ref_1): Formatting fix.
3719 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3720 (tree_add_const_value_attribute): Adjust caller.
3722 PR tree-optimization/90838
3723 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3724 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3725 CTZ_DEFINED_VALUE_AT_ZERO.
3727 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3730 * lra-constraints.c (match_reload): Permit input operands have the
3731 same mode as output while other input operands have a different
3734 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3736 PR tree-optimization/90838
3737 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3738 (check_ctz_string): Likewise.
3739 (optimize_count_trailing_zeroes): Likewise.
3740 (simplify_count_trailing_zeroes): Likewise.
3741 (pass_forwprop::execute): Try ctz simplification.
3742 * match.pd: Add matching for ctz idioms.
3744 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3746 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3748 (aarch64_invalid_unary_op): New function for target hook.
3749 (aarch64_invalid_binary_op): New function for target hook.
3751 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3753 * config.gcc: Add arm_bf16.h.
3754 * config/aarch64/aarch64-builtins.c
3755 (aarch64_simd_builtin_std_type): Add BFmode.
3756 (aarch64_init_simd_builtin_types): Define element types for vector
3758 (aarch64_init_bf16_types): New function.
3759 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3760 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3762 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3763 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3765 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3766 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3767 * config/aarch64/aarch64.c
3768 (aarch64_classify_vector_mode): Add support for BF types.
3769 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3770 (aarch64_vq_mode): Add support for BF types.
3771 (aarch64_simd_container_mode): Add support for BF types.
3772 (aarch64_mangle_type): Add support for BF scalar type.
3773 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3774 * config/aarch64/arm_bf16.h: New file.
3775 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3776 * config/aarch64/iterators.md: Add BF types to mode attributes.
3777 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3779 2020-01-10 Jason Merrill <jason@redhat.com>
3781 PR c++/93173 - incorrect tree sharing.
3782 * gimplify.c (copy_if_shared): No longer static.
3783 * gimplify.h: Declare it.
3785 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3787 * doc/invoke.texi (-msve-vector-bits=): Document that
3788 -msve-vector-bits=128 now generates VL-specific code for
3789 little-endian targets.
3790 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3791 build_vector_type_for_mode to construct the data vector types.
3792 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3793 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3794 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3795 for 128-bit vectors.
3797 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3799 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3802 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3804 * config/aarch64/aarch64-builtins.c
3805 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3806 rather than checking the number of elements and the element mode.
3808 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3810 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3811 get_related_vectype_for_scalar_type rather than build_vector_type
3812 to create the index type for a conditional reduction.
3814 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3816 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3817 for any type of gather or scatter, including strided accesses.
3819 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3821 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3824 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3826 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3828 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3829 parameter and its use to reset DR_OFFSET's.
3830 (vect_transform_loop): Remove orig_drs_init argument.
3831 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3832 member of dr_vec_info rather than the offset of the associated
3833 data_reference's innermost_loop_behavior.
3834 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3835 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3836 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3837 get_dr_vinfo_offset.
3838 (vectorizable_store): Likewise.
3839 (vectorizable_load): Likewise.
3841 2020-01-10 Richard Biener <rguenther@suse.de>
3843 * gimple-ssa-store-merging
3844 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3846 2020-01-10 Martin Liska <mliska@suse.cz>
3849 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3850 encapsulation that was there before r280040.
3852 2020-01-10 Richard Biener <rguenther@suse.de>
3855 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3856 sequences to avoid walking them again for secondary opportunities.
3857 (pass_lower_eh_dispatch::execute): Instead actually insert
3860 2020-01-10 Richard Biener <rguenther@suse.de>
3863 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3864 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3865 avoid quadraticness.
3867 2020-01-10 Martin Jambor <mjambor@suse.cz>
3869 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3870 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3871 to get param_ipa_sra_max_replacements.
3872 (param_splitting_across_edge): Pass the caller to
3873 pull_accesses_from_callee.
3875 2020-01-10 Martin Jambor <mjambor@suse.cz>
3877 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3878 * ipa-cp.c (max_new_size): Removed.
3879 (orig_overall_size): New variable.
3880 (get_max_overall_size): New function.
3881 (estimate_local_effects): Use it. Adjust dump.
3882 (decide_about_value): Likewise.
3883 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3884 orig_overall_size. Adjust dump.
3885 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3887 2020-01-10 Martin Jambor <mjambor@suse.cz>
3889 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3890 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3891 instead of param_ipa_max_agg_items.
3892 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3893 optimization info for the callee.
3895 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3897 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3898 markers if debug_inline_points is false.
3900 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3902 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3904 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3905 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3906 aarch64-sve-builtins-sve2.h.
3907 (aarch64-sve-builtins-sve2.o): New rule.
3908 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3909 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3910 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3911 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3912 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3913 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3915 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3916 instructions that are handled here.
3917 (@cond_asrd<mode>): Generalize to...
3918 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3919 (*cond_asrd<mode>_2): Generalize to...
3920 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3921 (*cond_asrd<mode>_z): Generalize to...
3922 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3923 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3924 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3925 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3926 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3928 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3929 (@aarch64_scatter_stnt<mode>): Likewise.
3930 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3931 (@aarch64_mul_lane_<mode>): Likewise.
3932 (@aarch64_sve_suqadd<mode>_const): Likewise.
3933 (*<sur>h<addsub><mode>): Generalize to...
3934 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3936 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3937 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3938 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3939 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3940 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3941 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3942 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3943 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3944 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3945 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3946 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3947 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3948 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3949 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3950 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3951 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3952 (@aarch64_sve2_xar<mode>): Likewise.
3953 (@aarch64_sve2_bcax<mode>): Likewise.
3954 (*aarch64_sve2_eor3<mode>): Rename to...
3955 (@aarch64_sve2_eor3<mode>): ...this.
3956 (@aarch64_sve2_bsl<mode>): New expander.
3957 (@aarch64_sve2_nbsl<mode>): Likewise.
3958 (@aarch64_sve2_bsl1n<mode>): Likewise.
3959 (@aarch64_sve2_bsl2n<mode>): Likewise.
3960 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3961 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3962 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3963 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3964 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3965 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3966 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3967 (<su>mull<bt><Vwide>): Generalize to...
3968 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3970 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3971 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3972 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3973 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3974 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3975 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3976 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3977 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3978 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3979 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3980 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3981 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3982 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3983 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3984 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3985 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3986 (<SHRNB:r>shrnb<mode>): Generalize to...
3987 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3989 (<SHRNT:r>shrnt<mode>): Generalize to...
3990 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3992 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3993 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3994 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3995 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3996 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3997 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3998 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3999 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4000 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4001 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4002 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4003 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4004 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4005 (@aarch64_sve2_cvtnt<mode>): Likewise.
4006 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4007 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4008 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4009 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4010 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4011 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4012 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4013 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4014 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4015 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4016 (@aarch64_sve2_pmul<mode>): Likewise.
4017 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4018 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4019 (@aarch64_sve2_tbl2<mode>): Likewise.
4020 (@aarch64_sve2_tbx<mode>): Likewise.
4021 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4022 (@aarch64_sve2_histcnt<mode>): Likewise.
4023 (@aarch64_sve2_histseg<mode>): Likewise.
4024 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4025 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4026 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4027 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4028 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4029 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4030 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4031 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4032 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4033 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4034 (SVE2_PMULL_PAIR_I): New mode iterators.
4035 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4036 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4037 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4038 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4039 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4040 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4041 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4042 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4043 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4044 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4045 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4046 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4047 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4048 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4049 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4050 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4051 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4052 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4053 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4054 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4055 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4056 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4057 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4058 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4059 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4060 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4061 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4062 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4063 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4064 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4065 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4067 (VNARROW, Ventype): New mode attributes.
4068 (Vewtype): Handle VNx2DI. Fix typo in comment.
4069 (VDOUBLE): New mode attribute.
4070 (sve_lane_con): Handle VNx8HI.
4071 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4072 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4073 (sve_int_op, sve_int_op_rev): Handle the above codes.
4074 (sve_pred_int_rhs2_operand): Likewise.
4075 (MULLBT, SHRNB, SHRNT): Delete.
4076 (SVE_INT_SHIFT_IMM): New int iterator.
4077 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4078 and UNSPEC_WHILEHS for TARGET_SVE2.
4079 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4080 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4081 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4082 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4083 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4084 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4085 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4086 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4087 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4088 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4089 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4090 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4091 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4092 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4093 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4094 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4095 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4096 (optab): Handle the new unspecs.
4097 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4099 (lr): Handle the new unspecs.
4101 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4102 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4103 (sve_int_qsub_op): New int attributes.
4104 (sve_fp_op, rot): Handle the new unspecs.
4105 * config/aarch64/aarch64-sve-builtins.h
4106 (function_resolver::require_matching_pointer_type): Declare.
4107 (function_resolver::resolve_unary): Add an optional boolean argument.
4108 (function_resolver::finish_opt_n_resolution): Add an optional
4109 type_suffix_index argument.
4110 (gimple_folder::redirect_call): Declare.
4111 (gimple_expander::prepare_gather_address_operands): Add an optional
4113 * config/aarch64/aarch64-sve-builtins.cc: Include
4114 aarch64-sve-builtins-sve2.h.
4115 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4116 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4117 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4118 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4119 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4120 (TYPES_s_unsigned): Likewise.
4121 (TYPES_s_integer): Use TYPES_s_unsigned.
4122 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4123 (TYPES_sd_integer): Use them.
4124 (TYPES_d_unsigned): New macro.
4125 (TYPES_d_integer): Use it.
4126 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4127 (TYPES_cvt_narrow): Likewise.
4128 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4129 (preds_mx): New variable.
4130 (function_builder::add_overloaded_function): Allow the new feature
4131 set to be more restrictive than the original one.
4132 (function_resolver::infer_pointer_type): Remove qualifiers from
4133 the pointer type before printing it.
4134 (function_resolver::require_matching_pointer_type): New function.
4135 (function_resolver::resolve_sv_displacement): Handle functions
4136 that don't support 32-bit vector indices or svint32_t vector offsets.
4137 (function_resolver::finish_opt_n_resolution): Take the inferred type
4138 as a separate argument.
4139 (function_resolver::resolve_unary): Optionally treat all forms in
4140 the same way as normal merging functions.
4141 (gimple_folder::redirect_call): New function.
4142 (function_expander::prepare_gather_address_operands): Add an argument
4143 that says whether scaled forms are available. If they aren't,
4144 handle scaling of vector indices and don't add the extension and
4146 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4147 fall back to using cond_* instead.
4148 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4149 Split out the member variables into...
4150 (rtx_code_function_base): ...this new base class.
4151 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4152 (unspec_based_function): Split out the member variables into...
4153 (unspec_based_function_base): ...this new base class.
4154 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4155 (unspec_based_function_exact_insn): New class.
4156 (unspec_based_add_function, unspec_based_add_lane_function)
4157 (unspec_based_lane_function, unspec_based_pred_function)
4158 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4159 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4160 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4162 (unspec_based_fused_function): New class.
4163 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4164 (unspec_based_fused_lane_function): New class.
4165 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4167 (CODE_FOR_MODE1): New macro.
4168 (fixed_insn_function): New class.
4169 (while_comparison): Likewise.
4170 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4171 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4172 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4173 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4174 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4175 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4176 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4177 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4178 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4179 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4180 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4181 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4182 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4183 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4184 Also add an initial argument for unary_convert_narrowt, regardless
4185 of the predication type.
4186 (build_32_64): Allow loads and stores to specify MODE_none.
4187 (build_sv_index64, build_sv_uint_offset): New functions.
4188 (long_type_suffix): New function.
4189 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4190 (binary_imm_long_base, load_gather_sv_base): Likewise.
4191 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4192 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4193 (unary_narrowb_base, unary_narrowt_base): Likewise.
4194 (binary_long_lane_def, binary_long_lane): New shape.
4195 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4196 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4197 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4198 (binary_to_uint_def, binary_to_uint): Likewise.
4199 (binary_wide_def, binary_wide): Likewise.
4200 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4201 (compare_def, compare): Likewise.
4202 (compare_ptr_def, compare_ptr): Likewise.
4203 (load_ext_gather_index_restricted_def,
4204 load_ext_gather_index_restricted): Likewise.
4205 (load_ext_gather_offset_restricted_def,
4206 load_ext_gather_offset_restricted): Likewise.
4207 (load_gather_sv_def): Inherit from load_gather_sv_base.
4208 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4209 (shift_left_imm_def, shift_left_imm): Likewise.
4210 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4211 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4212 (store_scatter_index_restricted_def,
4213 store_scatter_index_restricted): Likewise.
4214 (store_scatter_offset_restricted_def,
4215 store_scatter_offset_restricted): Likewise.
4216 (tbl_tuple_def, tbl_tuple): Likewise.
4217 (ternary_long_lane_def, ternary_long_lane): Likewise.
4218 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4219 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4220 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4221 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4222 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4223 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4224 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4225 (ternary_uint_def, ternary_uint): Likewise.
4226 (unary_convert): Fix typo in comment.
4227 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4228 (unary_long_def, unary_long): Likewise.
4229 (unary_narrowb_def, unary_narrowb): Likewise.
4230 (unary_narrowt_def, unary_narrowt): Likewise.
4231 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4232 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4233 (unary_to_int_def, unary_to_int): Likewise.
4234 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4235 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4236 (svasrd_impl): Delete.
4237 (svcadd_impl::expand): Handle integer operations too.
4238 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4239 new functions to derive the unspec numbers.
4240 (svmla_svmls_lane_impl): Replace with...
4241 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4242 integer operations too.
4243 (svwhile_impl): Rename to...
4244 (svwhilelx_impl): ...this and inherit from while_comparison.
4245 (svasrd): Use unspec_based_function.
4246 (svmla_lane): Use svmla_lane_impl.
4247 (svmls_lane): Use svmls_lane_impl.
4248 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4249 (svwhilele, svwhilelt): Use svwhilelx_impl.
4250 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4251 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4252 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4253 * config/aarch64/aarch64-sve-builtins.def: Include
4254 aarch64-sve-builtins-sve2.def.
4256 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4258 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4259 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4260 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4261 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4262 immediates as well as vector ones.
4263 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4264 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4265 (aarch64_sve_qsub_immediate): Update calls accordingly.
4267 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4269 * config/aarch64/aarch64-sve2.md: Add banner comments.
4270 (<su>mulh<r>s<mode>3): Move further up file.
4271 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4272 (*aarch64_sve2_sra<mode>): Move further down file.
4273 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4275 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4277 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4279 (while_optab_cmp): Handle them.
4280 * config/aarch64/aarch64-sve.md
4281 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4282 and add a "@" marker.
4283 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4284 instead of gen_aarch64_sve2_while_ptest.
4285 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4287 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4289 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4290 (UNSPEC_WHILELE): ...this.
4291 (UNSPEC_WHILE_LO): Rename to...
4292 (UNSPEC_WHILELO): ...this.
4293 (UNSPEC_WHILE_LS): Rename to...
4294 (UNSPEC_WHILELS): ...this.
4295 (UNSPEC_WHILE_LT): Rename to...
4296 (UNSPEC_WHILELT): ...this.
4297 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4298 (cmp_op, while_optab_cmp): Likewise.
4299 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4300 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4301 (svwhilelt): Likewise.
4303 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4305 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4306 (unary_to_uint): Define.
4307 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4308 (unary_count): Rename to...
4309 (unary_to_uint_def, unary_to_uint): ...this.
4310 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4312 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4314 * config/aarch64/aarch64-sve-builtins-functions.h
4315 (code_for_mode_function): New class.
4316 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4317 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4318 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4319 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4320 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4322 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4324 * config/aarch64/iterators.md (addsub): New code attribute.
4325 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4327 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4328 in the asm string and attributes. Fix indentation.
4329 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4331 (@aarch64_sve_<optab><mode>): ...this.
4332 * config/aarch64/aarch64-sve-builtins.h
4333 (function_expander::expand_signed_unpred_op): Delete.
4334 * config/aarch64/aarch64-sve-builtins.cc
4335 (function_expander::expand_signed_unpred_op): Likewise.
4336 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4337 try using code_for_aarch64_sve instead.
4338 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4339 (svqsub_impl): Likewise.
4340 (svqadd, svqsub): Use rtx_code_function instead.
4342 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4344 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4345 (HADDSUB, sur, addsub): Remove them.
4347 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4349 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4350 functions the same way, rather than singling out those that
4351 aren't mapped directly to optabs.
4353 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4355 * target.def (compatible_vector_types_p): New target hook.
4356 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4357 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4358 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4359 * doc/tm.texi: Regenerate.
4360 * gimple-expr.c: Include target.h.
4361 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4362 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4364 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4365 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4366 Use the original predicate if it already has a suitable type.
4368 2020-01-09 Martin Jambor <mjambor@suse.cz>
4370 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4371 resolve_speculation and redirect_call_stmt_to_callee static. Change
4372 return type of set_call_stmt to cgraph_edge *.
4373 * auto-profile.c (afdo_indirect_call): Adjust call to
4374 redirect_call_stmt_to_callee.
4375 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4376 make the this pointer explicit, adjust self-recursive calls and the
4377 call top make_direct. Return the resulting edge.
4378 (cgraph_edge::remove): Make this pointer explicit.
4379 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4380 (cgraph_edge::make_direct): Likewise, adjust call to
4381 resolve_speculation.
4382 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4383 call to set_call_stmt.
4384 (cgraph_update_edges_for_call_stmt_node): Update call to
4385 set_call_stmt and remove.
4386 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4387 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4388 (cgraph_node::create_edge_including_clones): Moved "first" definition
4389 of edge to the block where it was used. Adjusted calls to
4391 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4392 cgraph_edge::remove.
4393 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4394 make_direct and redirect_call_stmt_to_callee.
4395 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4396 resolve_speculation and make_direct.
4397 * ipa-inline-transform.c (inline_transform): Adjust call to
4398 redirect_call_stmt_to_callee.
4399 (check_speculations_1):: Adjust call to resolve_speculation.
4400 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4401 resolve-speculation.
4402 (inline_small_functions): Adjust call to resolve_speculation.
4403 (ipa_inline): Likewise.
4404 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4406 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4407 safe with regards to edge removal, adjust calls to
4408 redirect_call_stmt_to_callee.
4409 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4410 and redirect_call_stmt_to_callee.
4411 * multiple_target.c (create_dispatcher_calls): Adjust call to
4412 redirect_call_stmt_to_callee
4413 (redirect_to_specific_clone): Likewise.
4414 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4415 Adjust calls to cgraph_edge::remove.
4416 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4417 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4418 (expand_call_inline): Adjust call to cgraph_edge::remove.
4420 2020-01-09 Martin Liska <mliska@suse.cz>
4422 * params.opt: Set Optimization for
4423 param_max_speculative_devirt_maydefs.
4425 2020-01-09 Martin Sebor <msebor@redhat.com>
4429 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4431 2020-01-09 Martin Liska <mliska@suse.cz>
4433 * auto-profile.c (auto_profile): Use opt_for_fn
4435 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4436 (propagate_vals_across_arith_jfunc): Likewise.
4437 (hint_time_bonus): Likewise.
4438 (incorporate_penalties): Likewise.
4439 (good_cloning_opportunity_p): Likewise.
4440 (perform_estimation_of_a_value): Likewise.
4441 (estimate_local_effects): Likewise.
4442 (ipcp_propagate_stage): Likewise.
4443 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4444 (set_switch_stmt_execution_predicate): Likewise.
4445 (analyze_function_body): Likewise.
4446 * ipa-inline-analysis.c (offline_size): Likewise.
4447 * ipa-inline.c (early_inliner): Likewise.
4448 * ipa-prop.c (ipa_analyze_node): Likewise.
4449 (ipcp_transform_function): Likewise.
4450 * ipa-sra.c (process_scan_results): Likewise.
4451 (ipa_sra_summarize_function): Likewise.
4452 * params.opt: Rename ipcp-unit-growth to
4453 ipa-cp-unit-growth. Add Optimization for various
4454 IPA-related parameters.
4456 2020-01-09 Richard Biener <rguenther@suse.de>
4459 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4461 2020-01-09 Richard Biener <rguenther@suse.de>
4463 PR tree-optimization/93040
4464 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4466 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4468 * common/config/avr/avr-common.c (avr_option_optimization_table)
4469 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4471 2020-01-09 Martin Liska <mliska@suse.cz>
4473 * cgraphclones.c (symbol_table::materialize_all_clones):
4474 Use cgraph_node::dump_name.
4476 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4479 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4480 output_operand_lossage instead of gcc_unreachable.
4481 * doc/md.texi (riscv f constraint): Fix typo.
4484 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4485 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4486 CONST_SCALAR_INT_P instead of CONST_INT_P.
4487 (*subv<mode>4_1): Rename to ...
4488 (subv<mode>4_1): ... this.
4489 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4490 define_insn_and_split patterns.
4491 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4494 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4496 * vec.c (class selftest::count_dtor): New class.
4497 (selftest::test_auto_delete_vec): New test.
4498 (selftest::vec_c_tests): Call it.
4499 * vec.h (class auto_delete_vec): New class template.
4500 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4502 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4504 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4506 2020-01-08 Jim Wilson <jimw@sifive.com>
4508 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4509 use of TLS_MODEL_LOCAL_EXEC when not pic.
4511 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4513 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4516 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4519 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4520 *stack_protect_set_3 peephole2): Also check that the second
4521 insns source is general_operand.
4524 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4525 predicate for output operand instead of register_operand.
4526 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4527 memory destination and non-memory operands[2].
4529 2020-01-08 Martin Liska <mliska@suse.cz>
4531 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4532 ::dump_asm_name instead of (::name or ::asm_name).
4533 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4534 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4535 (analyze_functions): Likewise.
4536 (expand_all_functions): Likewise.
4537 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4538 (propagate_bits_across_jump_function): Likewise.
4539 (dump_profile_updates): Likewise.
4540 (ipcp_store_bits_results): Likewise.
4541 (ipcp_store_vr_results): Likewise.
4542 * ipa-devirt.c (dump_targets): Likewise.
4543 * ipa-fnsummary.c (analyze_function_body): Likewise.
4544 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4545 (process_hsa_functions): Likewise.
4546 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4547 (set_alias_uids): Likewise.
4548 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4549 * ipa-inline.c (recursive_inlining): Likewise.
4550 (inline_to_all_callers_1): Likewise.
4551 (ipa_inline): Likewise.
4552 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4553 (ipa_propagate_frequency): Likewise.
4554 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4555 (remove_described_reference): Likewise.
4556 * ipa-pure-const.c (worse_state): Likewise.
4557 (check_retval_uses): Likewise.
4558 (analyze_function): Likewise.
4559 (propagate_pure_const): Likewise.
4560 (propagate_nothrow): Likewise.
4561 (dump_malloc_lattice): Likewise.
4562 (propagate_malloc): Likewise.
4563 (pass_local_pure_const::execute): Likewise.
4564 * ipa-visibility.c (optimize_weakref): Likewise.
4565 (function_and_variable_visibility): Likewise.
4566 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4567 (ipa_discover_variable_flags): Likewise.
4568 * lto-streamer-out.c (output_function): Likewise.
4569 (output_constructor): Likewise.
4570 * tree-inline.c (copy_bb): Likewise.
4571 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4572 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4574 2020-01-08 Richard Biener <rguenther@suse.de>
4577 * tree-eh.c (sink_clobbers): Update virtual operands for
4578 the first and last stmt only. Add a dry-run capability.
4579 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4580 after CFG manipulations and in RPO order to catch all
4581 secondary opportunities reliably.
4583 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4586 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4588 2019-01-08 Richard Biener <rguenther@suse.de>
4591 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4592 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4593 virtual operand, also updating SSA use.
4594 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4595 Update stmt after resetting virtual operand.
4596 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4597 * gimple-iterator.c (gsi_remove): When not removing the stmt
4598 permanently do not delink immediate uses or mark the stmt modified.
4600 2020-01-08 Martin Liska <mliska@suse.cz>
4602 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4603 (ipa_call_context::estimate_size_and_time): Likewise.
4604 (inline_analyze_function): Likewise.
4606 2020-01-08 Martin Liska <mliska@suse.cz>
4608 * cgraph.c (cgraph_node::dump): Use systematically
4611 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4613 Add -nodevicespecs option for avr.
4616 * config/avr/avr.opt (-nodevicespecs): New driver option.
4617 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4618 "-specs=device-specs/..." if that option is not set.
4619 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4621 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4623 Implement 64-bit double functions for avr.
4626 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4627 --with-double-comparison.
4628 * doc/install.texi: Document them.
4629 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4630 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4631 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4632 * doc/invoke.texi (AVR Built-in Macros): Document them.
4633 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4634 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4635 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4637 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4640 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4641 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4642 when only building rm-profile multilibs.
4644 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4647 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4648 lattice for a value to check.
4649 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4650 finite propagation in self-recursive scc.
4652 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4654 * ipa-inline.c (caller_growth_limits): Restore the AND.
4656 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4658 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4659 (VEC_ALLREG_ALT): New iterator.
4660 (VEC_ALLREG_INT_MODE): New iterator.
4661 (VCMP_MODE): New iterator.
4662 (VCMP_MODE_INT): New iterator.
4663 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4664 (vec_cmp<u>v64qidi): New define_expand.
4665 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4666 (vec_cmpu<mode>di_exec): New define_expand.
4667 (vec_cmp<u>v64qidi_exec): New define_expand.
4668 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4669 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4670 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4671 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4672 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4673 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4674 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4675 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4676 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4677 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4679 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4680 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4682 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4684 * config/gcn/constraints.md (DA): Update description and match.
4686 (Db): New constraint.
4687 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4689 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4690 Implement 'Db' mixed immediate type.
4691 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4692 (addcv64si3_dup<exec_vcc>): Delete.
4693 (subcv64si3<exec_vcc>): Rework constraints.
4694 (addv64di3): Rework constraints.
4695 (addv64di3_exec): Rework constraints.
4696 (subv64di3): Rework constraints.
4697 (addv64di3_dup): Delete.
4698 (addv64di3_dup_exec): Delete.
4699 (addv64di3_zext): Rework constraints.
4700 (addv64di3_zext_exec): Rework constraints.
4701 (addv64di3_zext_dup): Rework constraints.
4702 (addv64di3_zext_dup_exec): Rework constraints.
4703 (addv64di3_zext_dup2): Rework constraints.
4704 (addv64di3_zext_dup2_exec): Rework constraints.
4705 (addv64di3_sext_dup2): Rework constraints.
4706 (addv64di3_sext_dup2_exec): Rework constraints.
4708 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4710 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4711 existing target checks.
4713 2020-01-07 Richard Biener <rguenther@suse.de>
4715 * doc/install.texi: Bump minimal supported MPC version.
4717 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4719 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4720 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4721 * langhooks.c: Include stor-layout.h.
4722 (lhd_simulate_enum_decl): New function.
4723 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4724 handle_arm_sve_h for the LTO frontend.
4725 (register_vector_type): Cope with null returns from pushdecl.
4727 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4729 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4730 (aarch64_sve::nvectors_if_data_type): Replace with...
4731 (aarch64_sve::builtin_type_p): ...this.
4732 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4733 (find_vector_type): Delete.
4734 (add_sve_type_attribute): New function.
4735 (lookup_sve_type_attribute): Likewise.
4736 (register_builtin_types): Add an "SVE type" attribute to each type.
4737 (register_tuple_type): Likewise.
4738 (svbool_type_p, nvectors_if_data_type): Delete.
4739 (mangle_builtin_type): Use lookup_sve_type_attribute.
4740 (builtin_type_p): Likewise. Add an overload that returns the
4741 number of constituent vector and predicate registers.
4742 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4743 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4744 instead of aarch64_sve_argument_p.
4745 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4746 (aarch64_pass_by_reference): Likewise.
4747 (aarch64_function_value_1): Likewise.
4748 (aarch64_return_in_memory): Likewise.
4749 (aarch64_layout_arg): Likewise.
4751 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4753 PR tree-optimization/93156
4754 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4755 least significant bit is always clear.
4757 PR tree-optimization/93118
4758 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4759 simplifier with two intermediate conversions.
4761 2020-01-07 Martin Liska <mliska@suse.cz>
4763 * params.opt: Add Optimization for various parameters.
4765 2020-01-07 Martin Liska <mliska@suse.cz>
4768 * doc/extend.texi: Explain cloning for target_clone
4771 2020-01-07 Martin Liska <mliska@suse.cz>
4773 PR tree-optimization/92860
4774 * common.opt: Make in Optimization option
4775 as it is affected by -O0, which is an Optimization
4777 * tree-inline.c (tree_inlinable_function_p):
4778 Use opt_for_fn for warn_inline.
4779 (expand_call_inline): Likewise.
4781 2020-01-07 Martin Liska <mliska@suse.cz>
4783 PR tree-optimization/92860
4784 * common.opt: Make flag_ree as optimization
4787 2020-01-07 Martin Liska <mliska@suse.cz>
4789 PR optimization/92860
4790 * params.opt: Mark param_min_crossjump_insns with Optimization
4793 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4795 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4796 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4798 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4800 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4801 helper function to return the valid addressing formats for a given
4802 hard register and mode.
4803 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4805 * config/rs6000/constraints.md (Q constraint): Update
4807 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4810 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4811 Use 'Q' for doing vector extract from memory.
4812 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4814 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4815 doing vector extract from memory.
4816 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4817 extract from memory.
4819 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4820 for the offset being 34-bits when -mcpu=future is used.
4822 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4824 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4825 instead of cmpib_comparison_operator in cmpib patterns.
4826 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4827 of cmpib_comparison_operator. Revise comment.
4829 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4831 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4832 in an IFN_DIV_POW2 node to be equal.
4834 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4836 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4837 (vect_check_scalar_mask): ...this.
4838 (vectorizable_store, vectorizable_load): Update call accordingly.
4839 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4840 argument in calls to conditional internal functions.
4842 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4844 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4845 '0' matching inputs.
4846 (subv64di3_exec): Likewise.
4848 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4850 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4851 * doc/md.texi (movstr): Likewise.
4853 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4855 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4858 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4860 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4862 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4863 to a temporary file and use move-if-change to update the real
4864 file where necessary.
4866 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4868 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4869 rather than Upa for CPY /M.
4871 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4873 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4876 2020-01-06 Martin Liska <mliska@suse.cz>
4878 PR tree-optimization/92860
4879 * params.opt: Mark param_max_combine_insns with Optimization
4882 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4885 * config/i386/i386.md (SWIDWI): New mode iterator.
4886 (DWI, dwi): Add TImode variants.
4887 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4888 <general_hilo_operand> instead of <general_operand>. Use
4889 CONST_SCALAR_INT_P instead of CONST_INT_P.
4890 (*addv<mode>4_1): Rename to ...
4891 (addv<mode>4_1): ... this.
4892 (QWI): New mode attribute.
4893 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4894 define_insn_and_split patterns.
4895 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4897 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4898 <general_hilo_operand> instead of <general_operand>.
4899 (*addcarry<mode>_1): New define_insn.
4900 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4902 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4904 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4905 Use "call" instead of "set".
4907 2020-01-03 Martin Jambor <mjambor@suse.cz>
4910 * ipa-cp.c (print_all_lattices): Skip functions without info.
4912 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4915 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4916 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4917 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4918 for 'e' simd clones.
4921 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4923 (mprefer-vector-width=): Add Save.
4924 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4925 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4926 (ix86_debug_options, ix86_function_specific_print): Adjust
4927 ix86_target_string callers.
4928 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4929 (ix86_valid_target_attribute_tree): Likewise.
4930 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4931 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4932 ix86_target_string caller.
4935 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4936 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4937 instead of gen_int_shift_amount + convert_modes.
4939 PR rtl-optimization/93088
4940 * loop-iv.c (find_single_def_src): Punt after looking through
4941 128 reg copies for regs with single definitions. Move definitions
4944 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4946 * config/arm/arm-c.c (arm_cpu_builtins): Define
4947 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4948 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4949 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4950 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4951 * config/arm/arm-tables.opt: Regenerated.
4952 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4953 arm_arch_i8mm and arm_arch_bf16 when enabled.
4954 * config/arm/arm.h (TARGET_I8MM): New macro.
4955 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4956 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4957 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4958 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4959 (v8_6_a_simd_variants): New.
4960 (v8_*_a_simd_variants): Add i8mm and bf16.
4961 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4963 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4966 * predict.c (compute_function_frequency): Don't call
4967 warn_function_cold on functions that already have cold attribute.
4969 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4972 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4973 COMDAT group function labels in .data.rel.ro.local section.
4974 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4977 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4978 comparison_operator in B and S integer comparisons. Likewise, use
4979 ordered_comparison_operator instead of cmpib_comparison_operator in
4981 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4983 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4985 Update copyright years.
4987 * gcc.c (process_command): Update copyright notice dates.
4988 * gcov-dump.c (print_version): Ditto.
4989 * gcov.c (print_version): Ditto.
4990 * gcov-tool.c (print_version): Ditto.
4991 * gengtype.c (create_file): Ditto.
4992 * doc/cpp.texi: Bump @copying's copyright year.
4993 * doc/cppinternals.texi: Ditto.
4994 * doc/gcc.texi: Ditto.
4995 * doc/gccint.texi: Ditto.
4996 * doc/gcov.texi: Ditto.
4997 * doc/install.texi: Ditto.
4998 * doc/invoke.texi: Ditto.
5000 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5002 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5005 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5007 PR tree-optimization/93098
5008 * match.pd (popcount): For shift amounts, use integer_onep
5009 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5010 tests. Make sure that precision is power of two larger than or equal
5011 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5012 instead of ULL suffixed constants. Formatting fixes.
5014 Copyright (C) 2020 Free Software Foundation, Inc.
5016 Copying and distribution of this file, with or without modification,
5017 are permitted in any medium without royalty provided the copyright
5018 notice and this notice are preserved.