1 2020-03-03 Jakub Jelinek <jakub@redhat.com>
4 * config/s390/s390.h (OPTION_DEFAULT_SPECS): Reorder.
6 PR rtl-optimization/94002
7 * explow.c (plus_constant): Punt if cst has VOIDmode and
8 get_pool_mode is different from mode.
10 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
12 * config/arc/arc.c (leigitimate_small_data_address_p): Check if an
13 address has an offset which fits the scalling constraint for a
15 (legitimate_scaled_address_p): Update use
16 leigitimate_small_data_address_p.
17 (arc_print_operand): Likewise.
18 (arc_legitimate_address_p): Likewise.
19 (legitimate_small_data_address_p): Likewise.
21 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
23 * config/arc/arc.md (fmasf4_fpu): Use accl_operand predicate.
24 (fnmasf4_fpu): Likewise.
26 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
28 * config/arc/arc.md (adddi3): Early expand the 64bit operation into
31 (adddi3_i): Remove pattern.
34 2020-03-03 Claudiu Zissulescu <claziss@synopsys.com>
36 * config/arc/arc.md (eh_return): Add length info.
38 2020-03-02 David Malcolm <dmalcolm@redhat.com>
40 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
42 2020-03-02 David Malcolm <dmalcolm@redhat.com>
44 * doc/invoke.texi (Static Analyzer Options): Add
45 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
48 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
51 * config/i386/i386.md (movstrict<mode>): Allow only
52 registers with VALID_INT_MODE_P modes.
54 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
56 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
57 (reduc_insn): Use 'U' and 'B' operand codes.
58 (reduc_<reduc_op>_scal_<mode>): Allow all types.
59 (reduc_<reduc_op>_scal_v64di): Delete.
60 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
61 (*plus_carry_dpp_shr_v64si): Change to ...
62 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
63 (mov_from_lane63_v64di): Change to ...
64 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
65 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
66 Support UNSPEC_MOV_DPP_SHR output formats.
67 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
68 Add "use_extends" reductions.
69 (print_operand_address): Add 'I' and 'U' codes.
70 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
72 2020-03-02 Martin Liska <mliska@suse.cz>
74 * lto-wrapper.c: Fix typo in comment about
77 2020-03-01 Martin Sebor <msebor@redhat.com>
80 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
82 2020-03-01 Martin Sebor <msebor@redhat.com>
85 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
86 of a pointer in the outermost ADDR_EXPRs.
88 2020-02-28 Jeff Law <law@redhat.com>
90 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
91 * config/v850/v850.c (v850_asm_trampoline_template): Update
94 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
97 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
100 2020-02-28 Martin Liska <mliska@suse.cz>
103 * configure.ac: Improve detection of ld_date by requiring
104 either two dashes or none.
105 * configure: Regenerate.
107 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
109 PR rtl-optimization/93564
110 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
111 do not honor reg alloc order.
113 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
116 * config/aarch64/aarch64.c (aarch64_override_options): Fix
117 misleading warning string.
119 2020-02-27 Martin Sebor <msebor@redhat.com>
121 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
123 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
126 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
127 Split the insn into two parts. This insn only does variable
128 extract from a register.
129 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
130 variable extract from memory.
131 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
132 only does variable extract from a register.
133 (vsx_extract_v4sf_var_load): New insn, do variable extract from
135 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
136 into two parts. This insn only does variable extract from a
138 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
139 do variable extract from memory.
141 2020-02-27 Martin Jambor <mjambor@suse.cz>
142 Feng Xue <fxue@os.amperecomputing.com>
145 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
146 new function calls_same_node_or_its_all_contexts_clone_p.
147 (cgraph_edge_brings_value_p): Use it.
148 (cgraph_edge_brings_value_p): Likewise.
149 (self_recursive_pass_through_p): Return false if caller is a clone.
150 (self_recursive_agg_pass_through_p): Likewise.
152 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
155 * alias.c (ends_tbaa_access_path_p): Break out from ...
156 (component_uses_parent_alias_set_from): ... here.
157 * alias.h (ends_tbaa_access_path_p): Declare.
158 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
159 handle trailing arrays past end of tbaa access path.
160 (aliasing_component_refs_p): ... here; likewise.
161 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
162 path; disambiguate also past end of it.
163 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
166 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
168 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
169 beginning of the file.
170 (vcreate_bf16, vcombine_bf16): New.
171 (vdup_n_bf16, vdupq_n_bf16): New.
172 (vdup_lane_bf16, vdup_laneq_bf16): New.
173 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
174 (vduph_lane_bf16, vduph_laneq_bf16): New.
175 (vset_lane_bf16, vsetq_lane_bf16): New.
176 (vget_lane_bf16, vgetq_lane_bf16): New.
177 (vget_high_bf16, vget_low_bf16): New.
178 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
179 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
180 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
181 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
182 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
183 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
184 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
185 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
186 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
187 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
188 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
189 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
190 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
191 (vreinterpretq_bf16_p128): New.
192 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
193 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
194 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
195 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
196 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
197 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
198 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
199 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
200 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
201 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
202 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
203 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
204 (vreinterpretq_p128_bf16): New.
205 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
207 (V_elem_l): Likewise.
210 (V_DOUBLE): Likewise.
211 (VDQX): Add V4BF and V8BF.
212 (V_two_elem, V_three_elem, V_four_elem): Likewise.
215 (V_double_vector_mode): Likewise.
216 (V_cmp_result): Likewise.
217 (V_uf_sclr): Likewise.
218 (V_sz_elem): Likewise.
219 (Is_d_reg): Likewise.
220 (V_mode_nunits): Likewise.
221 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
223 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
225 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
226 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
227 (<expander><mode>3<exec>): Likewise.
228 (<expander><mode>3): New.
229 (v<expander><mode>3): New.
230 (<expander><mode>3): New.
231 (<expander><mode>3<exec>): Rename to ...
232 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
233 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
235 2020-02-27 Alexandre Oliva <oliva@adacore.com>
237 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
240 2020-02-27 Richard Biener <rguenther@suse.de>
242 PR tree-optimization/93508
243 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
244 non-_CHK variants. Valueize their length arguments.
246 2020-02-27 Richard Biener <rguenther@suse.de>
248 PR tree-optimization/93953
249 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
250 to the hash-map entry.
252 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
254 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
256 2020-02-27 Mark Williams <mwilliams@fb.com>
258 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
259 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
260 -ffile-prefix-map and -fmacro-prefix-map.
261 * lto-streamer-out.c: Include file-prefix-map.h.
262 (lto_output_location): Remap the file part of locations.
264 2020-02-27 Jakub Jelinek <jakub@redhat.com>
267 * gimplify.c (gimplify_init_constructor): Don't promote readonly
268 DECL_REGISTER variables to TREE_STATIC.
270 PR tree-optimization/93582
271 PR tree-optimization/93945
272 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
273 non-zero INTEGER_CST second argument and ref->offset or ref->size
274 not a multiple of BITS_PER_UNIT.
276 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
278 * doc/install.texi (Binaries): Update description of BullFreeware.
280 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
284 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
285 C++ Language Options, Warning Options, and Static Analyzer
286 Options lists. Document negative form of options enabled by
287 default. Move some things around to more accurately sort
288 warnings by category.
289 (C++ Dialect Options, Warning Options, Static Analyzer
290 Options): Document negative form of options when enabled by
291 default. Move some things around to more accurately sort
292 warnings by category. Add some missing index entries.
295 2020-02-26 Carl Love <cel@us.ibm.com>
298 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
299 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
300 for the vector unsigned short arguments. It is also listed as the
301 name of the built-in for arguments vector unsigned short,
302 vector unsigned int and vector unsigned long long built-ins. The
303 name of the builtins for these arguments should be:
304 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
305 __builtin_crypto_vpmsumd respectively.
307 2020-02-26 Richard Biener <rguenther@suse.de>
309 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
310 and load permutation.
312 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
315 * optabs-tree.c (supportable_convert_operation): Reject types with
318 2020-02-26 David Malcolm <dmalcolm@redhat.com>
320 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
322 2020-02-26 Jakub Jelinek <jakub@redhat.com>
324 PR tree-optimization/93820
325 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
326 argument to ALL_INTEGER_CST_P boolean.
327 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
328 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
329 adjacent INTEGER_CST store into merged_store->only_constants like
332 2020-02-25 Jakub Jelinek <jakub@redhat.com>
335 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
337 * cfghooks.c (verify_flow_info): Likewise.
338 * predict.c (combine_predictions_for_bb): Likewise.
339 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
340 sucessor -> successor.
341 (find_traces_1_round): Fix comment typo, destinarion -> destination.
342 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
344 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
345 message typo, sucessors -> successors.
347 2020-02-25 Martin Sebor <msebor@redhat.com>
349 * doc/extend.texi (attribute access): Correct an example.
351 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
353 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
355 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
357 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
358 (VD): Enable for V4BF.
360 (VQ): Enable for V8BF.
363 (VDBL, Vdbl): Add V4BF.
364 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
365 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
366 (bfloat16x8x2_t): Likewise.
367 (bfloat16x4x3_t): Likewise.
368 (bfloat16x8x3_t): Likewise.
369 (bfloat16x4x4_t): Likewise.
370 (bfloat16x8x4_t): Likewise.
371 (vcombine_bf16): New.
372 (vld1_bf16, vld1_bf16_x2): New.
373 (vld1_bf16_x3, vld1_bf16_x4): New.
374 (vld1q_bf16, vld1q_bf16_x2): New.
375 (vld1q_bf16_x3, vld1q_bf16_x4): New.
376 (vld1_lane_bf16): New.
377 (vld1q_lane_bf16): New.
378 (vld1_dup_bf16): New.
379 (vld1q_dup_bf16): New.
382 (vld2_dup_bf16): New.
383 (vld2q_dup_bf16): New.
386 (vld3_dup_bf16): New.
387 (vld3q_dup_bf16): New.
390 (vld4_dup_bf16): New.
391 (vld4q_dup_bf16): New.
392 (vst1_bf16, vst1_bf16_x2): New.
393 (vst1_bf16_x3, vst1_bf16_x4): New.
394 (vst1q_bf16, vst1q_bf16_x2): New.
395 (vst1q_bf16_x3, vst1q_bf16_x4): New.
396 (vst1_lane_bf16): New.
397 (vst1q_lane_bf16): New.
405 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
407 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
408 (VALL_F16): Likewise.
409 (VALLDI_F16): Likewise.
412 (vswap_width_name): Likewise.
413 (VSWAP_WIDTH): Likewise.
417 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
418 (vget_lane_bf16, vgetq_lane_bf16): New.
420 (vdup_n_bf16, vdupq_n_bf16): New.
421 (vdup_lane_bf16, vdup_laneq_bf16): New.
422 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
423 (vduph_lane_bf16, vduph_laneq_bf16): New.
424 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
425 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
426 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
427 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
428 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
429 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
430 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
431 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
432 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
433 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
434 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
435 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
436 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
437 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
438 (vreinterpretq_bf16_p128): New.
439 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
440 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
441 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
442 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
443 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
444 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
445 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
446 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
447 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
448 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
449 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
450 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
451 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
452 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
453 (vreinterpretq_p128_bf16): New.
455 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
457 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
458 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
459 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
460 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
461 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
462 * config/arm/iterators.md (VSF2BF): New attribute.
463 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
464 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
465 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
467 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
469 * config/arm/arm.md (required_for_purecode): New attribute.
470 (enabled): Handle required_for_purecode.
471 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
472 work with -mpure-code.
474 2020-02-25 Jakub Jelinek <jakub@redhat.com>
476 PR rtl-optimization/93908
477 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
480 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
482 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
484 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
486 * doc/install.texi (--enable-checking): Adjust wording.
488 2020-02-25 Richard Biener <rguenther@suse.de>
490 PR tree-optimization/93868
491 * tree-vect-slp.c (slp_copy_subtree): New function.
492 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
493 re-arranging stmts in it.
495 2020-02-25 Jakub Jelinek <jakub@redhat.com>
498 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
499 dummy function and remove it at the end.
502 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
503 paramter -> parameter.
504 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
505 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
507 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
509 * doc/install.texi (--enable-checking): Properly document current
511 (--enable-stage1-checking): Minor clarification about bootstrap.
513 2020-02-24 David Malcolm <dmalcolm@redhat.com>
516 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
517 -fanalyzer-checker=taint is also required.
518 (-fanalyzer-checker=): Note that providing this option enables the
519 given checker, and doing so may be required for checkers that are
522 2020-02-24 David Malcolm <dmalcolm@redhat.com>
524 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
525 significant control flow events; add a "3" which shows all
526 control flow events; the old "3" becomes "4".
528 2020-02-24 Jakub Jelinek <jakub@redhat.com>
530 PR tree-optimization/93582
531 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
532 pd.offset and pd.size to be counted in bits rather than bytes, add
533 support for maxsizei that is not a multiple of BITS_PER_UNIT and
534 handle bitfield stores and loads.
535 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
536 uncomparable quantities - bytes vs. bits. Allow push_partial_def
537 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
538 pd.offset/pd.size to be counted in bits rather than bytes.
539 Formatting fix. Rename shadowed len variable to buflen.
541 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
542 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
545 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
546 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
547 * opts-common.c (parse_options_from_collect_gcc_options): New function.
548 (prepend_xassembler_to_collect_as_options): Likewise.
549 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
550 (prepend_xassembler_to_collect_as_options): Likewise.
551 * lto-opts.c (lto_write_options): Stream assembler options
552 in COLLECT_AS_OPTIONS.
553 * lto-wrapper.c (xassembler_options_error): New static variable.
554 (get_options_from_collect_gcc_options): Move parsing options code to
555 parse_options_from_collect_gcc_options and call it.
556 (merge_and_complain): Validate -Xassembler options.
557 (append_compiler_options): Handle OPT_Xassembler.
558 (run_gcc): Append command line -Xassembler options to
560 * doc/invoke.texi: Add documentation about using Xassembler
563 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
565 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
567 (riscv_rtx_costs): Update cost model for LTGT.
569 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
571 PR rtl-optimization/93564
572 * ira-color.c (struct update_cost_queue_elem): New member start.
573 (queue_update_cost, get_next_update_cost): Add new arg start.
574 (allocnos_conflict_p): New function.
575 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
576 Add checking conflicts with allocnos_conflict_p.
577 (update_costs_from_prefs, restore_costs_from_copies): Adjust
578 update_costs_from_allocno calls.
579 (update_conflict_hard_regno_costs): Add checking conflicts with
580 allocnos_conflict_p. Adjust calls of queue_update_cost and
581 get_next_update_cost.
582 (assign_hard_reg): Adjust calls of queue_update_cost. Add
584 (bucket_allocno_compare_func): Restore previous version.
586 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
588 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
589 double-word size when handling aggregate return values.
590 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
591 that homogeneous SFmode and DFmode aggregates are passed and returned
592 in general registers.
594 2020-02-21 Jakub Jelinek <jakub@redhat.com>
597 * opts.c (print_filtered_help): Translate help before appending
598 messages to it rather than after that.
600 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
602 PR rtl-optimization/PR92989
603 * lra-lives.c (process_bb_lives): Restore the original order
604 of the bb liveness update. Call make_hard_regno_dead for each
605 register clobbered at the start of an EH receiver.
607 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
610 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
611 self-recursively generated.
613 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
616 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
619 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
621 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
622 Document new target supports option.
624 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
626 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
627 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
628 * config/arm/iterators.md (MATMUL): New iterator.
629 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
630 (mmla_sfx): New attribute.
631 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
632 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
633 (UNSPEC_MATMUL_US): New.
635 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
637 * config/arm/arm.md: Prevent scalar shifts from being used when big
640 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
641 Richard Biener <rguenther@suse.de>
643 PR tree-optimization/93586
644 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
645 after mismatched array refs; do not sure type size information to
646 recover from unmatched referneces with !flag_strict_aliasing_p.
648 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
650 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
651 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
652 (scatter_store<mode>): Rename to ...
653 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
654 (scatter<mode>_exec): Delete. Move contents ...
655 (mask_scatter_store<mode>): ... here, and rename that to ...
656 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
657 Remove mode conversion.
658 (mask_gather_load<mode>): Rename to ...
659 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
660 Remove mode conversion.
661 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
663 2020-02-21 Martin Jambor <mjambor@suse.cz>
665 PR tree-optimization/93845
666 * tree-sra.c (verify_sra_access_forest): Only test access size of
669 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
671 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
672 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
673 (addv64di3_exec): Likewise.
674 (subv64di3): Likewise.
675 (subv64di3_exec): Likewise.
676 (addv64di3_zext): Likewise.
677 (addv64di3_zext_exec): Likewise.
678 (addv64di3_zext_dup): Likewise.
679 (addv64di3_zext_dup_exec): Likewise.
680 (addv64di3_zext_dup2): Likewise.
681 (addv64di3_zext_dup2_exec): Likewise.
682 (addv64di3_sext_dup2): Likewise.
683 (addv64di3_sext_dup2_exec): Likewise.
684 (<expander>v64di3): Likewise.
685 (<expander>v64di3_exec): Likewise.
686 (*<reduc_op>_dpp_shr_v64di): Likewise.
687 (*plus_carry_dpp_shr_v64di): Likewise.
688 * config/gcn/gcn.md (adddi3): Likewise.
689 (addptrdi3): Likewise.
690 (<expander>di3): Likewise.
692 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
694 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
696 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
698 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
699 support. Use aarch64_emit_mult instead of emitting multiplication
700 instructions directly.
701 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
702 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
704 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
706 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
707 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
708 instead of emitting multiplication instructions directly.
709 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
710 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
711 (@aarch64_frecps<mode>): New expanders.
713 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
715 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
716 on and produce uint64_ts rather than ints.
717 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
718 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
720 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
722 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
723 an unused xmsk register when handling approximate rsqrt.
725 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
727 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
728 flag_finite_math_only condition.
730 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
733 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
734 to destination operand for shufps alternative.
735 (*vec_extractv2si_1): Ditto.
737 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
740 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
743 2020-02-20 Martin Liska <mliska@suse.cz>
746 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
748 2020-02-20 Martin Liska <mliska@suse.cz>
751 * common/config/avr/avr-common.c: Remote trailing "|".
753 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
755 * collect2.c (maybe_run_lto_and_relink): Fix typo in
758 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
760 PR tree-optimization/93767
761 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
762 access-size bias from the offset calculations for negative strides.
764 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
766 * collect2.c (c_file, o_file): Make const again.
767 (ldout,lderrout, dump_ld_file): Remove.
768 (tool_cleanup): Avoid calling not signal-safe functions.
769 (maybe_run_lto_and_relink): Avoid possible signal handler
770 access to unintialzed memory (lto_o_files).
771 (main): Avoid leaking temp files in $TMPDIR.
772 Initialize c_file/o_file with concat, which avoids exposing
773 uninitialized memory to signal handler, which calls unlink(!).
774 Avoid calling maybe_unlink when the main function returns,
775 since the atexit handler is already doing this.
776 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
778 2020-02-19 Martin Jambor <mjambor@suse.cz>
780 PR tree-optimization/93776
781 * tree-sra.c (create_access): Do not create zero size accesses.
782 (get_access_for_expr): Do not search for zero sized accesses.
784 2020-02-19 Martin Jambor <mjambor@suse.cz>
786 PR tree-optimization/93667
787 * tree-sra.c (scalarizable_type_p): Return false if record fields
788 do not follow wach other.
790 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
792 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
793 rather than fmv.x.s/fmv.s.x.
795 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
797 * config/aarch64/aarch64-simd-builtins.def
798 (intrinsic_vec_smult_lo_): New.
799 (intrinsic_vec_umult_lo_): Likewise.
800 (vec_widen_smult_hi_): Likewise.
801 (vec_widen_umult_hi_): Likewise.
802 * config/aarch64/aarch64-simd.md
803 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
804 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
805 (vmull_high_s16): Likewise.
806 (vmull_high_s32): Likewise.
807 (vmull_high_u8): Likewise.
808 (vmull_high_u16): Likewise.
809 (vmull_high_u32): Likewise.
810 (vmull_s8): Likewise.
811 (vmull_s16): Likewise.
812 (vmull_s32): Likewise.
813 (vmull_u8): Likewise.
814 (vmull_u16): Likewise.
815 (vmull_u32): Likewise.
817 2020-02-18 Martin Liska <mliska@suse.cz>
819 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
820 bootstrap by missing removal of invalid sanity check.
822 2020-02-18 Martin Liska <mliska@suse.cz>
825 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
826 Always compare LHS of gimple_assign.
828 2020-02-18 Martin Liska <mliska@suse.cz>
831 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
832 and return type of functions.
833 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
834 Drop MALLOC attribute for void functions.
835 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
836 malloc_state for a new VOID clone.
838 2020-02-18 Martin Liska <mliska@suse.cz>
841 * common.opt: Add -fprofile-reproducibility.
842 * doc/invoke.texi: Document it.
843 * value-prof.c (dump_histogram_value):
844 Document and support behavior for counters[0]
845 being a negative value.
846 (get_nth_most_common_value): Handle negative
847 counters[0] in respect to flag_profile_reproducible.
849 2020-02-18 Jakub Jelinek <jakub@redhat.com>
852 * cgraph.c (verify_speculative_call): Use speculative_id instead of
853 speculative_uid in messages. Remove trailing whitespace from error
854 message. Use num_speculative_call_targets instead of
855 num_speculative_targets in a message.
856 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
857 edge messages and stmt instead of cal_stmt in reference message.
859 PR tree-optimization/93780
860 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
861 before calling build_vector_type.
862 (execute_update_addresses_taken): Likewise.
865 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
866 typo, functoin -> function.
867 * tree.c (free_lang_data_in_decl): Fix comment typo,
868 functoin -> function.
869 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
871 2020-02-17 David Malcolm <dmalcolm@redhat.com>
873 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
875 (print_option_information): Don't call get_option_url if URLs
878 2020-02-17 Alexandre Oliva <oliva@adacore.com>
880 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
881 handling of register_common-less targets.
883 2020-02-17 Martin Liska <mliska@suse.cz>
886 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
888 2020-02-17 Martin Liska <mliska@suse.cz>
891 * config/rs6000/rs6000.c (rs6000_option_override_internal):
894 2020-02-17 Martin Liska <mliska@suse.cz>
897 * config/rx/elf.opt: Fix typo.
899 2020-02-17 Richard Biener <rguenther@suse.de>
902 * opts-global.c (print_ignored_options): Use inform and
905 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
908 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
910 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
913 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
914 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
916 2020-02-15 Jason Merrill <jason@redhat.com>
918 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
920 2020-02-15 Jakub Jelinek <jakub@redhat.com>
922 PR tree-optimization/93744
923 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
924 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
925 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
926 sure @2 in the first and @1 in the other patterns has no side-effects.
928 2020-02-15 David Malcolm <dmalcolm@redhat.com>
929 Bernd Edlinger <bernd.edlinger@hotmail.de>
933 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
934 * configure.ac (--with-diagnostics-urls): New configuration
935 option, based on --with-diagnostics-color.
936 (DIAGNOSTICS_URLS_DEFAULT): New define.
937 * config.h: Regenerate.
938 * configure: Regenerate.
939 * diagnostic.c (diagnostic_urls_init): Handle -1 for
940 DIAGNOSTICS_URLS_DEFAULT from configure-time
941 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
942 and TERM_URLS environment variable.
943 * diagnostic-url.h (diagnostic_url_format): New enum type.
944 (diagnostic_urls_enabled_p): rename to...
945 (determine_url_format): ... this, and change return type.
946 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
947 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
948 the linux console, and mingw.
949 (diagnostic_urls_enabled_p): rename to...
950 (determine_url_format): ... this, and adjust.
951 * pretty-print.h (pretty_printer::show_urls): rename to...
952 (pretty_printer::url_format): ... this, and change to enum.
953 * pretty-print.c (pretty_printer::pretty_printer,
954 pp_begin_url, pp_end_url, test_urls): Adjust.
955 * doc/install.texi (--with-diagnostics-urls): Document the new
956 configuration option.
957 (--with-diagnostics-color): Document the existing interaction
958 with GCC_COLORS better.
959 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
960 vindex reference. Update description of defaults based on the above.
961 (-fdiagnostics-color): Update description of how -fdiagnostics-color
962 interacts with GCC_COLORS.
964 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
967 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
968 conjunction with TARGET_GNU_TLS in early return.
970 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
972 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
973 the mode is not wider than UNITS_PER_WORD.
975 2020-02-14 Martin Jambor <mjambor@suse.cz>
977 PR tree-optimization/93516
978 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
979 access of the same type as the parent.
980 (propagate_subaccesses_from_lhs): Likewise.
982 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
985 * config/i386/avx512vbmi2intrin.h
986 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
987 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
988 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
989 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
990 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
991 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
992 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
993 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
994 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
995 of lacking a closing parenthesis.
996 * config/i386/avx512vbmi2vlintrin.h
997 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
998 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
999 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
1000 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
1001 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
1002 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
1003 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
1004 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
1005 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
1006 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
1007 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
1008 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
1009 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
1010 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
1011 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
1012 _mm_shldi_epi32, _mm_mask_shldi_epi32,
1013 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
1014 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
1016 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
1019 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
1020 the target function entry.
1022 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1024 * common/config/arc/arc-common.c (arc_option_optimization_table):
1025 Disable if-conversion step when optimized for size.
1027 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1029 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
1030 R12-R15 are always in ARCOMPACT16_REGS register class.
1031 * config/arc/arc.opt (mq-class): Deprecate.
1032 * config/arc/constraint.md ("q"): Remove dependency on mq-class
1034 * doc/invoke.texi (mq-class): Update text.
1035 * common/config/arc/arc-common.c (arc_option_optimization_table):
1038 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1040 * config/arc/arc.c (arc_insn_cost): New function.
1041 (TARGET_INSN_COST): Define.
1042 * config/arc/arc.md (cost): New attribute.
1043 (add_n): Use arc_nonmemory_operand.
1044 (ashlsi3_insn): Likewise, also update constraints.
1045 (ashrsi3_insn): Likewise.
1046 (rotrsi3): Likewise.
1047 (add_shift): Likewise.
1048 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1050 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1052 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1054 (umulsidi_600): Likewise.
1056 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1059 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1060 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1061 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1062 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1063 pass __A to the builtin followed by __W instead of __A followed by
1065 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1066 _mm512_mask_popcnt_epi64): Likewise.
1067 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1068 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1069 _mm256_mask_popcnt_epi64): Likewise.
1071 PR tree-optimization/93582
1072 * fold-const.h (shift_bytes_in_array_left,
1073 shift_bytes_in_array_right): Declare.
1074 * fold-const.c (shift_bytes_in_array_left,
1075 shift_bytes_in_array_right): New function, moved from
1076 gimple-ssa-store-merging.c, no longer static.
1077 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1078 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1079 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1080 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1081 shift_bytes_in_array.
1082 (verify_shift_bytes_in_array): Rename to ...
1083 (verify_shift_bytes_in_array_left): ... this. Use
1084 shift_bytes_in_array_left instead of shift_bytes_in_array.
1085 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1086 instead of verify_shift_bytes_in_array.
1087 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1088 / native_interpret_expr where the store covers all needed bits,
1089 punt on PDP-endian, otherwise allow all involved offsets and sizes
1090 not to be byte-aligned.
1093 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1094 use const_0_to_255_operand predicate instead of immediate_operand.
1095 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1096 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1097 vgf2p8affineinvqb_<mode><mask_name>,
1098 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1099 const_0_to_255_operand predicated operands.
1101 2020-02-12 Jeff Law <law@redhat.com>
1103 * config/h8300/h8300.md (comparison shortening peepholes): Use
1104 a mode iterator to merge the HImode and SImode peepholes.
1106 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1109 * real.c (is_even): Make static. Function comment fix.
1110 (is_halfway_below): Make static, don't assert R is not inf/nan,
1111 instead return false for those. Small formatting fixes.
1113 2020-02-12 Martin Sebor <msebor@redhat.com>
1116 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1117 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1118 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1119 (strlen_check_and_optimize_call): Adjust callee name.
1121 2020-02-12 Jeff Law <law@redhat.com>
1123 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1124 (and (xor)) variant. Combine other two into single peephole.
1126 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1128 PR rtl-optimization/93565
1129 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1131 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1133 * config/aarch64/aarch64-simd.md
1134 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1135 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1136 generating separate ADDV and zero_extend patterns.
1137 * config/aarch64/iterators.md (VDQV_E): New iterator.
1139 2020-02-12 Jeff Law <law@redhat.com>
1141 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1142 expanders, splits, etc.
1143 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1144 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1145 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1146 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1147 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1148 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1150 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1152 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1155 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1156 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1157 TARGET_AVX512DQ from condition.
1158 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1159 instead of <mask_mode512bit_condition> in condition. If
1160 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1162 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1165 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1168 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1170 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1172 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1173 where strlen is more legible.
1174 (rs6000_builtin_vectorized_libmass): Ditto.
1175 (rs6000_print_options_internal): Ditto.
1177 2020-02-11 Martin Sebor <msebor@redhat.com>
1179 PR tree-optimization/93683
1180 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1182 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1184 * config/rs6000/predicates.md (cint34_operand): Rename the
1185 -mprefixed-addr option to be -mprefixed.
1186 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1187 the -mprefixed-addr option to be -mprefixed.
1188 (OTHER_FUTURE_MASKS): Likewise.
1189 (POWERPC_MASKS): Likewise.
1190 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1191 the -mprefixed-addr option to be -mprefixed. Change error
1192 messages to refer to -mprefixed.
1193 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1195 (rs6000_legitimate_offset_address_p): Likewise.
1196 (rs6000_mode_dependent_address): Likewise.
1197 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1198 "-mprefixed" for target attributes and pragmas.
1199 (address_to_insn_form): Rename the -mprefixed-addr option to be
1201 (rs6000_adjust_insn_length): Likewise.
1202 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1203 -mprefixed-addr option to be -mprefixed.
1204 (ASM_OUTPUT_OPCODE): Likewise.
1205 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1206 -mprefixed-addr option to be -mprefixed.
1207 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1208 -mprefixed-addr option to be prefixed. Change the option from
1209 being undocumented to being documented.
1210 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1211 -mprefixed option. Update the -mpcrel documentation to mention
1214 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1216 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1217 including FIRST_PSEUDO_REGISTER - 1.
1218 * ira-color.c (print_hard_reg_set): Ditto.
1220 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1222 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1223 (USTERNOP_QUALIFIERS): New define.
1224 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1225 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1226 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1227 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1228 * config/arm/arm_neon.h (vusdot_s32): New.
1229 (vusdot_lane_s32): New.
1230 (vusdotq_lane_s32): New.
1231 (vsudot_lane_s32): New.
1232 (vsudotq_lane_s32): New.
1233 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1234 * config/arm/iterators.md (DOTPROD_I8MM): New.
1235 (sup, opsuffix): Add <us/su>.
1236 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1237 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1239 2020-02-11 Richard Biener <rguenther@suse.de>
1241 PR tree-optimization/93661
1242 PR tree-optimization/93662
1243 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1245 * tree-sra.c (get_access_for_expr): Likewise.
1247 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1250 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1251 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1252 Change condition from TARGET_AVX2 to TARGET_AVX.
1254 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1257 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1258 argument of strncmp.
1260 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1262 Try to generate zero-based comparisons.
1263 * config/cris/cris.c (cris_reduce_compare): New function.
1264 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1265 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1266 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1268 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1271 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1272 in Thumb state and also as a destination in Arm state. Add T16
1275 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1277 * md.texi (Define Subst): Match closing paren in example.
1279 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1283 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1284 arguments of strncmp.
1286 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1289 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1290 but different source value.
1291 (adjust_callers_for_value_intersection): New function.
1292 (gather_edges_for_value): Adjust order of callers to let a
1293 non-self-recursive caller be the first element.
1294 (self_recursive_pass_through_p): Add a new parameter "simple", and
1295 check generalized self-recursive pass-through jump function.
1296 (self_recursive_agg_pass_through_p): Likewise.
1297 (find_more_scalar_values_for_callers_subset): Compute value from
1298 pass-through jump function for self-recursive.
1299 (intersect_with_plats): Cleanup previous implementation code for value
1300 itersection with self-recursive call edge.
1301 (intersect_with_agg_replacements): Likewise.
1302 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1303 function for self-recursive call edge. Cleanup previous implementation
1304 code for value intersection with self-recursive call edge.
1305 (decide_whether_version_node): Remove dead callers and adjust order
1306 to let a non-self-recursive caller be the first element.
1308 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1310 * recog.c: Move pass_split_before_sched2 code in front of
1311 pass_split_before_regstack.
1312 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1313 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1314 (rest_of_handle_split_before_sched2): Remove.
1315 (pass_split_before_sched2::execute): Unconditionally call
1317 (enable_split_before_sched2): New function.
1318 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1319 (pass_split_before_regstack::gate): Ditto.
1320 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1321 Update name check for renamed split4 pass.
1322 * config/sh/sh.c (register_sh_passes): Update pass insertion
1323 point for renamed split4 pass.
1325 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1327 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1328 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1329 copying them around between host and target.
1331 2020-02-08 Andrew Pinski <apinski@marvell.com>
1334 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1335 STRICT_ALIGNMENT also.
1337 2020-02-08 Jim Wilson <jimw@sifive.com>
1340 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1342 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1343 Jakub Jelinek <jakub@redhat.com>
1346 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1347 xmm16-xmm31 call-used even in 64-bit ms-abi.
1349 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1351 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1352 (simd_ummla, simd_usmmla): Likewise.
1353 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1354 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1355 (vusmmlaq_s32): New.
1357 2020-02-07 Richard Biener <rguenther@suse.de>
1360 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1361 skipping unreachable regions.
1362 (optimize_inline_calls): Skip folding stmts when we didn't
1365 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1368 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1369 Don't return aggregates with only SFmode and DFmode in SSE
1371 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1373 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1376 * config/rs6000/rs6000-logue.c
1377 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1378 if it fails, move rs into end_addr and retry. Add
1379 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1380 the insn pattern doesn't describe well what exactly happens to
1384 * config/i386/predicates.md (avx_identity_operand): Remove.
1385 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1386 (avx_<castmode><avxsizesuffix>_<castmode>,
1387 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1388 a VEC_CONCAT of the operand and UNSPEC_CAST.
1389 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1390 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1394 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1395 recog_data.insn if distance_non_agu_define changed it.
1397 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1400 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1401 we only had X-FORM (reg+reg) addressing for vectors. Also before
1402 ISA 3.0, we only had X-FORM addressing for scalars in the
1403 traditional Altivec registers.
1405 2020-02-06 <zhongyunde@huawei.com>
1406 Vladimir Makarov <vmakarov@redhat.com>
1408 PR rtl-optimization/93561
1409 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1410 hard register range.
1412 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1414 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1417 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1419 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1420 where the low and the high 32 bits are equal to each other specially,
1421 with an rldimi instruction.
1423 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1425 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1427 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1429 * config/arm/arm-tables.opt: Regenerate.
1431 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1434 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1435 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1436 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1438 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1440 PR rtl-optimization/87763
1441 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1443 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1445 * config/aarch64/aarch64-simd-builtins.def
1446 (bfmlaq): New built-in function.
1447 (bfmlalb): New built-in function.
1448 (bfmlalt): New built-in function.
1449 (bfmlalb_lane): New built-in function.
1450 (bfmlalt_lane): New built-in function.
1451 * config/aarch64/aarch64-simd.md
1452 (aarch64_bfmmlaqv4sf): New pattern.
1453 (aarch64_bfmlal<bt>v4sf): New pattern.
1454 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1455 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1456 (vbfmlalbq_f32): New intrinsic.
1457 (vbfmlaltq_f32): New intrinsic.
1458 (vbfmlalbq_lane_f32): New intrinsic.
1459 (vbfmlaltq_lane_f32): New intrinsic.
1460 (vbfmlalbq_laneq_f32): New intrinsic.
1461 (vbfmlaltq_laneq_f32): New intrinsic.
1462 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1463 (bt): New int attribute.
1465 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1467 * config/i386/i386.md (*pushtf): Emit "#" instead of
1468 calling gcc_unreachable in insn output.
1471 (*pushsf_rex64): Ditto for alternatives other than 1.
1472 (*pushsf): Ditto for alternatives other than 1.
1474 2020-02-06 Martin Liska <mliska@suse.cz>
1476 PR gcov-profile/91971
1477 PR gcov-profile/93466
1478 * coverage.c (coverage_init): Revert mangling of
1479 path into filename. It can lead to huge filename length.
1480 Creation of subfolders seem more natural.
1482 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1485 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1486 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1487 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1489 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1492 * config/i386/predicates.md (avx_identity_operand): New predicate.
1493 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1494 define_insn_and_split.
1497 * omp-low.c (use_pointer_for_field): For nested constructs, also
1498 look for map clauses on target construct.
1499 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1500 taskreg_nesting_level.
1503 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1504 shared clause, call omp_notice_variable on outer context if any.
1506 2020-02-05 Jason Merrill <jason@redhat.com>
1509 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1510 non-zero address even if weak and not yet defined.
1512 2020-02-05 Martin Sebor <msebor@redhat.com>
1514 PR tree-optimization/92765
1515 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1516 * tree-ssa-strlen.c (compute_string_length): Remove.
1517 (determine_min_objsize): Remove.
1518 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1519 Avoid using type size as the upper bound on string length.
1520 (handle_builtin_string_cmp): Add an argument. Adjust.
1521 (strlen_check_and_optimize_call): Pass additional argument to
1522 handle_builtin_string_cmp.
1524 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1526 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1527 (*pushdi2_rex64 peephole2): Unconditionally split after
1529 (*ashl<mode>3_doubleword): Ditto.
1530 (*<shift_insn><mode>3_doubleword): Ditto.
1532 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1535 * config/rs6000/rs6000.c (get_vector_offset): Fix
1537 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1539 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1541 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1544 (Special Functions for Debugging the Analyzer): Update description
1545 of __analyzer_dump_exploded_nodes.
1547 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1550 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1551 include sets and not clobbers in the vzeroupper pattern.
1552 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1553 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1554 (*avx_vzeroupper_1): New define_insn_and_split.
1557 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1558 don't run when !optimize.
1559 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1562 2020-02-05 Richard Biener <rguenther@suse.de>
1565 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1566 checks before matching calls.
1568 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1570 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1571 function comment typo.
1574 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1575 simd_clone_create failed when i == 0, adjust clone->nargs by
1578 2020-02-05 Martin Liska <mliska@suse.cz>
1581 * doc/invoke.texi: Document that one should
1582 not combine ASLR and -fpch.
1584 2020-02-04 Richard Biener <rguenther@suse.de>
1586 PR tree-optimization/93538
1587 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1589 2020-02-04 Richard Biener <rguenther@suse.de>
1591 PR tree-optimization/91123
1592 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1593 (vn_walk_cb_data::last_vuse): New member.
1594 (vn_walk_cb_data::saved_operands): Likewsie.
1595 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1596 (vn_walk_cb_data::push_partial_def): Use finish.
1597 (vn_reference_lookup_2): Update last_vuse and use finish if
1598 we've saved operands.
1599 (vn_reference_lookup_3): Use finish and update calls to
1600 push_partial_defs everywhere. When translating through
1601 memcpy or aggregate copies save off operands and alias-set.
1602 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1603 operation for redundant store removal.
1605 2020-02-04 Richard Biener <rguenther@suse.de>
1607 PR tree-optimization/92819
1608 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1609 generating more stmts than before.
1611 2020-02-04 Martin Liska <mliska@suse.cz>
1613 * config/arm/arm.c (arm_gen_far_branch): Move the function
1614 outside of selftests.
1616 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1618 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1619 function to adjust PC-relative vector addresses.
1620 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1621 handle vectors with PC-relative addresses.
1623 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1625 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1627 (hard_reg_and_mode_to_addr_mask): Delete.
1628 (rs6000_adjust_vec_address): If the original vector address
1629 was REG+REG or REG+OFFSET and the element is not zero, do the add
1630 of the elements in the original address before adding the offset
1631 for the vector element. Use address_to_insn_form to validate the
1632 address using the register being loaded, rather than guessing
1633 whether the address is a DS-FORM or DQ-FORM address.
1635 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1637 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1638 to calculate the offset in memory from the start of a vector of a
1639 particular element. Add code to keep the element number in
1640 bounds if the element number is variable.
1641 (rs6000_adjust_vec_address): Move calculation of offset of the
1642 vector element to get_vector_offset.
1643 (rs6000_split_vec_extract_var): Do not do the initial AND of
1644 element here, move the code to get_vector_offset.
1646 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1648 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1651 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1653 * config/rs6000/constraints.md: Improve documentation.
1655 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1658 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1659 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1661 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1663 * config.gcc: Remove "carrizo" support.
1664 * config/gcn/gcn-opts.h (processor_type): Likewise.
1665 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1666 * config/gcn/gcn.opt (gpu_type): Likewise.
1667 * config/gcn/t-omp-device: Likewise.
1669 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1672 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1673 * config/arm/arm.c (arm_gen_far_branch): New function
1675 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1677 2020-02-03 Julian Brown <julian@codesourcery.com>
1678 Tobias Burnus <tobias@codesourcery.com>
1680 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1682 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1685 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1686 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1689 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1691 PR rtl-optimization/91333
1692 * ira-color.c (struct allocno_color_data): Add member
1694 (init_allocno_threads): Set the member up.
1695 (bucket_allocno_compare_func): Add compare hard reg
1698 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1700 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1702 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1703 * config.in: Regenerated.
1704 * configure: Regenerated.
1705 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1706 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1707 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1709 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1711 * configure: Regenerate.
1713 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1715 PR rtl-optimization/91333
1716 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1717 reg preferences comparison up.
1719 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1721 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1722 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1723 aarch64-sve-builtins-base.h.
1724 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1725 aarch64-sve-builtins-base.cc.
1726 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1727 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1729 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1730 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1731 (svcvtnt): New functions.
1732 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1733 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1734 (svcvtnt): New functions.
1735 (svcvt): Add a form that converts f32 to bf16.
1736 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1737 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1739 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1740 Treat B as bfloat16_t.
1741 (ternary_bfloat_lane_base): New class.
1742 (ternary_bfloat_def): Likewise.
1743 (ternary_bfloat): New shape.
1744 (ternary_bfloat_lane_def): New class.
1745 (ternary_bfloat_lane): New shape.
1746 (ternary_bfloat_lanex2_def): New class.
1747 (ternary_bfloat_lanex2): New shape.
1748 (ternary_bfloat_opt_n_def): New class.
1749 (ternary_bfloat_opt_n): New shape.
1750 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1751 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1752 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1753 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1754 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1755 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1756 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1757 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1758 the pattern off the narrow mode instead of the wider one.
1759 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1760 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1761 (sve_fp_op): Handle them.
1762 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1763 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1765 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1767 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1768 * config/aarch64/aarch64-modes.def (BF): Move definition before
1769 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1770 (SVE_MODES): Handle BF modes.
1771 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1773 (aarch64_full_sve_mode): Likewise.
1774 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1776 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1777 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1778 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1779 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1781 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1783 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1784 (TYPES_all_data): Add bf16.
1785 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1786 (register_tuple_type): Increase buffer size.
1787 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1788 (bf16): New type suffix.
1789 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1790 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1791 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1792 Change type from all_data to all_arith.
1793 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1796 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1797 Matthew Malcomson <matthew.malcomson@arm.com>
1798 Richard Sandiford <richard.sandiford@arm.com>
1800 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1801 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1802 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1803 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1804 __ARM_FEATURE_MATMUL_FP64.
1805 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1806 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1807 be disabled at the same time.
1808 (f32mm): New extension.
1809 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1810 (AARCH64_FL_F64MM): Bump to the next bit up.
1811 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1812 (TARGET_SVE_F64MM): New macros.
1813 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1814 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1815 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1816 (UNSPEC_ZIP2Q): New unspeccs.
1817 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1818 (optab, sur, perm_insn): Handle the new unspecs.
1819 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1820 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1821 TARGET_SVE_F64MM instead of separate tests.
1822 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1823 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1824 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1825 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1826 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1827 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1828 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1829 (TYPES_s_signed): New macro.
1830 (TYPES_s_integer): Use it.
1831 (TYPES_d_float): New macro.
1832 (TYPES_d_data): Use it.
1833 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1834 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1835 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1836 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1837 (svmmla): New shape.
1838 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1839 template parameters.
1840 (ternary_resize2_lane_base): Likewise.
1841 (ternary_resize2_base): New class.
1842 (ternary_qq_lane_base): Likewise.
1843 (ternary_intq_uintq_lane_def): Likewise.
1844 (ternary_intq_uintq_lane): New shape.
1845 (ternary_intq_uintq_opt_n_def): New class
1846 (ternary_intq_uintq_opt_n): New shape.
1847 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1848 (ternary_uintq_intq_def): New class.
1849 (ternary_uintq_intq): New shape.
1850 (ternary_uintq_intq_lane_def): New class.
1851 (ternary_uintq_intq_lane): New shape.
1852 (ternary_uintq_intq_opt_n_def): New class.
1853 (ternary_uintq_intq_opt_n): New shape.
1854 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1855 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1856 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1857 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1859 (svdotprod_lane_impl): ...this new class.
1860 (svmmla_impl, svusdot_impl): New classes.
1861 (svdot_lane): Update to use svdotprod_lane_impl.
1862 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1863 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1865 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1866 function, with no types defined.
1867 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1868 AARCH64_FL_I8MM functions.
1869 (svmmla): New AARCH64_FL_F32MM function.
1870 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1871 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1872 AARCH64_FL_F64MM function.
1873 (REQUIRED_EXTENSIONS):
1875 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1877 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1880 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1882 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1883 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1884 (*movti_internal): Do not check for
1885 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1886 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1887 just after check for TARGET_AVX.
1888 (*movdf_internal): Ditto.
1889 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1890 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1891 * config/i386/sse.md (mov<mode>_internal): Only check
1892 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1893 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1894 (<sse>_andnot<mode>3<mask_name>): Move check for
1895 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1896 (<code><mode>3<mask_name>): Ditto.
1897 (*andnot<mode>3): Ditto.
1898 (*andnottf3): Ditto.
1899 (*<code><mode>3): Ditto.
1900 (*<code>tf3): Ditto.
1901 (*andnot<VI:mode>3): Remove
1902 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1903 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1904 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1905 (sse4_1_blendv<ssemodesuffix>): Ditto.
1906 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1907 Explain that tune applies to 128bit instructions only.
1909 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1911 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1912 to definition of hsa_kernel_description. Parse assembly to find SGPR
1913 and VGPR count of kernel and store in hsa_kernel_description.
1915 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1917 PR rtl-optimization/91838
1918 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1919 to truncate if allowed or reject combination.
1921 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1923 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1924 (find_inv_vars_cb): Likewise.
1926 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1928 * calls.c (special_function_p): Split out the check for DECL_NAME
1929 being non-NULL and fndecl being extern at file scope into a
1930 new maybe_special_function_p and call it. Drop check for fndecl
1931 being non-NULL that was after a usage of DECL_NAME (fndecl).
1932 * tree.h (maybe_special_function_p): New inline function.
1934 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1936 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1937 (mask_gather_load<mode>): ... here, and zero-initialize the
1939 (maskload<mode>di): Zero-initialize the destination.
1942 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1945 * doc/analyzer.texi (Limitations): Note that constraints on
1946 floating-point values are currently ignored.
1948 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1951 * symtab.c (symtab_node::noninterposable_alias): If localalias
1952 already exists, but is not usable, append numbers after it until
1953 a unique name is found. Formatting fix.
1956 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1959 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1961 * config/gcn/gcn.c (print_operand): Handle LTGT.
1962 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1964 2020-01-30 Richard Biener <rguenther@suse.de>
1966 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1967 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1969 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1971 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1972 without a DECL in .data.rel.ro.local.
1974 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1977 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1981 * config/i386/sse.md
1982 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1983 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1984 any_extend code iterator instead of always zero_extend.
1985 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1986 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1987 Use any_extend code iterator instead of always zero_extend.
1988 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1989 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1990 Use any_extend code iterator instead of always zero_extend.
1991 (*sse2_pmovmskb_ext): New define_insn.
1992 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1995 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1996 (*popcountsi2_zext_falsedep): New define_insn.
1998 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2000 * config.in: Regenerated.
2001 * configure: Regenerated.
2003 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
2006 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
2007 LLVM's assembler changed the default in version 9.
2009 2020-01-24 Jeff Law <law@redhat.com>
2011 PR tree-optimization/89689
2012 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
2014 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
2018 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2020 PR rtl-optimization/87763
2021 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2022 simplification to handle subregs as well as bare regs.
2023 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2025 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
2028 * ira.c (ira): Revert use of simplified LRA algorithm.
2030 2020-01-29 Martin Jambor <mjambor@suse.cz>
2032 PR tree-optimization/92706
2033 * tree-sra.c (struct access): Fields first_link, last_link,
2034 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
2035 next_rhs_queued and grp_rhs_queued respectively, new fields
2036 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2037 (struct assign_link): Field next renamed to next_rhs, new field
2038 next_lhs. Updated comment.
2039 (work_queue_head): Renamed to rhs_work_queue_head.
2040 (lhs_work_queue_head): New variable.
2041 (add_link_to_lhs): New function.
2042 (relink_to_new_repr): Also relink LHS lists.
2043 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2044 (add_access_to_lhs_work_queue): New function.
2045 (pop_access_from_work_queue): Renamed to
2046 pop_access_from_rhs_work_queue.
2047 (pop_access_from_lhs_work_queue): New function.
2048 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2050 (child_would_conflict_in_lacc): Renamed to
2051 child_would_conflict_in_acc. Adjusted parameter names.
2052 (create_artificial_child_access): New parameter set_grp_read, use it.
2053 (subtree_mark_written_and_enqueue): Renamed to
2054 subtree_mark_written_and_rhs_enqueue.
2055 (propagate_subaccesses_across_link): Renamed to
2056 propagate_subaccesses_from_rhs.
2057 (propagate_subaccesses_from_lhs): New function.
2058 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2061 2020-01-29 Martin Jambor <mjambor@suse.cz>
2063 PR tree-optimization/92706
2064 * tree-sra.c (struct access): Adjust comment of
2065 grp_total_scalarization.
2066 (find_access_in_subtree): Look for single children spanning an entire
2068 (scalarizable_type_p): Allow register accesses, adjust callers.
2069 (completely_scalarize): Remove function.
2070 (scalarize_elem): Likewise.
2071 (create_total_scalarization_access): Likewise.
2072 (sort_and_splice_var_accesses): Do not track total scalarization
2074 (analyze_access_subtree): New parameter totally, adjust to new meaning
2075 of grp_total_scalarization.
2076 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2077 (can_totally_scalarize_forest_p): New function.
2078 (create_total_scalarization_access): Likewise.
2079 (create_total_access_and_reshape): Likewise.
2080 (total_should_skip_creating_access): Likewise.
2081 (totally_scalarize_subtree): Likewise.
2082 (analyze_all_variable_accesses): Perform total scalarization after
2083 subaccess propagation using the new functions above.
2084 (initialize_constant_pool_replacements): Output initializers by
2085 traversing the access tree.
2087 2020-01-29 Martin Jambor <mjambor@suse.cz>
2089 * tree-sra.c (verify_sra_access_forest): New function.
2090 (verify_all_sra_access_forests): Likewise.
2091 (create_artificial_child_access): Set parent.
2092 (analyze_all_variable_accesses): Call the verifier.
2094 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2096 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2097 if called on indirect edge.
2098 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2099 speculative call if needed.
2101 2020-01-29 Richard Biener <rguenther@suse.de>
2103 PR tree-optimization/93428
2104 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2105 permutation when the load node is created.
2106 (vect_analyze_slp_instance): Re-use it here.
2108 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2110 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2112 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2114 PR rtl-optimization/93272
2115 * ira-lives.c (process_out_of_region_eh_regs): New function.
2116 (process_bb_node_lives): Call it.
2118 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2120 * coverage.c (read_counts_file): Make error message lowercase.
2122 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2124 * profile-count.c (profile_quality_display_names): Fix ordering.
2126 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2129 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2130 hash only when edge is first within the sequence.
2131 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2132 (symbol_table::create_edge): Do not set target_prob.
2133 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2135 (cgraph_edge::make_speculative): Drop target_prob parameter.
2136 (cgraph_edge::speculative_call_info): Remove.
2137 (cgraph_edge::first_speculative_call_target): New member function.
2138 (update_call_stmt_hash_for_removing_direct_edge): New function.
2139 (cgraph_edge::resolve_speculation): Rewrite to new API.
2140 (cgraph_edge::speculative_call_for_target): New member function.
2141 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2142 multiple speculation targets.
2143 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2145 (verify_speculative_call): Verify that targets form an interval.
2146 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2147 (cgraph_edge::first_speculative_call_target): New member function.
2148 (cgraph_edge::next_speculative_call_target): New member function.
2149 (cgraph_edge::speculative_call_target_ref): New member function.
2150 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2151 (cgraph_edge): Remove target_prob.
2152 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2153 Fix handling of speculative calls.
2154 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2155 * ipa-fnsummary.c (analyze_function_body): Likewise.
2156 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2157 * ipa-profile.c (dump_histogram): Fix formating.
2158 (ipa_profile_generate_summary): Watch for overflows.
2159 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2160 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2161 (update_indirect_edges_after_inlining): Update to new API.
2162 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2164 * profile-count.h: (profile_probability::adjusted): New.
2165 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2166 updating of profile.
2167 * value-prof.c (gimple_ic_transform): Rename to ...
2168 (dump_ic_profile): ... this one; update dumping.
2169 (stream_in_histogram_value): Fix formating.
2170 (gimple_value_profile_transformations): Update.
2172 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2175 * config/i386/i386.md (*movoi_internal_avx): Remove
2176 TARGET_SSE_TYPELESS_STORES check.
2177 (*movti_internal): Prefer TARGET_AVX over
2178 TARGET_SSE_TYPELESS_STORES.
2179 (*movtf_internal): Likewise.
2180 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2181 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2182 from TARGET_SSE_TYPELESS_STORES.
2184 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2186 * diagnostic-core.h (warning_at): Rename overload to...
2187 (warning_meta): ...this.
2188 (emit_diagnostic_valist): Delete decl of overload taking
2189 diagnostic_metadata.
2190 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2191 (warning_at): Rename overload taking diagnostic_metadata to...
2192 (warning_meta): ...this.
2194 2020-01-28 Richard Biener <rguenther@suse.de>
2196 PR tree-optimization/93439
2197 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2198 * tree-cfg.c (move_sese_region_to_fn): ... here.
2199 (verify_types_in_gimple_reference): Verify used cliques are
2202 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2205 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2206 argument of a pointer to struct gcc_options and pass it to
2207 parse_mtune_ctrl_str.
2208 (ix86_function_specific_restore): Pass opts to
2209 set_ix86_tune_features.
2210 (ix86_option_override_internal): Likewise.
2211 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2212 gcc_options and use it for x_ix86_tune_ctrl_string.
2214 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2216 PR rtl-optimization/87763
2217 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2218 simplification to handle subregs as well as bare regs.
2219 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2221 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2223 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2224 for reduction chains that (now) include a call.
2226 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2228 PR tree-optimization/92822
2229 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2230 out the don't-care elements of a vector whose significant elements
2231 are duplicates, make the don't-care elements duplicates too.
2233 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2235 PR tree-optimization/93434
2236 * tree-predcom.c (split_data_refs_to_components): Record which
2237 components have had aliasing loads removed. Prevent store-store
2238 commoning for all such components.
2240 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2243 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2244 -1 or is_vshift is true, use new_vector with number of elts npatterns
2245 rather than new_unary_operation.
2247 PR tree-optimization/93454
2248 * gimple-fold.c (fold_array_ctor_reference): Perform
2249 elt_size.to_uhwi () just once, instead of calling it in every
2250 iteration. Punt if that value is above size of the temporary
2251 buffer. Decrease third native_encode_expr argument when
2252 bufoff + elt_sz is above size of buf.
2254 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2256 * config/mips/mips.c (mips_declare_object_name)
2257 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2259 2020-01-27 Martin Liska <mliska@suse.cz>
2261 PR gcov-profile/93403
2262 * tree-profile.c (gimple_init_gcov_profiler): Generate
2263 both __gcov_indirect_call_profiler_v4 and
2264 __gcov_indirect_call_profiler_v4_atomic.
2266 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2269 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2271 (@aarch64_split_simd_mov<mode>): Use it.
2272 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2273 Leave the vec_extract patterns to handle 2-element vectors.
2274 (aarch64_simd_mov_from_<mode>high): Likewise.
2275 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2276 (vec_extractv2dfv1df): Likewise.
2278 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2280 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2281 jump conditions for *compare_condjump<GPI:mode>.
2283 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2286 * digraph.cc (test_edge::test_edge): Specify template for base
2289 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2291 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2293 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2295 * config/arc/arc-protos.h (gen_mlo): Remove.
2296 (gen_mhi): Likewise.
2297 * config/arc/arc.c (AUX_MULHI): Define.
2298 (arc_must_save_reister): Special handling for r58/59.
2299 (arc_compute_frame_size): Consider mlo/mhi registers.
2300 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2302 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2303 mlo/mhi name selection.
2304 (arc_restore_callee_saves): Don't early restore blink when ISR.
2305 (arc_expand_prologue): Add mlo/mhi saving.
2306 (arc_expand_epilogue): Add mlo/mhi restoring.
2309 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2310 numbering when MUL64 option is used.
2311 (DWARF2_FRAME_REG_OUT): Define.
2312 * config/arc/arc.md (arc600_stall): New pattern.
2313 (VUNSPEC_ARC_ARC600_STALL): Define.
2314 (mulsi64): Use correct mlo/mhi registers.
2315 (mulsi_600): Clean it up.
2316 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2318 (mhi_operand): Likewise.
2320 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2321 Petro Karashchenko <petro.karashchenko@ring.com>
2323 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2324 attributes if needed.
2325 (prepare_move_operands): Generate special unspec instruction for
2327 (arc_isuncached_mem_p): Propagate uncached attribute to each
2329 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2330 (VUNSPEC_ARC_STDI): Likewise.
2331 (ALLI): New mode iterator.
2332 (mALLI): New mode attribute.
2333 (lddi): New instruction pattern.
2335 (stdidi_split): Split instruction for architectures which are not
2336 supporting ll64 option.
2337 (lddidi_split): Likewise.
2339 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2341 PR rtl-optimization/92989
2342 * lra-lives.c (process_bb_lives): Update the live-in set before
2343 processing additional clobbers.
2345 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2347 PR rtl-optimization/93170
2348 * cselib.c (cselib_invalidate_regno_val): New function, split out
2350 (cselib_invalidate_regno): ...here.
2351 (cselib_invalidated_by_call_p): New function.
2352 (cselib_process_insn): Iterate over all the hard-register entries in
2353 REG_VALUES and invalidate any that cross call-clobbered registers.
2355 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2357 * dojump.c (split_comparison): Use HONOR_NANS rather than
2358 HONOR_SNANS when splitting LTGT.
2360 2020-01-27 Martin Liska <mliska@suse.cz>
2363 * opts.c (print_filtered_help): Exclude language-specific
2364 options from --help=common unless enabled in all FEs.
2366 2020-01-27 Martin Liska <mliska@suse.cz>
2368 * opts.c (print_help): Exclude params from
2369 all except --help=param.
2371 2020-01-27 Martin Liska <mliska@suse.cz>
2374 * config/i386/i386-features.c (make_resolver_func):
2375 Align the code with ppc64 target implementation.
2376 Do not generate a unique name for resolver function.
2378 2020-01-27 Richard Biener <rguenther@suse.de>
2380 PR tree-optimization/93397
2381 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2382 converted reduction chain SLP graph adjustment.
2384 2020-01-26 Marek Polacek <polacek@redhat.com>
2387 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2390 2020-01-26 Jason Merrill <jason@redhat.com>
2393 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2396 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2398 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2399 (rx_setmem): Likewise.
2401 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2404 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2405 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2406 drop <di> from constraint of last operand.
2409 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2410 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2411 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2413 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2416 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2419 2020-01-24 Jeff Law <law@redhat.com>
2421 PR tree-optimization/92788
2422 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2425 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2428 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2429 *avx_vperm_broadcast_<mode>,
2430 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2431 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2432 Move before avx2_perm<mode>/avx512f_perm<mode>.
2435 * simplify-rtx.c (simplify_const_unary_operation,
2436 simplify_const_binary_operation): Punt for mode precision above
2437 MAX_BITSIZE_MODE_ANY_INT.
2439 2020-01-24 Andrew Pinski <apinski@marvell.com>
2441 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2444 2020-01-24 Jeff Law <law@redhat.com>
2447 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2448 for REGs. Call output_operand_lossage to get more reasonable
2451 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2453 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2454 gcn_fp_compare_operator.
2455 (vec_cmpu<mode>di): Use gcn_compare_operator.
2456 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2457 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2458 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2459 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2460 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2461 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2462 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2463 gcn_fp_compare_operator.
2464 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2465 gcn_fp_compare_operator.
2466 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2467 gcn_fp_compare_operator.
2468 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2469 gcn_fp_compare_operator.
2471 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2473 * doc/install.texi (Cross-Compiler-Specific Options): Document
2474 `--with-toolexeclibdir' option.
2476 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2478 * target.def (flags_regnum): Also mention effect on delay slot filling.
2479 * doc/tm.texi: Regenerate.
2481 2020-01-23 Jeff Law <law@redhat.com>
2483 PR translation/90162
2484 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2486 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2489 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2492 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2494 PR rtl-optimization/93402
2495 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2498 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2500 * config.in: Regenerated.
2501 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2502 for TARGET_LIBC_GNUSTACK.
2503 * configure: Regenerated.
2504 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2505 found to be 2.31 or greater.
2507 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2509 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2511 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2512 (mips_asm_file_end): New function. Delegate to
2513 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2514 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2516 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2519 * config/i386/i386-modes.def (POImode): New mode.
2520 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2521 * config/i386/i386.md (DPWI): New mode attribute.
2522 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2524 (QPWI): ... this. Use POI instead of OI for TImode.
2525 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2526 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2529 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2532 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2534 (speculation_tracker_rev): New pattern.
2535 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2536 Use speculation_tracker_rev to track the inverse condition.
2538 2020-01-23 Richard Biener <rguenther@suse.de>
2540 PR tree-optimization/93381
2541 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2542 alias-set of the def as argument and record the first one.
2543 (vn_walk_cb_data::first_set): New member.
2544 (vn_reference_lookup_3): Pass the alias-set of the current def
2545 to push_partial_def. Fix alias-set used in the aggregate copy
2547 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2548 * real.c (clear_significand_below): Fix out-of-bound access.
2550 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2553 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2554 New define_insn patterns.
2556 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2558 * doc/sourcebuild.texi (check-function-bodies): Add an
2559 optional target/xfail selector.
2561 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2563 PR rtl-optimization/93124
2564 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2565 bare USE and CLOBBER insns.
2567 2020-01-22 Andrew Pinski <apinski@marvell.com>
2569 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2571 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2574 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2575 diagnostic_manager into "ana" namespace.
2576 * selftest-run-tests.c (selftest::run_tests): Update for move of
2577 selftest::run_analyzer_selftests to
2578 ana::selftest::run_analyzer_selftests.
2580 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2582 * cfgexpand.c (union_stack_vars): Update the size.
2584 2020-01-22 Richard Biener <rguenther@suse.de>
2586 PR tree-optimization/93381
2587 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2588 throughout, handle all conversions the same.
2590 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2593 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2594 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2595 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2596 Call force_reg on high_in2 unconditionally.
2598 2020-01-22 Martin Liska <mliska@suse.cz>
2600 PR tree-optimization/92924
2601 * profile.c (compute_value_histograms): Divide
2604 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2607 * output.h (assemble_name_resolve): Declare.
2608 * varasm.c (assemble_name_resolve): New function.
2609 (assemble_name): Use it.
2610 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2612 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2614 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2615 update_web_docs_git instead of update_web_docs_svn.
2617 2020-01-21 Andrew Pinski <apinski@marvell.com>
2620 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2621 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2622 (*tlsgd_small_<mode>): Likewise.
2623 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2624 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2625 register. Convert that register back to dest using convert_mode.
2627 2020-01-21 Jim Wilson <jimw@sifive.com>
2629 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2632 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2633 Uros Bizjak <ubizjak@gmail.com>
2636 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2638 (legitimize_tls_address): Do GNU2 TLS address computation in
2639 ptr_mode and zero-extend result to Pmode.
2640 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2641 :P with :PTR and Pmode with ptr_mode.
2642 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2643 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2644 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2646 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2649 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2650 the last two operands are CONST_INT_P before using them as such.
2652 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2654 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2655 to get the integer element types.
2657 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2659 * config/aarch64/aarch64-sve-builtins.h
2660 (function_expander::convert_to_pmode): Declare.
2661 * config/aarch64/aarch64-sve-builtins.cc
2662 (function_expander::convert_to_pmode): New function.
2663 (function_expander::get_contiguous_base): Use it.
2664 (function_expander::prepare_gather_address_operands): Likewise.
2665 * config/aarch64/aarch64-sve-builtins-sve2.cc
2666 (svwhilerw_svwhilewr_impl::expand): Likewise.
2668 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2671 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2672 cfun->machine->label_is_assembled.
2673 (aarch64_print_patchable_function_entry): New.
2674 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2675 * config/aarch64/aarch64.h (struct machine_function): New field,
2678 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2681 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2684 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2687 * cgraph.c (cgraph_edge::resolve_speculation,
2688 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2689 call_stmt_site_hash.
2691 2020-01-21 Martin Liska <mliska@suse.cz>
2693 * config/rs6000/rs6000.c (common_mode_defined): Remove
2696 2020-01-21 Richard Biener <rguenther@suse.de>
2698 PR tree-optimization/92328
2699 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2700 type when value-numbering same-sized store by inserting a
2702 (eliminate_dom_walker::eliminate_stmt): When eliminating
2703 a redundant store handle bit-reinterpretation of the same value.
2705 2020-01-21 Andrew Pinski <apinski@marvel.com>
2708 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2710 (prepare_block_for_update): This. Use a worklist instead of
2713 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2715 * gcc/config/arm/arm.c (clear_operation_p):
2716 Initialise last_regno, skip first iteration
2717 based on the first_set value and use ints instead
2718 of the unnecessary HOST_WIDE_INTs.
2720 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2723 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2724 compare_mode other than SFmode or DFmode.
2726 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2729 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2730 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2731 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2733 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2735 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2737 2020-01-20 Andrew Pinski <apinski@marvell.com>
2740 * targhooks.c (default_print_patchable_function_entry): Use
2741 output_asm_insn to emit the nop instruction.
2743 2020-01-20 Fangrui Song <maskray@google.com>
2746 * targhooks.c (default_print_patchable_function_entry): Align to
2749 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2752 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2753 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2754 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2755 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2756 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2757 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2758 Remove the {q} suffix from lea.
2759 (*tls_dynamic_gnu2_call_64): Renamed to ...
2760 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2761 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2762 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2763 Pass Pmode to gen_tls_dynamic_gnu2_64.
2765 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2767 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2769 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2771 * config/aarch64/aarch64-sve-builtins-base.cc
2772 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2774 2020-01-20 Richard Biener <rguenther@suse.de>
2777 * dwarf2out.c (prune_unused_types): Unconditionally mark
2778 called function DIEs.
2780 2020-01-20 Martin Liska <mliska@suse.cz>
2782 PR tree-optimization/93199
2783 * tree-eh.c (struct leh_state): Add
2784 new field outer_non_cleanup.
2785 (cleanup_is_dead_in): Pass leh_state instead
2786 of eh_region. Add a checking that state->outer_non_cleanup
2787 points to outer non-clean up region.
2788 (lower_try_finally): Record outer_non_cleanup
2790 (lower_catch): Likewise.
2791 (lower_eh_filter): Likewise.
2792 (lower_eh_must_not_throw): Likewise.
2793 (lower_cleanup): Likewise.
2795 2020-01-20 Richard Biener <rguenther@suse.de>
2797 PR tree-optimization/93094
2798 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2799 (vect_transform_loop): Likewise.
2800 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2801 loop_vectorized_call to vect_transform_loop.
2802 * tree-vect-loop.c (vect_transform_loop): Pass down
2803 loop_vectorized_call to vect_loop_versioning.
2804 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2805 the earlier discovered loop_vectorized_call.
2807 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2809 * doc/contribute.texi: Update for SVN -> Git transition.
2810 * doc/install.texi: Likewise.
2812 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2815 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2816 speculative targets.
2817 (verify_speculative_call): New function
2818 (cgraph_node::verify_node): Use it.
2819 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2822 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2825 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2826 (cgraph_edge::make_direct): Remove all indirect targets.
2827 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2828 (cgraph_node::verify_node): Verify that only one call_stmt or
2829 lto_stmt_uid is set.
2830 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2832 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2833 (lto_output_ref): Simplify streaming of stmt.
2834 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2836 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2838 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2839 Mark parameter unused.
2841 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2843 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2845 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2847 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2849 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2851 * Makefile.in: Add coroutine-passes.o.
2852 * builtin-types.def (BT_CONST_SIZE): New.
2853 (BT_FN_BOOL_PTR): New.
2854 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2855 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2856 * coroutine-builtins.def: New file.
2857 * coroutine-passes.cc: New file.
2858 * function.h (struct GTY function): Add a bit to indicate that the
2859 function is a coroutine component.
2860 * internal-fn.c (expand_CO_FRAME): New.
2861 (expand_CO_YIELD): New.
2862 (expand_CO_SUSPN): New.
2863 (expand_CO_ACTOR): New.
2864 * internal-fn.def (CO_ACTOR): New.
2868 * passes.def: Add pass_coroutine_lower_builtins,
2869 pass_coroutine_early_expand_ifns.
2870 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2871 (make_pass_coroutine_early_expand_ifns): New.
2872 * doc/invoke.texi: Document the fcoroutines command line
2875 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2877 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2880 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2881 after checking the argument is a REG. Don't use REGNO (reg)
2882 again to set last_regno, reuse regno variable instead.
2884 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2886 * doc/analyzer.texi (Limitations): Add note about NaN.
2888 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2889 Sudakshina Das <sudi.das@arm.com>
2891 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2892 and valid immediate.
2893 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2894 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2895 * config/arm/constraints.md (Pg): New.
2896 * config/arm/predicates.md (long_shift_imm): New.
2897 (arm_reg_or_long_shift_imm): Likewise.
2898 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2899 (thumb2_lsll): Likewise.
2902 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2903 Sudakshina Das <sudi.das@arm.com>
2905 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2906 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2907 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2908 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2909 * config/arm/thumb2.md (thumb2_asrl): New.
2910 (thumb2_lsll): Likewise.
2912 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2914 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2917 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2919 * gdbinit.in (help-gcc-hooks): New command.
2920 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2921 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2924 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2926 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2927 correct target macro.
2929 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2931 * config/aarch64/aarch64-protos.h
2932 (aarch64_sve_ld1ro_operand_p): New.
2933 * config/aarch64/aarch64-sve-builtins-base.cc
2934 (class load_replicate): New.
2935 (class svld1ro_impl): New.
2936 (class svld1rq_impl): Change to inherit from load_replicate.
2937 (svld1ro): New sve intrinsic function base.
2938 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2939 New DEF_SVE_FUNCTION.
2940 * config/aarch64/aarch64-sve-builtins-base.h
2941 (svld1ro): New decl.
2942 * config/aarch64/aarch64-sve-builtins.cc
2943 (function_expander::add_mem_operand): Modify assert to allow
2945 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2947 * config/aarch64/aarch64.c
2948 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2949 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2950 (aarch64_sve_ld1ro_operand_p): New.
2951 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2952 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2953 * config/aarch64/predicates.md
2954 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2956 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2958 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2959 Introduce this ACLE specified predefined macro.
2960 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2961 (fp): Disabling this disables f64mm.
2962 (simd): Disabling this disables f64mm.
2963 (fp16): Disabling this disables f64mm.
2964 (sve): Disabling this disables f64mm.
2965 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2966 (AARCH64_ISA_F64MM): New.
2967 (TARGET_F64MM): New.
2968 * doc/invoke.texi (f64mm): Document new option.
2970 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2972 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2973 (neoversen1_tunings): Likewise.
2975 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2978 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2979 Add assert to ensure prolog has been emitted.
2980 (aarch64_split_atomic_op): Likewise.
2981 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2982 Use epilogue_completed rather than reload_completed.
2983 (aarch64_atomic_exchange<mode>): Likewise.
2984 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2985 (atomic_nand<mode>): Likewise.
2986 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2987 (atomic_fetch_nand<mode>): Likewise.
2988 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2989 (atomic_nand_fetch<mode>): Likewise.
2991 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2994 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2996 (REVERSE_CONDITION): Delete.
2997 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2998 (CCFP_CCFPE): Likewise.
2999 (e): New mode attribute.
3000 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
3001 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
3002 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
3003 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
3004 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
3005 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
3006 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
3007 name of generator from gen_ccmpdi to gen_ccmpccdi.
3008 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
3009 the previous comparison but aren't able to, use the new ccmp_rev
3012 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
3014 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
3015 than testing directly for INTEGER_CST.
3016 (gimplify_target_expr, gimplify_omp_depend): Likewise.
3018 2020-01-17 Jakub Jelinek <jakub@redhat.com>
3020 PR tree-optimization/93292
3021 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
3022 get_vectype_for_scalar_type returns NULL.
3024 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3026 * params.opt (-param=max-predicted-iterations): Increase range from 0.
3027 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
3029 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
3031 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
3033 * params.opt: (max-predicted-iterations): Set bounds.
3034 * predict.c (real_almost_one, real_br_prob_base,
3035 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
3036 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3037 probabilities; do not truncate to reg_br_prob_bases.
3038 (estimate_loops_at_level): Pass max_cyclic_prob.
3039 (estimate_loops): Compute max_cyclic_prob.
3040 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3042 * profile-count.c (profile_probability::to_sreal): New.
3043 * profile-count.h (class sreal): Move up in file.
3044 (profile_probability::to_sreal): Declare.
3046 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3049 (arm_invalid_conversion): New function for target hook.
3050 (arm_invalid_unary_op): New function for target hook.
3051 (arm_invalid_binary_op): New function for target hook.
3053 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3055 * config.gcc: Add arm_bf16.h.
3056 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3057 (arm_simd_builtin_std_type): Add BFmode.
3058 (arm_init_simd_builtin_types): Define element types for vector types.
3059 (arm_init_bf16_types): New function.
3060 (arm_init_builtins): Add arm_init_bf16_types function call.
3061 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3062 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3063 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3064 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3065 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3066 (arm_mangle_type): Add __bf16.
3067 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3068 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3069 arm_bf16_ptr_type_node.
3070 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3071 define_split between ARM registers.
3072 * config/arm/arm_bf16.h: New file.
3073 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3074 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3076 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3077 * config/arm/vfp.md: Add BFmode to movhf patterns.
3079 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3080 Andre Vieira <andre.simoesdiasvieira@arm.com>
3082 * config/arm/arm-cpus.in (mve, mve_float): New features.
3083 (dsp, mve, mve.fp): New options.
3084 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3085 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3086 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3088 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3089 Thomas Preud'homme <thomas.preudhomme@arm.com>
3091 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3093 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3094 error for using -mcmse when targeting Armv8.1-M Mainline.
3096 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3097 Thomas Preud'homme <thomas.preudhomme@arm.com>
3099 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3100 address in r4 when targeting Armv8.1-M Mainline.
3101 (nonsecure_call_value_internal): Likewise.
3102 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3103 a register match_operand again. Emit BLXNS when targeting
3105 (nonsecure_call_value_reg_thumb2): Likewise.
3107 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3108 Thomas Preud'homme <thomas.preudhomme@arm.com>
3110 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3111 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3112 variable as true when floating-point ABI is not hard. Replace
3113 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3114 Generate VLSTM and VLLDM instruction respectively before and
3115 after a function call to cmse_nonsecure_call function.
3116 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3117 (VUNSPEC_VLLDM): Likewise.
3118 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3119 (lazy_load_multiple_insn): Likewise.
3121 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3122 Thomas Preud'homme <thomas.preudhomme@arm.com>
3124 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3125 (arm_emit_vfp_multi_reg_pop): Likewise.
3126 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3127 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3128 restore callee-saved VFP registers.
3130 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3131 Thomas Preud'homme <thomas.preudhomme@arm.com>
3133 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3134 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3135 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3136 callee-saved GPRs as well as clear ip register before doing a nonsecure
3137 call then restore callee-saved GPRs after it when targeting
3139 (arm_reorg): Adapt to function rename.
3141 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3142 Thomas Preud'homme <thomas.preudhomme@arm.com>
3144 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3145 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3146 clear_vfp_multiple pattern based on a new vfp parameter.
3147 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3148 targeting Armv8.1-M Mainline.
3149 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3150 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3151 whether VFP registers are available before looking call_used_regs for a
3153 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3154 of prototype of clear_operation_p.
3155 (clear_vfp_multiple_operation): New predicate.
3156 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3157 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3159 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3160 Thomas Preud'homme <thomas.preudhomme@arm.com>
3162 * config/arm/arm-protos.h (clear_operation_p): Declare.
3163 * config/arm/arm.c (clear_operation_p): New function.
3164 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3165 targeting Armv8.1-M Mainline or successor.
3166 (output_return_instruction): Only output APSR register clearing if
3167 Armv8.1-M Mainline instructions not available.
3168 (thumb_exit): Likewise.
3169 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3170 * config/arm/thumb2.md (clear_apsr): New define_insn.
3171 (clear_multiple): Likewise.
3172 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3174 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3175 Thomas Preud'homme <thomas.preudhomme@arm.com>
3177 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3178 (use_return_insn): Also return false for Armv8.1-M Mainline.
3179 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3180 Mainline instructions are available.
3181 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3182 when targeting Armv8.1-M Mainline Security Extensions.
3183 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3184 Mainline entry function.
3185 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3186 targeting Armv8.1-M Mainline or successor.
3187 (arm_expand_epilogue): Fix indentation of caller-saved register
3188 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3190 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3191 (FP_SYSREGS): Likewise.
3192 (enum vfp_sysregs_encoding): Define enum.
3193 (fp_sysreg_names): Declare.
3194 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3195 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3196 (pop_fpsysreg_insn): Likewise.
3198 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3199 Thomas Preud'homme <thomas.preudhomme@arm.com>
3201 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3202 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3203 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3204 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3205 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3206 (ARMv8_1m_main): New feature group.
3207 (armv8.1-m.main): New architecture.
3208 * config/arm/arm-tables.opt: Regenerate.
3209 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3210 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3211 (arm_options_perform_arch_sanity_checks): Error out when targeting
3212 Armv8.1-M Mainline Security Extensions.
3213 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3215 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3217 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3218 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3219 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3220 aarch64_bfdot_laneq): New.
3221 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3222 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3223 vbfdotq_laneq_f32): New.
3224 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3225 VBFMLA_W, VBF): New.
3226 (isquadop): Add V4BF, V8BF.
3228 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3230 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3231 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3232 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3233 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3234 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3235 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3236 usdot_laneq, sudot_lane,sudot_laneq): New.
3237 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3238 (aarch64_<sur>dot_lane): New.
3239 * config/aarch64/arm_neon.h (vusdot_s32): New.
3241 (vusdot_lane_s32): New.
3242 (vsudot_lane_s32): New.
3243 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3244 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3246 2020-01-16 Martin Liska <mliska@suse.cz>
3248 * value-prof.c (dump_histogram_value): Fix
3249 obvious spacing issue.
3251 2020-01-16 Andrew Pinski <apinski@marvell.com>
3253 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3254 !storage_order_barrier_p.
3256 2020-01-16 Andrew Pinski <apinski@marvell.com>
3258 * sched-int.h (_dep): Add unused bit-field field for the padding.
3259 * sched-deps.c (init_dep_1): Init unused field.
3261 2020-01-16 Andrew Pinski <apinski@marvell.com>
3263 * optabs.h (create_expand_operand): Initialize target field also.
3265 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3267 PR tree-optimization/92429
3268 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3269 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3271 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3274 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3276 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3277 aarch64_sve_int_mode to each mode.
3279 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3281 * doc/analyzer.texi (Overview): Add note about
3282 -fdump-ipa-analyzer.
3284 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3286 PR tree-optimization/93231
3287 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3288 input_type is unsigned. Use tree_to_shwi for shift constant.
3289 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3290 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3291 inputs more efficiently.
3293 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3295 * config/i386/i386.md (*movsf_internal): Do not require
3296 SSE2 ISA for alternatives 14 and 15.
3298 2020-01-15 Richard Biener <rguenther@suse.de>
3301 * tree-eh.c (sink_clobbers): If we already visited the destination
3302 block do not defer insertion.
3303 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3304 the purpose of defered insertion.
3306 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3308 * BASE-VER: Bump to 10.0.1.
3310 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3312 PR tree-optimization/93247
3313 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3314 type of the stmt that we're going to vectorize.
3316 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3318 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3319 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3322 2020-01-15 Martin Liska <mliska@suse.cz>
3324 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3325 2 calls of streamer_read_hwi in a function call.
3327 2020-01-15 Richard Biener <rguenther@suse.de>
3329 * alias.c (record_alias_subset): Avoid redundant work when
3330 subset is already recorded.
3332 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3334 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3335 the analyzer options provide CWE identifiers.
3337 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3339 * tree-diagnostic-path.cc (path_summary::event_range::print):
3340 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3341 using get_pure_location.
3343 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3345 PR tree-optimization/93262
3346 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3347 perform head trimming only if the last argument is constant,
3348 either all ones, or larger or equal to head trim, in the latter
3349 case decrease the last argument by head_trim.
3351 PR tree-optimization/93249
3352 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3353 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3354 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3355 perform head trim unless we can prove there are no '\0' chars
3356 from the source among the first head_trim chars.
3358 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3360 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3362 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3365 * config/i386/sse.md
3366 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3367 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3368 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3369 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3370 just a single alternative instead of two, make operands 1 and 2
3373 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3376 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3379 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3381 * Makefile.in (lang_opt_files): Add analyzer.opt.
3382 (ANALYZER_OBJS): New.
3383 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3384 tristate.o and ANALYZER_OBJS.
3385 (TEXI_GCCINT_FILES): Add analyzer.texi.
3386 * common.opt (-fanalyzer): New driver option.
3387 * config.in: Regenerate.
3388 * configure: Regenerate.
3389 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3390 (gccdepdir): Also create depdir for "analyzer" subdir.
3391 * digraph.cc: New file.
3392 * digraph.h: New file.
3393 * doc/analyzer.texi: New file.
3394 * doc/gccint.texi ("Static Analyzer") New menu item.
3395 (analyzer.texi): Include it.
3396 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3397 ("Warning Options"): Add static analysis warnings to the list.
3398 (-Wno-analyzer-double-fclose): New option.
3399 (-Wno-analyzer-double-free): New option.
3400 (-Wno-analyzer-exposure-through-output-file): New option.
3401 (-Wno-analyzer-file-leak): New option.
3402 (-Wno-analyzer-free-of-non-heap): New option.
3403 (-Wno-analyzer-malloc-leak): New option.
3404 (-Wno-analyzer-possible-null-argument): New option.
3405 (-Wno-analyzer-possible-null-dereference): New option.
3406 (-Wno-analyzer-null-argument): New option.
3407 (-Wno-analyzer-null-dereference): New option.
3408 (-Wno-analyzer-stale-setjmp-buffer): New option.
3409 (-Wno-analyzer-tainted-array-index): New option.
3410 (-Wno-analyzer-use-after-free): New option.
3411 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3412 (-Wno-analyzer-use-of-uninitialized-value): New option.
3413 (-Wanalyzer-too-complex): New option.
3414 (-fanalyzer-call-summaries): New warning.
3415 (-fanalyzer-checker=): New warning.
3416 (-fanalyzer-fine-grained): New warning.
3417 (-fno-analyzer-state-merge): New warning.
3418 (-fno-analyzer-state-purge): New warning.
3419 (-fanalyzer-transitivity): New warning.
3420 (-fanalyzer-verbose-edges): New warning.
3421 (-fanalyzer-verbose-state-changes): New warning.
3422 (-fanalyzer-verbosity=): New warning.
3423 (-fdump-analyzer): New warning.
3424 (-fdump-analyzer-callgraph): New warning.
3425 (-fdump-analyzer-exploded-graph): New warning.
3426 (-fdump-analyzer-exploded-nodes): New warning.
3427 (-fdump-analyzer-exploded-nodes-2): New warning.
3428 (-fdump-analyzer-exploded-nodes-3): New warning.
3429 (-fdump-analyzer-supergraph): New warning.
3430 * doc/sourcebuild.texi (dg-require-dot): New.
3431 (dg-check-dot): New.
3432 * gdbinit.in (break-on-saved-diagnostic): New command.
3433 * graphviz.cc: New file.
3434 * graphviz.h: New file.
3435 * ordered-hash-map-tests.cc: New file.
3436 * ordered-hash-map.h: New file.
3437 * passes.def (pass_analyzer): Add before
3438 pass_ipa_whole_program_visibility.
3439 * selftest-run-tests.c (selftest::run_tests): Call
3440 selftest::ordered_hash_map_tests_cc_tests.
3441 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3443 * shortest-paths.h: New file.
3444 * timevar.def (TV_ANALYZER): New timevar.
3445 (TV_ANALYZER_SUPERGRAPH): Likewise.
3446 (TV_ANALYZER_STATE_PURGE): Likewise.
3447 (TV_ANALYZER_PLAN): Likewise.
3448 (TV_ANALYZER_SCC): Likewise.
3449 (TV_ANALYZER_WORKLIST): Likewise.
3450 (TV_ANALYZER_DUMP): Likewise.
3451 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3452 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3453 * tree-pass.h (make_pass_analyzer): New decl.
3454 * tristate.cc: New file.
3455 * tristate.h: New file.
3457 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3460 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3461 alternatives 9 and 10.
3463 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3465 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3466 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3467 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3468 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3469 (selftest::hash_map_tests_c_tests): Call it.
3470 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3471 New static constant, using the value of = H::empty_zero_p.
3472 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3473 from default_hash_traits <Value>.
3474 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3476 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3477 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3478 calls to mark_empty with !Descriptor::empty_zero_p.
3479 (hash_table::empty_slow): Conditionalize the memset call with a
3480 check that Descriptor::empty_zero_p; otherwise, loop through the
3481 entries calling mark_empty on them.
3482 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3483 (pointer_hash::empty_zero_p): Likewise.
3484 (pair_hash::empty_zero_p): Likewise.
3485 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3487 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3488 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3489 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3490 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3491 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3492 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3493 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3494 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3496 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3499 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3501 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3502 fix typo on return value.
3504 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3507 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3509 (cgraph_edge::make_speculative): Add param for setting speculative_id
3511 (cgraph_edge::speculative_call_info): Update comments and find reference
3512 by speculative_id for multiple indirect targets.
3513 (cgraph_edge::resolve_speculation): Decrease the speculations
3514 for indirect edge, drop it's speculative if not direct target
3515 left. Update comments.
3516 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3517 (cgraph_node::dump): Print num_speculative_call_targets.
3518 (cgraph_node::verify_node): Don't report error if speculative
3519 edge not include statement.
3520 (cgraph_edge::num_speculative_call_targets_p): New function.
3521 * cgraph.h (int common_target_id): Remove.
3522 (int common_target_probability): Remove.
3523 (num_speculative_call_targets): New variable.
3524 (make_speculative): Add param for setting speculative_id.
3525 (cgraph_edge::num_speculative_call_targets_p): New declare.
3526 (target_prob): New variable.
3527 (speculative_id): New variable.
3528 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3529 call summaries for multiple speculative call targets.
3530 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3531 * ipa-profile.c (struct speculative_call_target): New struct.
3532 (class speculative_call_summary): New class.
3533 (class speculative_call_summaries): New class.
3534 (call_sums): New variable.
3535 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3536 (ipa_profile_write_edge_summary): New function.
3537 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3538 (ipa_profile_dump_all_summaries): New function.
3539 (ipa_profile_read_edge_summary): New function.
3540 (ipa_profile_read_summary_section): New function.
3541 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3542 (ipa_profile): Generate num_speculative_call_targets from
3544 * ipa-ref.h (speculative_id): New variable.
3545 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3546 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3547 common_target_probability. Stream out speculative_id and
3548 num_speculative_call_targets.
3549 (input_edge): Likewise.
3550 * predict.c (dump_prediction): Remove edges count assert to be
3552 * symtab.c (symtab_node::create_reference): Init speculative_id.
3553 (symtab_node::clone_references): Clone speculative_id.
3554 (symtab_node::clone_referring): Clone speculative_id.
3555 (symtab_node::clone_reference): Clone speculative_id.
3556 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3557 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3558 if indirect call contains multiple speculative targets.
3559 * value-prof.h (check_ic_target): Remove.
3560 * value-prof.c (gimple_value_profile_transformations):
3561 Use void function gimple_ic_transform.
3562 * value-prof.c (gimple_ic_transform): Handle topn case.
3563 Fix comment typos. Change it to a void function.
3565 2020-01-13 Andrew Pinski <apinski@marvell.com>
3567 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3568 (octeontx2t98): New define.
3569 (octeontx2t96): New define.
3570 (octeontx2t93): New define.
3571 (octeontx2f95): New define.
3572 (octeontx2f95n): New define.
3573 (octeontx2f95mm): New define.
3574 * config/aarch64/aarch64-tune.md: Regenerate.
3575 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3577 2020-01-13 Jason Merrill <jason@redhat.com>
3579 PR c++/33799 - destroy return value if local cleanup throws.
3580 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3582 2020-01-13 Martin Liska <mliska@suse.cz>
3584 * ipa-cp.c (get_max_overall_size): Use newly
3585 renamed param param_ipa_cp_unit_growth.
3586 * params.opt: Remove legacy param name.
3588 2020-01-13 Martin Sebor <msebor@redhat.com>
3590 PR tree-optimization/93213
3591 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3592 stores to be eliminated.
3594 2020-01-13 Martin Liska <mliska@suse.cz>
3596 * opts.c (print_help): Do not print CL_PARAM
3597 and CL_WARNING for CL_OPTIMIZATION.
3599 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3602 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3603 depending on optimization settings.
3605 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3607 PR tree-optimization/90838
3608 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3609 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3610 argument rather than to initialize temporary for targets that
3611 don't use the mode argument at all. Initialize ctzval to avoid
3614 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3616 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3617 * tree-core.h: Document it.
3618 * gimplify.c (gimplify_omp_workshare): Set it.
3619 * omp-low.c (lower_omp_target): Use it.
3620 * tree-pretty-print.c (dump_omp_clause): Print it.
3622 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3623 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3625 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3627 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3628 * common.opt (fdiagnostics-path-format=): New option.
3629 (diagnostic_path_format): New enum.
3630 (fdiagnostics-show-path-depths): New option.
3631 * coretypes.h (diagnostic_event_id_t): New forward decl.
3632 * diagnostic-color.c (color_dict): Add "path".
3633 * diagnostic-event-id.h: New file.
3634 * diagnostic-format-json.cc (json_from_expanded_location): Make
3636 (json_end_diagnostic): Call context->make_json_for_path if it
3637 exists and the diagnostic has a path.
3638 (diagnostic_output_format_init): Clear context->print_path.
3639 * diagnostic-path.h: New file.
3640 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3641 when printing a run of events in a diagnostic_path so that they
3642 all get the same color.
3643 (layout::m_diagnostic_path_p): New field.
3644 (layout::layout): Initialize it.
3645 (layout::print_any_labels): Don't colorize the label text for an
3646 event in a diagnostic_path.
3647 (gcc_rich_location::add_location_if_nearby): Add
3648 "restrict_to_current_line_spans" and "label" params. Pass the
3649 former to layout.maybe_add_location_range; pass the latter
3650 when calling add_range.
3651 * diagnostic.c: Include "diagnostic-path.h".
3652 (diagnostic_initialize): Initialize context->path_format and
3653 context->show_path_depths.
3654 (diagnostic_show_any_path): New function.
3655 (diagnostic_path::interprocedural_p): New function.
3656 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3657 (simple_diagnostic_path::num_events): New function.
3658 (simple_diagnostic_path::get_event): New function.
3659 (simple_diagnostic_path::add_event): New function.
3660 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3661 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3662 (debug): New overload taking a diagnostic_path *.
3663 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3664 * diagnostic.h (enum diagnostic_path_format): New enum.
3665 (json::value): New forward decl.
3666 (diagnostic_context::path_format): New field.
3667 (diagnostic_context::show_path_depths): New field.
3668 (diagnostic_context::print_path): New callback field.
3669 (diagnostic_context::make_json_for_path): New callback field.
3670 (diagnostic_show_any_path): New decl.
3671 (json_from_expanded_location): New decl.
3672 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3673 (-fdiagnostics-show-path-depths): New option.
3674 (-fdiagnostics-color): Add "path" to description of default
3675 GCC_COLORS; describe it.
3676 (-fdiagnostics-format=json): Document how diagnostic paths are
3677 represented in the JSON output format.
3678 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3679 Add optional params "restrict_to_current_line_spans" and "label".
3680 * opts.c (common_handle_option): Handle
3681 OPT_fdiagnostics_path_format_ and
3682 OPT_fdiagnostics_show_path_depths.
3683 * pretty-print.c: Include "diagnostic-event-id.h".
3684 (pp_format): Implement "%@" format code for printing
3685 diagnostic_event_id_t *.
3686 (selftest::test_pp_format): Add tests for "%@".
3687 * selftest-run-tests.c (selftest::run_tests): Call
3688 selftest::tree_diagnostic_path_cc_tests.
3689 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3690 * toplev.c (general_init): Initialize global_dc->path_format and
3691 global_dc->show_path_depths.
3692 * tree-diagnostic-path.cc: New file.
3693 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3694 non-static. Drop "diagnostic" param in favor of storing the
3695 original value of "where" and re-using it.
3696 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3697 maybe_unwind_expanded_macro_loc.
3698 (tree_diagnostics_defaults): Initialize context->print_path and
3699 context->make_json_for_path.
3700 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3702 (default_tree_make_json_for_path): New decl.
3703 (maybe_unwind_expanded_macro_loc): New decl.
3705 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3707 PR tree-optimization/93210
3708 * fold-const.h (native_encode_initializer,
3709 can_native_interpret_type_p): Declare.
3710 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3712 (native_encode_initializer): New function, moved from dwarf2out.c.
3713 Adjust to native_encode_expr compatible arguments, including dry-run
3714 and partial extraction modes. Don't handle STRING_CST.
3715 (can_native_interpret_type_p): No longer static.
3716 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3717 offset / BITS_PER_UNIT fits into int and don't call it if
3718 can_native_interpret_type_p fails. If suboff is NULL and for
3719 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3720 native_encode_initializer.
3721 (fold_const_aggregate_ref_1): Formatting fix.
3722 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3723 (tree_add_const_value_attribute): Adjust caller.
3725 PR tree-optimization/90838
3726 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3727 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3728 CTZ_DEFINED_VALUE_AT_ZERO.
3730 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3733 * lra-constraints.c (match_reload): Permit input operands have the
3734 same mode as output while other input operands have a different
3737 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3739 PR tree-optimization/90838
3740 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3741 (check_ctz_string): Likewise.
3742 (optimize_count_trailing_zeroes): Likewise.
3743 (simplify_count_trailing_zeroes): Likewise.
3744 (pass_forwprop::execute): Try ctz simplification.
3745 * match.pd: Add matching for ctz idioms.
3747 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3749 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3751 (aarch64_invalid_unary_op): New function for target hook.
3752 (aarch64_invalid_binary_op): New function for target hook.
3754 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3756 * config.gcc: Add arm_bf16.h.
3757 * config/aarch64/aarch64-builtins.c
3758 (aarch64_simd_builtin_std_type): Add BFmode.
3759 (aarch64_init_simd_builtin_types): Define element types for vector
3761 (aarch64_init_bf16_types): New function.
3762 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3763 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3765 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3766 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3768 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3769 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3770 * config/aarch64/aarch64.c
3771 (aarch64_classify_vector_mode): Add support for BF types.
3772 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3773 (aarch64_vq_mode): Add support for BF types.
3774 (aarch64_simd_container_mode): Add support for BF types.
3775 (aarch64_mangle_type): Add support for BF scalar type.
3776 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3777 * config/aarch64/arm_bf16.h: New file.
3778 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3779 * config/aarch64/iterators.md: Add BF types to mode attributes.
3780 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3782 2020-01-10 Jason Merrill <jason@redhat.com>
3784 PR c++/93173 - incorrect tree sharing.
3785 * gimplify.c (copy_if_shared): No longer static.
3786 * gimplify.h: Declare it.
3788 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3790 * doc/invoke.texi (-msve-vector-bits=): Document that
3791 -msve-vector-bits=128 now generates VL-specific code for
3792 little-endian targets.
3793 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3794 build_vector_type_for_mode to construct the data vector types.
3795 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3796 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3797 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3798 for 128-bit vectors.
3800 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3802 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3805 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3807 * config/aarch64/aarch64-builtins.c
3808 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3809 rather than checking the number of elements and the element mode.
3811 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3813 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3814 get_related_vectype_for_scalar_type rather than build_vector_type
3815 to create the index type for a conditional reduction.
3817 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3819 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3820 for any type of gather or scatter, including strided accesses.
3822 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3824 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3827 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3829 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3831 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3832 parameter and its use to reset DR_OFFSET's.
3833 (vect_transform_loop): Remove orig_drs_init argument.
3834 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3835 member of dr_vec_info rather than the offset of the associated
3836 data_reference's innermost_loop_behavior.
3837 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3838 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3839 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3840 get_dr_vinfo_offset.
3841 (vectorizable_store): Likewise.
3842 (vectorizable_load): Likewise.
3844 2020-01-10 Richard Biener <rguenther@suse.de>
3846 * gimple-ssa-store-merging
3847 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3849 2020-01-10 Martin Liska <mliska@suse.cz>
3852 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3853 encapsulation that was there before r280040.
3855 2020-01-10 Richard Biener <rguenther@suse.de>
3858 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3859 sequences to avoid walking them again for secondary opportunities.
3860 (pass_lower_eh_dispatch::execute): Instead actually insert
3863 2020-01-10 Richard Biener <rguenther@suse.de>
3866 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3867 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3868 avoid quadraticness.
3870 2020-01-10 Martin Jambor <mjambor@suse.cz>
3872 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3873 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3874 to get param_ipa_sra_max_replacements.
3875 (param_splitting_across_edge): Pass the caller to
3876 pull_accesses_from_callee.
3878 2020-01-10 Martin Jambor <mjambor@suse.cz>
3880 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3881 * ipa-cp.c (max_new_size): Removed.
3882 (orig_overall_size): New variable.
3883 (get_max_overall_size): New function.
3884 (estimate_local_effects): Use it. Adjust dump.
3885 (decide_about_value): Likewise.
3886 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3887 orig_overall_size. Adjust dump.
3888 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3890 2020-01-10 Martin Jambor <mjambor@suse.cz>
3892 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3893 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3894 instead of param_ipa_max_agg_items.
3895 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3896 optimization info for the callee.
3898 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3900 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3901 markers if debug_inline_points is false.
3903 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3905 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3907 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3908 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3909 aarch64-sve-builtins-sve2.h.
3910 (aarch64-sve-builtins-sve2.o): New rule.
3911 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3912 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3913 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3914 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3915 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3916 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3918 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3919 instructions that are handled here.
3920 (@cond_asrd<mode>): Generalize to...
3921 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3922 (*cond_asrd<mode>_2): Generalize to...
3923 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3924 (*cond_asrd<mode>_z): Generalize to...
3925 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3926 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3927 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3928 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3929 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3931 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3932 (@aarch64_scatter_stnt<mode>): Likewise.
3933 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3934 (@aarch64_mul_lane_<mode>): Likewise.
3935 (@aarch64_sve_suqadd<mode>_const): Likewise.
3936 (*<sur>h<addsub><mode>): Generalize to...
3937 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3939 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3940 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3941 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3942 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3943 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3944 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3945 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3946 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3947 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3948 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3949 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3950 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3951 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3952 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3953 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3954 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3955 (@aarch64_sve2_xar<mode>): Likewise.
3956 (@aarch64_sve2_bcax<mode>): Likewise.
3957 (*aarch64_sve2_eor3<mode>): Rename to...
3958 (@aarch64_sve2_eor3<mode>): ...this.
3959 (@aarch64_sve2_bsl<mode>): New expander.
3960 (@aarch64_sve2_nbsl<mode>): Likewise.
3961 (@aarch64_sve2_bsl1n<mode>): Likewise.
3962 (@aarch64_sve2_bsl2n<mode>): Likewise.
3963 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3964 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3965 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3966 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3967 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3968 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3969 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3970 (<su>mull<bt><Vwide>): Generalize to...
3971 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3973 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3974 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3975 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3976 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3977 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3978 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3979 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3980 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3981 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3982 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3983 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3984 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3985 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3986 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3987 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3988 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3989 (<SHRNB:r>shrnb<mode>): Generalize to...
3990 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3992 (<SHRNT:r>shrnt<mode>): Generalize to...
3993 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3995 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3996 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3997 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3998 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3999 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
4000 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
4001 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
4002 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
4003 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
4004 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
4005 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
4006 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
4007 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
4008 (@aarch64_sve2_cvtnt<mode>): Likewise.
4009 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
4010 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
4011 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
4012 (@aarch64_sve2_cvtxnt<mode>): Likewise.
4013 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
4014 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
4015 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
4016 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
4017 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
4018 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
4019 (@aarch64_sve2_pmul<mode>): Likewise.
4020 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
4021 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
4022 (@aarch64_sve2_tbl2<mode>): Likewise.
4023 (@aarch64_sve2_tbx<mode>): Likewise.
4024 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
4025 (@aarch64_sve2_histcnt<mode>): Likewise.
4026 (@aarch64_sve2_histseg<mode>): Likewise.
4027 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
4028 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
4029 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
4030 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
4031 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
4032 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
4033 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
4034 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
4035 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
4036 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4037 (SVE2_PMULL_PAIR_I): New mode iterators.
4038 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4039 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4040 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4041 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4042 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4043 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4044 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4045 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4046 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4047 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4048 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4049 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4050 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4051 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4052 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4053 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4054 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4055 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4056 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4057 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4058 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4059 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4060 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4061 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4062 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4063 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4064 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4065 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4066 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4067 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4068 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4070 (VNARROW, Ventype): New mode attributes.
4071 (Vewtype): Handle VNx2DI. Fix typo in comment.
4072 (VDOUBLE): New mode attribute.
4073 (sve_lane_con): Handle VNx8HI.
4074 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4075 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4076 (sve_int_op, sve_int_op_rev): Handle the above codes.
4077 (sve_pred_int_rhs2_operand): Likewise.
4078 (MULLBT, SHRNB, SHRNT): Delete.
4079 (SVE_INT_SHIFT_IMM): New int iterator.
4080 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4081 and UNSPEC_WHILEHS for TARGET_SVE2.
4082 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4083 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4084 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4085 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4086 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4087 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4088 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4089 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4090 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4091 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4092 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4093 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4094 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4095 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4096 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4097 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4098 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4099 (optab): Handle the new unspecs.
4100 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4102 (lr): Handle the new unspecs.
4104 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4105 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4106 (sve_int_qsub_op): New int attributes.
4107 (sve_fp_op, rot): Handle the new unspecs.
4108 * config/aarch64/aarch64-sve-builtins.h
4109 (function_resolver::require_matching_pointer_type): Declare.
4110 (function_resolver::resolve_unary): Add an optional boolean argument.
4111 (function_resolver::finish_opt_n_resolution): Add an optional
4112 type_suffix_index argument.
4113 (gimple_folder::redirect_call): Declare.
4114 (gimple_expander::prepare_gather_address_operands): Add an optional
4116 * config/aarch64/aarch64-sve-builtins.cc: Include
4117 aarch64-sve-builtins-sve2.h.
4118 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4119 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4120 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4121 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4122 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4123 (TYPES_s_unsigned): Likewise.
4124 (TYPES_s_integer): Use TYPES_s_unsigned.
4125 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4126 (TYPES_sd_integer): Use them.
4127 (TYPES_d_unsigned): New macro.
4128 (TYPES_d_integer): Use it.
4129 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4130 (TYPES_cvt_narrow): Likewise.
4131 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4132 (preds_mx): New variable.
4133 (function_builder::add_overloaded_function): Allow the new feature
4134 set to be more restrictive than the original one.
4135 (function_resolver::infer_pointer_type): Remove qualifiers from
4136 the pointer type before printing it.
4137 (function_resolver::require_matching_pointer_type): New function.
4138 (function_resolver::resolve_sv_displacement): Handle functions
4139 that don't support 32-bit vector indices or svint32_t vector offsets.
4140 (function_resolver::finish_opt_n_resolution): Take the inferred type
4141 as a separate argument.
4142 (function_resolver::resolve_unary): Optionally treat all forms in
4143 the same way as normal merging functions.
4144 (gimple_folder::redirect_call): New function.
4145 (function_expander::prepare_gather_address_operands): Add an argument
4146 that says whether scaled forms are available. If they aren't,
4147 handle scaling of vector indices and don't add the extension and
4149 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4150 fall back to using cond_* instead.
4151 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4152 Split out the member variables into...
4153 (rtx_code_function_base): ...this new base class.
4154 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4155 (unspec_based_function): Split out the member variables into...
4156 (unspec_based_function_base): ...this new base class.
4157 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4158 (unspec_based_function_exact_insn): New class.
4159 (unspec_based_add_function, unspec_based_add_lane_function)
4160 (unspec_based_lane_function, unspec_based_pred_function)
4161 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4162 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4163 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4165 (unspec_based_fused_function): New class.
4166 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4167 (unspec_based_fused_lane_function): New class.
4168 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4170 (CODE_FOR_MODE1): New macro.
4171 (fixed_insn_function): New class.
4172 (while_comparison): Likewise.
4173 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4174 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4175 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4176 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4177 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4178 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4179 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4180 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4181 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4182 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4183 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4184 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4185 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4186 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4187 Also add an initial argument for unary_convert_narrowt, regardless
4188 of the predication type.
4189 (build_32_64): Allow loads and stores to specify MODE_none.
4190 (build_sv_index64, build_sv_uint_offset): New functions.
4191 (long_type_suffix): New function.
4192 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4193 (binary_imm_long_base, load_gather_sv_base): Likewise.
4194 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4195 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4196 (unary_narrowb_base, unary_narrowt_base): Likewise.
4197 (binary_long_lane_def, binary_long_lane): New shape.
4198 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4199 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4200 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4201 (binary_to_uint_def, binary_to_uint): Likewise.
4202 (binary_wide_def, binary_wide): Likewise.
4203 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4204 (compare_def, compare): Likewise.
4205 (compare_ptr_def, compare_ptr): Likewise.
4206 (load_ext_gather_index_restricted_def,
4207 load_ext_gather_index_restricted): Likewise.
4208 (load_ext_gather_offset_restricted_def,
4209 load_ext_gather_offset_restricted): Likewise.
4210 (load_gather_sv_def): Inherit from load_gather_sv_base.
4211 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4212 (shift_left_imm_def, shift_left_imm): Likewise.
4213 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4214 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4215 (store_scatter_index_restricted_def,
4216 store_scatter_index_restricted): Likewise.
4217 (store_scatter_offset_restricted_def,
4218 store_scatter_offset_restricted): Likewise.
4219 (tbl_tuple_def, tbl_tuple): Likewise.
4220 (ternary_long_lane_def, ternary_long_lane): Likewise.
4221 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4222 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4223 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4224 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4225 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4226 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4227 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4228 (ternary_uint_def, ternary_uint): Likewise.
4229 (unary_convert): Fix typo in comment.
4230 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4231 (unary_long_def, unary_long): Likewise.
4232 (unary_narrowb_def, unary_narrowb): Likewise.
4233 (unary_narrowt_def, unary_narrowt): Likewise.
4234 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4235 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4236 (unary_to_int_def, unary_to_int): Likewise.
4237 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4238 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4239 (svasrd_impl): Delete.
4240 (svcadd_impl::expand): Handle integer operations too.
4241 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4242 new functions to derive the unspec numbers.
4243 (svmla_svmls_lane_impl): Replace with...
4244 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4245 integer operations too.
4246 (svwhile_impl): Rename to...
4247 (svwhilelx_impl): ...this and inherit from while_comparison.
4248 (svasrd): Use unspec_based_function.
4249 (svmla_lane): Use svmla_lane_impl.
4250 (svmls_lane): Use svmls_lane_impl.
4251 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4252 (svwhilele, svwhilelt): Use svwhilelx_impl.
4253 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4254 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4255 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4256 * config/aarch64/aarch64-sve-builtins.def: Include
4257 aarch64-sve-builtins-sve2.def.
4259 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4261 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4262 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4263 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4264 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4265 immediates as well as vector ones.
4266 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4267 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4268 (aarch64_sve_qsub_immediate): Update calls accordingly.
4270 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4272 * config/aarch64/aarch64-sve2.md: Add banner comments.
4273 (<su>mulh<r>s<mode>3): Move further up file.
4274 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4275 (*aarch64_sve2_sra<mode>): Move further down file.
4276 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4278 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4280 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4282 (while_optab_cmp): Handle them.
4283 * config/aarch64/aarch64-sve.md
4284 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4285 and add a "@" marker.
4286 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4287 instead of gen_aarch64_sve2_while_ptest.
4288 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4290 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4292 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4293 (UNSPEC_WHILELE): ...this.
4294 (UNSPEC_WHILE_LO): Rename to...
4295 (UNSPEC_WHILELO): ...this.
4296 (UNSPEC_WHILE_LS): Rename to...
4297 (UNSPEC_WHILELS): ...this.
4298 (UNSPEC_WHILE_LT): Rename to...
4299 (UNSPEC_WHILELT): ...this.
4300 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4301 (cmp_op, while_optab_cmp): Likewise.
4302 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4303 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4304 (svwhilelt): Likewise.
4306 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4308 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4309 (unary_to_uint): Define.
4310 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4311 (unary_count): Rename to...
4312 (unary_to_uint_def, unary_to_uint): ...this.
4313 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4315 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4317 * config/aarch64/aarch64-sve-builtins-functions.h
4318 (code_for_mode_function): New class.
4319 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4320 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4321 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4322 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4323 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4325 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4327 * config/aarch64/iterators.md (addsub): New code attribute.
4328 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4330 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4331 in the asm string and attributes. Fix indentation.
4332 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4334 (@aarch64_sve_<optab><mode>): ...this.
4335 * config/aarch64/aarch64-sve-builtins.h
4336 (function_expander::expand_signed_unpred_op): Delete.
4337 * config/aarch64/aarch64-sve-builtins.cc
4338 (function_expander::expand_signed_unpred_op): Likewise.
4339 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4340 try using code_for_aarch64_sve instead.
4341 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4342 (svqsub_impl): Likewise.
4343 (svqadd, svqsub): Use rtx_code_function instead.
4345 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4347 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4348 (HADDSUB, sur, addsub): Remove them.
4350 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4352 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4353 functions the same way, rather than singling out those that
4354 aren't mapped directly to optabs.
4356 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4358 * target.def (compatible_vector_types_p): New target hook.
4359 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4360 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4361 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4362 * doc/tm.texi: Regenerate.
4363 * gimple-expr.c: Include target.h.
4364 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4365 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4367 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4368 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4369 Use the original predicate if it already has a suitable type.
4371 2020-01-09 Martin Jambor <mjambor@suse.cz>
4373 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4374 resolve_speculation and redirect_call_stmt_to_callee static. Change
4375 return type of set_call_stmt to cgraph_edge *.
4376 * auto-profile.c (afdo_indirect_call): Adjust call to
4377 redirect_call_stmt_to_callee.
4378 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4379 make the this pointer explicit, adjust self-recursive calls and the
4380 call top make_direct. Return the resulting edge.
4381 (cgraph_edge::remove): Make this pointer explicit.
4382 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4383 (cgraph_edge::make_direct): Likewise, adjust call to
4384 resolve_speculation.
4385 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4386 call to set_call_stmt.
4387 (cgraph_update_edges_for_call_stmt_node): Update call to
4388 set_call_stmt and remove.
4389 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4390 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4391 (cgraph_node::create_edge_including_clones): Moved "first" definition
4392 of edge to the block where it was used. Adjusted calls to
4394 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4395 cgraph_edge::remove.
4396 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4397 make_direct and redirect_call_stmt_to_callee.
4398 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4399 resolve_speculation and make_direct.
4400 * ipa-inline-transform.c (inline_transform): Adjust call to
4401 redirect_call_stmt_to_callee.
4402 (check_speculations_1):: Adjust call to resolve_speculation.
4403 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4404 resolve-speculation.
4405 (inline_small_functions): Adjust call to resolve_speculation.
4406 (ipa_inline): Likewise.
4407 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4409 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4410 safe with regards to edge removal, adjust calls to
4411 redirect_call_stmt_to_callee.
4412 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4413 and redirect_call_stmt_to_callee.
4414 * multiple_target.c (create_dispatcher_calls): Adjust call to
4415 redirect_call_stmt_to_callee
4416 (redirect_to_specific_clone): Likewise.
4417 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4418 Adjust calls to cgraph_edge::remove.
4419 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4420 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4421 (expand_call_inline): Adjust call to cgraph_edge::remove.
4423 2020-01-09 Martin Liska <mliska@suse.cz>
4425 * params.opt: Set Optimization for
4426 param_max_speculative_devirt_maydefs.
4428 2020-01-09 Martin Sebor <msebor@redhat.com>
4432 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4434 2020-01-09 Martin Liska <mliska@suse.cz>
4436 * auto-profile.c (auto_profile): Use opt_for_fn
4438 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4439 (propagate_vals_across_arith_jfunc): Likewise.
4440 (hint_time_bonus): Likewise.
4441 (incorporate_penalties): Likewise.
4442 (good_cloning_opportunity_p): Likewise.
4443 (perform_estimation_of_a_value): Likewise.
4444 (estimate_local_effects): Likewise.
4445 (ipcp_propagate_stage): Likewise.
4446 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4447 (set_switch_stmt_execution_predicate): Likewise.
4448 (analyze_function_body): Likewise.
4449 * ipa-inline-analysis.c (offline_size): Likewise.
4450 * ipa-inline.c (early_inliner): Likewise.
4451 * ipa-prop.c (ipa_analyze_node): Likewise.
4452 (ipcp_transform_function): Likewise.
4453 * ipa-sra.c (process_scan_results): Likewise.
4454 (ipa_sra_summarize_function): Likewise.
4455 * params.opt: Rename ipcp-unit-growth to
4456 ipa-cp-unit-growth. Add Optimization for various
4457 IPA-related parameters.
4459 2020-01-09 Richard Biener <rguenther@suse.de>
4462 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4464 2020-01-09 Richard Biener <rguenther@suse.de>
4466 PR tree-optimization/93040
4467 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4469 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4471 * common/config/avr/avr-common.c (avr_option_optimization_table)
4472 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4474 2020-01-09 Martin Liska <mliska@suse.cz>
4476 * cgraphclones.c (symbol_table::materialize_all_clones):
4477 Use cgraph_node::dump_name.
4479 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4482 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4483 output_operand_lossage instead of gcc_unreachable.
4484 * doc/md.texi (riscv f constraint): Fix typo.
4487 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4488 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4489 CONST_SCALAR_INT_P instead of CONST_INT_P.
4490 (*subv<mode>4_1): Rename to ...
4491 (subv<mode>4_1): ... this.
4492 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4493 define_insn_and_split patterns.
4494 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4497 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4499 * vec.c (class selftest::count_dtor): New class.
4500 (selftest::test_auto_delete_vec): New test.
4501 (selftest::vec_c_tests): Call it.
4502 * vec.h (class auto_delete_vec): New class template.
4503 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4505 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4507 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4509 2020-01-08 Jim Wilson <jimw@sifive.com>
4511 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4512 use of TLS_MODEL_LOCAL_EXEC when not pic.
4514 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4516 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4519 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4522 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4523 *stack_protect_set_3 peephole2): Also check that the second
4524 insns source is general_operand.
4527 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4528 predicate for output operand instead of register_operand.
4529 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4530 memory destination and non-memory operands[2].
4532 2020-01-08 Martin Liska <mliska@suse.cz>
4534 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4535 ::dump_asm_name instead of (::name or ::asm_name).
4536 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4537 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4538 (analyze_functions): Likewise.
4539 (expand_all_functions): Likewise.
4540 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4541 (propagate_bits_across_jump_function): Likewise.
4542 (dump_profile_updates): Likewise.
4543 (ipcp_store_bits_results): Likewise.
4544 (ipcp_store_vr_results): Likewise.
4545 * ipa-devirt.c (dump_targets): Likewise.
4546 * ipa-fnsummary.c (analyze_function_body): Likewise.
4547 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4548 (process_hsa_functions): Likewise.
4549 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4550 (set_alias_uids): Likewise.
4551 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4552 * ipa-inline.c (recursive_inlining): Likewise.
4553 (inline_to_all_callers_1): Likewise.
4554 (ipa_inline): Likewise.
4555 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4556 (ipa_propagate_frequency): Likewise.
4557 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4558 (remove_described_reference): Likewise.
4559 * ipa-pure-const.c (worse_state): Likewise.
4560 (check_retval_uses): Likewise.
4561 (analyze_function): Likewise.
4562 (propagate_pure_const): Likewise.
4563 (propagate_nothrow): Likewise.
4564 (dump_malloc_lattice): Likewise.
4565 (propagate_malloc): Likewise.
4566 (pass_local_pure_const::execute): Likewise.
4567 * ipa-visibility.c (optimize_weakref): Likewise.
4568 (function_and_variable_visibility): Likewise.
4569 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4570 (ipa_discover_variable_flags): Likewise.
4571 * lto-streamer-out.c (output_function): Likewise.
4572 (output_constructor): Likewise.
4573 * tree-inline.c (copy_bb): Likewise.
4574 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4575 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4577 2020-01-08 Richard Biener <rguenther@suse.de>
4580 * tree-eh.c (sink_clobbers): Update virtual operands for
4581 the first and last stmt only. Add a dry-run capability.
4582 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4583 after CFG manipulations and in RPO order to catch all
4584 secondary opportunities reliably.
4586 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4589 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4591 2019-01-08 Richard Biener <rguenther@suse.de>
4594 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4595 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4596 virtual operand, also updating SSA use.
4597 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4598 Update stmt after resetting virtual operand.
4599 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4600 * gimple-iterator.c (gsi_remove): When not removing the stmt
4601 permanently do not delink immediate uses or mark the stmt modified.
4603 2020-01-08 Martin Liska <mliska@suse.cz>
4605 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4606 (ipa_call_context::estimate_size_and_time): Likewise.
4607 (inline_analyze_function): Likewise.
4609 2020-01-08 Martin Liska <mliska@suse.cz>
4611 * cgraph.c (cgraph_node::dump): Use systematically
4614 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4616 Add -nodevicespecs option for avr.
4619 * config/avr/avr.opt (-nodevicespecs): New driver option.
4620 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4621 "-specs=device-specs/..." if that option is not set.
4622 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4624 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4626 Implement 64-bit double functions for avr.
4629 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4630 --with-double-comparison.
4631 * doc/install.texi: Document them.
4632 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4633 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4634 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4635 * doc/invoke.texi (AVR Built-in Macros): Document them.
4636 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4637 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4638 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4640 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4643 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4644 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4645 when only building rm-profile multilibs.
4647 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4650 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4651 lattice for a value to check.
4652 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4653 finite propagation in self-recursive scc.
4655 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4657 * ipa-inline.c (caller_growth_limits): Restore the AND.
4659 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4661 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4662 (VEC_ALLREG_ALT): New iterator.
4663 (VEC_ALLREG_INT_MODE): New iterator.
4664 (VCMP_MODE): New iterator.
4665 (VCMP_MODE_INT): New iterator.
4666 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4667 (vec_cmp<u>v64qidi): New define_expand.
4668 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4669 (vec_cmpu<mode>di_exec): New define_expand.
4670 (vec_cmp<u>v64qidi_exec): New define_expand.
4671 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4672 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4673 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4674 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4675 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4676 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4677 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4678 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4679 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4680 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4682 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4683 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4685 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4687 * config/gcn/constraints.md (DA): Update description and match.
4689 (Db): New constraint.
4690 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4692 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4693 Implement 'Db' mixed immediate type.
4694 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4695 (addcv64si3_dup<exec_vcc>): Delete.
4696 (subcv64si3<exec_vcc>): Rework constraints.
4697 (addv64di3): Rework constraints.
4698 (addv64di3_exec): Rework constraints.
4699 (subv64di3): Rework constraints.
4700 (addv64di3_dup): Delete.
4701 (addv64di3_dup_exec): Delete.
4702 (addv64di3_zext): Rework constraints.
4703 (addv64di3_zext_exec): Rework constraints.
4704 (addv64di3_zext_dup): Rework constraints.
4705 (addv64di3_zext_dup_exec): Rework constraints.
4706 (addv64di3_zext_dup2): Rework constraints.
4707 (addv64di3_zext_dup2_exec): Rework constraints.
4708 (addv64di3_sext_dup2): Rework constraints.
4709 (addv64di3_sext_dup2_exec): Rework constraints.
4711 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4713 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4714 existing target checks.
4716 2020-01-07 Richard Biener <rguenther@suse.de>
4718 * doc/install.texi: Bump minimal supported MPC version.
4720 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4722 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4723 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4724 * langhooks.c: Include stor-layout.h.
4725 (lhd_simulate_enum_decl): New function.
4726 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4727 handle_arm_sve_h for the LTO frontend.
4728 (register_vector_type): Cope with null returns from pushdecl.
4730 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4732 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4733 (aarch64_sve::nvectors_if_data_type): Replace with...
4734 (aarch64_sve::builtin_type_p): ...this.
4735 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4736 (find_vector_type): Delete.
4737 (add_sve_type_attribute): New function.
4738 (lookup_sve_type_attribute): Likewise.
4739 (register_builtin_types): Add an "SVE type" attribute to each type.
4740 (register_tuple_type): Likewise.
4741 (svbool_type_p, nvectors_if_data_type): Delete.
4742 (mangle_builtin_type): Use lookup_sve_type_attribute.
4743 (builtin_type_p): Likewise. Add an overload that returns the
4744 number of constituent vector and predicate registers.
4745 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4746 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4747 instead of aarch64_sve_argument_p.
4748 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4749 (aarch64_pass_by_reference): Likewise.
4750 (aarch64_function_value_1): Likewise.
4751 (aarch64_return_in_memory): Likewise.
4752 (aarch64_layout_arg): Likewise.
4754 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4756 PR tree-optimization/93156
4757 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4758 least significant bit is always clear.
4760 PR tree-optimization/93118
4761 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4762 simplifier with two intermediate conversions.
4764 2020-01-07 Martin Liska <mliska@suse.cz>
4766 * params.opt: Add Optimization for various parameters.
4768 2020-01-07 Martin Liska <mliska@suse.cz>
4771 * doc/extend.texi: Explain cloning for target_clone
4774 2020-01-07 Martin Liska <mliska@suse.cz>
4776 PR tree-optimization/92860
4777 * common.opt: Make in Optimization option
4778 as it is affected by -O0, which is an Optimization
4780 * tree-inline.c (tree_inlinable_function_p):
4781 Use opt_for_fn for warn_inline.
4782 (expand_call_inline): Likewise.
4784 2020-01-07 Martin Liska <mliska@suse.cz>
4786 PR tree-optimization/92860
4787 * common.opt: Make flag_ree as optimization
4790 2020-01-07 Martin Liska <mliska@suse.cz>
4792 PR optimization/92860
4793 * params.opt: Mark param_min_crossjump_insns with Optimization
4796 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4798 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4799 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4801 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4803 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4804 helper function to return the valid addressing formats for a given
4805 hard register and mode.
4806 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4808 * config/rs6000/constraints.md (Q constraint): Update
4810 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4813 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4814 Use 'Q' for doing vector extract from memory.
4815 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4817 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4818 doing vector extract from memory.
4819 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4820 extract from memory.
4822 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4823 for the offset being 34-bits when -mcpu=future is used.
4825 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4827 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4828 instead of cmpib_comparison_operator in cmpib patterns.
4829 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4830 of cmpib_comparison_operator. Revise comment.
4832 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4834 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4835 in an IFN_DIV_POW2 node to be equal.
4837 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4839 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4840 (vect_check_scalar_mask): ...this.
4841 (vectorizable_store, vectorizable_load): Update call accordingly.
4842 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4843 argument in calls to conditional internal functions.
4845 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4847 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4848 '0' matching inputs.
4849 (subv64di3_exec): Likewise.
4851 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4853 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4854 * doc/md.texi (movstr): Likewise.
4856 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4858 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4861 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4863 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4865 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4866 to a temporary file and use move-if-change to update the real
4867 file where necessary.
4869 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4871 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4872 rather than Upa for CPY /M.
4874 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4876 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4879 2020-01-06 Martin Liska <mliska@suse.cz>
4881 PR tree-optimization/92860
4882 * params.opt: Mark param_max_combine_insns with Optimization
4885 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4888 * config/i386/i386.md (SWIDWI): New mode iterator.
4889 (DWI, dwi): Add TImode variants.
4890 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4891 <general_hilo_operand> instead of <general_operand>. Use
4892 CONST_SCALAR_INT_P instead of CONST_INT_P.
4893 (*addv<mode>4_1): Rename to ...
4894 (addv<mode>4_1): ... this.
4895 (QWI): New mode attribute.
4896 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4897 define_insn_and_split patterns.
4898 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4900 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4901 <general_hilo_operand> instead of <general_operand>.
4902 (*addcarry<mode>_1): New define_insn.
4903 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4905 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4907 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4908 Use "call" instead of "set".
4910 2020-01-03 Martin Jambor <mjambor@suse.cz>
4913 * ipa-cp.c (print_all_lattices): Skip functions without info.
4915 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4918 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4919 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4920 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4921 for 'e' simd clones.
4924 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4926 (mprefer-vector-width=): Add Save.
4927 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4928 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4929 (ix86_debug_options, ix86_function_specific_print): Adjust
4930 ix86_target_string callers.
4931 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4932 (ix86_valid_target_attribute_tree): Likewise.
4933 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4934 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4935 ix86_target_string caller.
4938 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4939 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4940 instead of gen_int_shift_amount + convert_modes.
4942 PR rtl-optimization/93088
4943 * loop-iv.c (find_single_def_src): Punt after looking through
4944 128 reg copies for regs with single definitions. Move definitions
4947 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4949 * config/arm/arm-c.c (arm_cpu_builtins): Define
4950 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4951 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4952 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4953 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4954 * config/arm/arm-tables.opt: Regenerated.
4955 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4956 arm_arch_i8mm and arm_arch_bf16 when enabled.
4957 * config/arm/arm.h (TARGET_I8MM): New macro.
4958 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4959 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4960 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4961 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4962 (v8_6_a_simd_variants): New.
4963 (v8_*_a_simd_variants): Add i8mm and bf16.
4964 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4966 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4969 * predict.c (compute_function_frequency): Don't call
4970 warn_function_cold on functions that already have cold attribute.
4972 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4975 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4976 COMDAT group function labels in .data.rel.ro.local section.
4977 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4980 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4981 comparison_operator in B and S integer comparisons. Likewise, use
4982 ordered_comparison_operator instead of cmpib_comparison_operator in
4984 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4986 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4988 Update copyright years.
4990 * gcc.c (process_command): Update copyright notice dates.
4991 * gcov-dump.c (print_version): Ditto.
4992 * gcov.c (print_version): Ditto.
4993 * gcov-tool.c (print_version): Ditto.
4994 * gengtype.c (create_file): Ditto.
4995 * doc/cpp.texi: Bump @copying's copyright year.
4996 * doc/cppinternals.texi: Ditto.
4997 * doc/gcc.texi: Ditto.
4998 * doc/gccint.texi: Ditto.
4999 * doc/gcov.texi: Ditto.
5000 * doc/install.texi: Ditto.
5001 * doc/invoke.texi: Ditto.
5003 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
5005 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
5008 2020-01-01 Jakub Jelinek <jakub@redhat.com>
5010 PR tree-optimization/93098
5011 * match.pd (popcount): For shift amounts, use integer_onep
5012 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
5013 tests. Make sure that precision is power of two larger than or equal
5014 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
5015 instead of ULL suffixed constants. Formatting fixes.
5017 Copyright (C) 2020 Free Software Foundation, Inc.
5019 Copying and distribution of this file, with or without modification,
5020 are permitted in any medium without royalty provided the copyright
5021 notice and this notice are preserved.