1 2020-03-02 David Malcolm <dmalcolm@redhat.com>
3 * doc/invoke.texi (-fanalyzer-show-duplicate-count): New.
5 2020-03-02 David Malcolm <dmalcolm@redhat.com>
7 * doc/invoke.texi (Static Analyzer Options): Add
8 -Wanalyzer-stale-setjmp-buffer to the list of options enabled
11 2020-03-02 Uroš Bizjak <ubizjak@gmail.com>
14 * config/i386/i386.md (movstrict<mode>): Allow only
15 registers with VALID_INT_MODE_P modes.
17 2020-03-02 Andrew Stubbs <ams@codesourcery.com>
19 * config/gcn/gcn-valu.md (dpp_move<mode>): New.
20 (reduc_insn): Use 'U' and 'B' operand codes.
21 (reduc_<reduc_op>_scal_<mode>): Allow all types.
22 (reduc_<reduc_op>_scal_v64di): Delete.
23 (*<reduc_op>_dpp_shr_<mode>): Allow all 1reg types.
24 (*plus_carry_dpp_shr_v64si): Change to ...
25 (*plus_carry_dpp_shr_<mode>): ... this and allow all 1reg int types.
26 (mov_from_lane63_v64di): Change to ...
27 (mov_from_lane63_<mode>): ... this, and allow all 64-bit modes.
28 * config/gcn/gcn.c (gcn_expand_dpp_shr_insn): Increase buffer size.
29 Support UNSPEC_MOV_DPP_SHR output formats.
30 (gcn_expand_reduc_scalar): Add "use_moves" reductions.
31 Add "use_extends" reductions.
32 (print_operand_address): Add 'I' and 'U' codes.
33 * config/gcn/gcn.md (unspec): Add UNSPEC_MOV_DPP_SHR.
35 2020-03-02 Martin Liska <mliska@suse.cz>
37 * lto-wrapper.c: Fix typo in comment about
40 2020-03-01 Martin Sebor <msebor@redhat.com>
43 * calls.c (init_attr_rdwr_indices): Correctly handle attribute.
45 2020-03-01 Martin Sebor <msebor@redhat.com>
48 * tree-ssa-strlen.c (count_nonzero_bytes): Set the size to that
49 of a pointer in the outermost ADDR_EXPRs.
51 2020-02-28 Jeff Law <law@redhat.com>
53 * config/v850/v850.h (STATIC_CHAIN_REGNUM): Change to r19.
54 * config/v850/v850.c (v850_asm_trampoline_template): Update
57 2020-02-28 Michael Meissner <meissner@linux.ibm.com>
60 * config/rs6000/vsx.md (vsx_extract_<mode>_<VS_scalar>mode_var):
63 2020-02-28 Martin Liska <mliska@suse.cz>
66 * configure.ac: Improve detection of ld_date by requiring
67 either two dashes or none.
68 * configure: Regenerate.
70 2020-02-28 Vladimir Makarov <vmakarov@redhat.com>
72 PR rtl-optimization/93564
73 * ira-color.c (assign_hard_reg): Prefer smaller hard regno when we
74 do not honor reg alloc order.
76 2020-02-27 Joel Hutton <Joel.Hutton@arm.com>
79 * config/aarch64/aarch64.c (aarch64_override_options): Fix
80 misleading warning string.
82 2020-02-27 Martin Sebor <msebor@redhat.com>
84 * doc/invoke.texi (-Wbuiltin-declaration-mismatch): Fix a typo.
86 2020-02-27 Michael Meissner <meissner@linux.ibm.com>
89 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
90 Split the insn into two parts. This insn only does variable
91 extract from a register.
92 (vsx_extract_<mode>_var_load, VSX_D iterator): New insn, do
93 variable extract from memory.
94 (vsx_extract_v4sf_var): Split the insn into two parts. This insn
95 only does variable extract from a register.
96 (vsx_extract_v4sf_var_load): New insn, do variable extract from
98 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Split the insn
99 into two parts. This insn only does variable extract from a
101 (vsx_extract_<mode>_var_load, VSX_EXTRACT_I iterator): New insn,
102 do variable extract from memory.
104 2020-02-27 Martin Jambor <mjambor@suse.cz>
105 Feng Xue <fxue@os.amperecomputing.com>
108 * ipa-cp.c (same_node_or_its_all_contexts_clone_p): Replaced with
109 new function calls_same_node_or_its_all_contexts_clone_p.
110 (cgraph_edge_brings_value_p): Use it.
111 (cgraph_edge_brings_value_p): Likewise.
112 (self_recursive_pass_through_p): Return false if caller is a clone.
113 (self_recursive_agg_pass_through_p): Likewise.
115 2020-02-27 Jan Hubicka <hubicka@ucw.cz>
118 * alias.c (ends_tbaa_access_path_p): Break out from ...
119 (component_uses_parent_alias_set_from): ... here.
120 * alias.h (ends_tbaa_access_path_p): Declare.
121 * tree-ssa-alias.c (access_path_may_continue_p): Break out from ...;
122 handle trailing arrays past end of tbaa access path.
123 (aliasing_component_refs_p): ... here; likewise.
124 (nonoverlapping_refs_since_match_p): Track TBAA segment of the access
125 path; disambiguate also past end of it.
126 (nonoverlapping_component_refs_p): Use only TBAA segment of the access
129 2020-02-27 Mihail Ionescu <mihail.ionescu@arm.com>
131 * (__ARM_NUM_LANES, __arm_lane, __arm_lane_q): Move to the
132 beginning of the file.
133 (vcreate_bf16, vcombine_bf16): New.
134 (vdup_n_bf16, vdupq_n_bf16): New.
135 (vdup_lane_bf16, vdup_laneq_bf16): New.
136 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
137 (vduph_lane_bf16, vduph_laneq_bf16): New.
138 (vset_lane_bf16, vsetq_lane_bf16): New.
139 (vget_lane_bf16, vgetq_lane_bf16): New.
140 (vget_high_bf16, vget_low_bf16): New.
141 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
142 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
143 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
144 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
145 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
146 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
147 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
148 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
149 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
150 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
151 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New.
152 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
153 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
154 (vreinterpretq_bf16_p128): New.
155 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
156 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
157 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
158 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
159 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
160 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
161 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
162 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
163 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
164 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
165 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
166 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
167 (vreinterpretq_p128_bf16): New.
168 * config/arm/arm_neon_builtins.def (VDX): Add V4BF.
170 (V_elem_l): Likewise.
173 (V_DOUBLE): Likewise.
174 (VDQX): Add V4BF and V8BF.
175 (V_two_elem, V_three_elem, V_four_elem): Likewise.
178 (V_double_vector_mode): Likewise.
179 (V_cmp_result): Likewise.
180 (V_uf_sclr): Likewise.
181 (V_sz_elem): Likewise.
182 (Is_d_reg): Likewise.
183 (V_mode_nunits): Likewise.
184 * config/arm/neon.md (neon_vdup_lane): Enable for BFloat16.
186 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
188 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
189 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
190 (<expander><mode>3<exec>): Likewise.
191 (<expander><mode>3): New.
192 (v<expander><mode>3): New.
193 (<expander><mode>3): New.
194 (<expander><mode>3<exec>): Rename to ...
195 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
196 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
198 2020-02-27 Alexandre Oliva <oliva@adacore.com>
200 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
203 2020-02-27 Richard Biener <rguenther@suse.de>
205 PR tree-optimization/93508
206 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
207 non-_CHK variants. Valueize their length arguments.
209 2020-02-27 Richard Biener <rguenther@suse.de>
211 PR tree-optimization/93953
212 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
213 to the hash-map entry.
215 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
217 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
219 2020-02-27 Mark Williams <mwilliams@fb.com>
221 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
222 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
223 -ffile-prefix-map and -fmacro-prefix-map.
224 * lto-streamer-out.c: Include file-prefix-map.h.
225 (lto_output_location): Remap the file part of locations.
227 2020-02-27 Jakub Jelinek <jakub@redhat.com>
230 * gimplify.c (gimplify_init_constructor): Don't promote readonly
231 DECL_REGISTER variables to TREE_STATIC.
233 PR tree-optimization/93582
234 PR tree-optimization/93945
235 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
236 non-zero INTEGER_CST second argument and ref->offset or ref->size
237 not a multiple of BITS_PER_UNIT.
239 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
241 * doc/install.texi (Binaries): Update description of BullFreeware.
243 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
247 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
248 C++ Language Options, Warning Options, and Static Analyzer
249 Options lists. Document negative form of options enabled by
250 default. Move some things around to more accurately sort
251 warnings by category.
252 (C++ Dialect Options, Warning Options, Static Analyzer
253 Options): Document negative form of options when enabled by
254 default. Move some things around to more accurately sort
255 warnings by category. Add some missing index entries.
258 2020-02-26 Carl Love <cel@us.ibm.com>
261 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
262 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
263 for the vector unsigned short arguments. It is also listed as the
264 name of the built-in for arguments vector unsigned short,
265 vector unsigned int and vector unsigned long long built-ins. The
266 name of the builtins for these arguments should be:
267 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
268 __builtin_crypto_vpmsumd respectively.
270 2020-02-26 Richard Biener <rguenther@suse.de>
272 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
273 and load permutation.
275 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
278 * optabs-tree.c (supportable_convert_operation): Reject types with
281 2020-02-26 David Malcolm <dmalcolm@redhat.com>
283 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
285 2020-02-26 Jakub Jelinek <jakub@redhat.com>
287 PR tree-optimization/93820
288 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
289 argument to ALL_INTEGER_CST_P boolean.
290 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
291 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
292 adjacent INTEGER_CST store into merged_store->only_constants like
295 2020-02-25 Jakub Jelinek <jakub@redhat.com>
298 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
300 * cfghooks.c (verify_flow_info): Likewise.
301 * predict.c (combine_predictions_for_bb): Likewise.
302 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
303 sucessor -> successor.
304 (find_traces_1_round): Fix comment typo, destinarion -> destination.
305 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
307 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
308 message typo, sucessors -> successors.
310 2020-02-25 Martin Sebor <msebor@redhat.com>
312 * doc/extend.texi (attribute access): Correct an example.
314 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
316 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
318 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
320 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
321 (VD): Enable for V4BF.
323 (VQ): Enable for V8BF.
326 (VDBL, Vdbl): Add V4BF.
327 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
328 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
329 (bfloat16x8x2_t): Likewise.
330 (bfloat16x4x3_t): Likewise.
331 (bfloat16x8x3_t): Likewise.
332 (bfloat16x4x4_t): Likewise.
333 (bfloat16x8x4_t): Likewise.
334 (vcombine_bf16): New.
335 (vld1_bf16, vld1_bf16_x2): New.
336 (vld1_bf16_x3, vld1_bf16_x4): New.
337 (vld1q_bf16, vld1q_bf16_x2): New.
338 (vld1q_bf16_x3, vld1q_bf16_x4): New.
339 (vld1_lane_bf16): New.
340 (vld1q_lane_bf16): New.
341 (vld1_dup_bf16): New.
342 (vld1q_dup_bf16): New.
345 (vld2_dup_bf16): New.
346 (vld2q_dup_bf16): New.
349 (vld3_dup_bf16): New.
350 (vld3q_dup_bf16): New.
353 (vld4_dup_bf16): New.
354 (vld4q_dup_bf16): New.
355 (vst1_bf16, vst1_bf16_x2): New.
356 (vst1_bf16_x3, vst1_bf16_x4): New.
357 (vst1q_bf16, vst1q_bf16_x2): New.
358 (vst1q_bf16_x3, vst1q_bf16_x4): New.
359 (vst1_lane_bf16): New.
360 (vst1q_lane_bf16): New.
368 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
370 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
371 (VALL_F16): Likewise.
372 (VALLDI_F16): Likewise.
375 (vswap_width_name): Likewise.
376 (VSWAP_WIDTH): Likewise.
380 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
381 (vget_lane_bf16, vgetq_lane_bf16): New.
383 (vdup_n_bf16, vdupq_n_bf16): New.
384 (vdup_lane_bf16, vdup_laneq_bf16): New.
385 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
386 (vduph_lane_bf16, vduph_laneq_bf16): New.
387 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
388 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
389 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
390 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
391 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
392 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
393 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
394 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
395 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
396 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
397 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
398 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
399 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
400 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
401 (vreinterpretq_bf16_p128): New.
402 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
403 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
404 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
405 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
406 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
407 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
408 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
409 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
410 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
411 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
412 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
413 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
414 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
415 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
416 (vreinterpretq_p128_bf16): New.
418 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
420 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
421 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
422 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
423 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
424 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
425 * config/arm/iterators.md (VSF2BF): New attribute.
426 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
427 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
428 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
430 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
432 * config/arm/arm.md (required_for_purecode): New attribute.
433 (enabled): Handle required_for_purecode.
434 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
435 work with -mpure-code.
437 2020-02-25 Jakub Jelinek <jakub@redhat.com>
439 PR rtl-optimization/93908
440 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
443 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
445 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
447 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
449 * doc/install.texi (--enable-checking): Adjust wording.
451 2020-02-25 Richard Biener <rguenther@suse.de>
453 PR tree-optimization/93868
454 * tree-vect-slp.c (slp_copy_subtree): New function.
455 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
456 re-arranging stmts in it.
458 2020-02-25 Jakub Jelinek <jakub@redhat.com>
461 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
462 dummy function and remove it at the end.
465 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
466 paramter -> parameter.
467 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
468 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
470 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
472 * doc/install.texi (--enable-checking): Properly document current
474 (--enable-stage1-checking): Minor clarification about bootstrap.
476 2020-02-24 David Malcolm <dmalcolm@redhat.com>
479 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
480 -fanalyzer-checker=taint is also required.
481 (-fanalyzer-checker=): Note that providing this option enables the
482 given checker, and doing so may be required for checkers that are
485 2020-02-24 David Malcolm <dmalcolm@redhat.com>
487 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
488 significant control flow events; add a "3" which shows all
489 control flow events; the old "3" becomes "4".
491 2020-02-24 Jakub Jelinek <jakub@redhat.com>
493 PR tree-optimization/93582
494 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
495 pd.offset and pd.size to be counted in bits rather than bytes, add
496 support for maxsizei that is not a multiple of BITS_PER_UNIT and
497 handle bitfield stores and loads.
498 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
499 uncomparable quantities - bytes vs. bits. Allow push_partial_def
500 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
501 pd.offset/pd.size to be counted in bits rather than bytes.
502 Formatting fix. Rename shadowed len variable to buflen.
504 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
505 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
508 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
509 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
510 * opts-common.c (parse_options_from_collect_gcc_options): New function.
511 (prepend_xassembler_to_collect_as_options): Likewise.
512 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
513 (prepend_xassembler_to_collect_as_options): Likewise.
514 * lto-opts.c (lto_write_options): Stream assembler options
515 in COLLECT_AS_OPTIONS.
516 * lto-wrapper.c (xassembler_options_error): New static variable.
517 (get_options_from_collect_gcc_options): Move parsing options code to
518 parse_options_from_collect_gcc_options and call it.
519 (merge_and_complain): Validate -Xassembler options.
520 (append_compiler_options): Handle OPT_Xassembler.
521 (run_gcc): Append command line -Xassembler options to
523 * doc/invoke.texi: Add documentation about using Xassembler
526 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
528 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
530 (riscv_rtx_costs): Update cost model for LTGT.
532 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
534 PR rtl-optimization/93564
535 * ira-color.c (struct update_cost_queue_elem): New member start.
536 (queue_update_cost, get_next_update_cost): Add new arg start.
537 (allocnos_conflict_p): New function.
538 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
539 Add checking conflicts with allocnos_conflict_p.
540 (update_costs_from_prefs, restore_costs_from_copies): Adjust
541 update_costs_from_allocno calls.
542 (update_conflict_hard_regno_costs): Add checking conflicts with
543 allocnos_conflict_p. Adjust calls of queue_update_cost and
544 get_next_update_cost.
545 (assign_hard_reg): Adjust calls of queue_update_cost. Add
547 (bucket_allocno_compare_func): Restore previous version.
549 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
551 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
552 double-word size when handling aggregate return values.
553 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
554 that homogeneous SFmode and DFmode aggregates are passed and returned
555 in general registers.
557 2020-02-21 Jakub Jelinek <jakub@redhat.com>
560 * opts.c (print_filtered_help): Translate help before appending
561 messages to it rather than after that.
563 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
565 PR rtl-optimization/PR92989
566 * lra-lives.c (process_bb_lives): Restore the original order
567 of the bb liveness update. Call make_hard_regno_dead for each
568 register clobbered at the start of an EH receiver.
570 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
573 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
574 self-recursively generated.
576 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
579 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
582 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
584 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
585 Document new target supports option.
587 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
589 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
590 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
591 * config/arm/iterators.md (MATMUL): New iterator.
592 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
593 (mmla_sfx): New attribute.
594 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
595 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
596 (UNSPEC_MATMUL_US): New.
598 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
600 * config/arm/arm.md: Prevent scalar shifts from being used when big
603 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
604 Richard Biener <rguenther@suse.de>
606 PR tree-optimization/93586
607 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
608 after mismatched array refs; do not sure type size information to
609 recover from unmatched referneces with !flag_strict_aliasing_p.
611 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
613 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
614 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
615 (scatter_store<mode>): Rename to ...
616 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
617 (scatter<mode>_exec): Delete. Move contents ...
618 (mask_scatter_store<mode>): ... here, and rename that to ...
619 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
620 Remove mode conversion.
621 (mask_gather_load<mode>): Rename to ...
622 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
623 Remove mode conversion.
624 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
626 2020-02-21 Martin Jambor <mjambor@suse.cz>
628 PR tree-optimization/93845
629 * tree-sra.c (verify_sra_access_forest): Only test access size of
632 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
634 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
635 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
636 (addv64di3_exec): Likewise.
637 (subv64di3): Likewise.
638 (subv64di3_exec): Likewise.
639 (addv64di3_zext): Likewise.
640 (addv64di3_zext_exec): Likewise.
641 (addv64di3_zext_dup): Likewise.
642 (addv64di3_zext_dup_exec): Likewise.
643 (addv64di3_zext_dup2): Likewise.
644 (addv64di3_zext_dup2_exec): Likewise.
645 (addv64di3_sext_dup2): Likewise.
646 (addv64di3_sext_dup2_exec): Likewise.
647 (<expander>v64di3): Likewise.
648 (<expander>v64di3_exec): Likewise.
649 (*<reduc_op>_dpp_shr_v64di): Likewise.
650 (*plus_carry_dpp_shr_v64di): Likewise.
651 * config/gcn/gcn.md (adddi3): Likewise.
652 (addptrdi3): Likewise.
653 (<expander>di3): Likewise.
655 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
657 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
659 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
661 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
662 support. Use aarch64_emit_mult instead of emitting multiplication
663 instructions directly.
664 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
665 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
667 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
669 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
670 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
671 instead of emitting multiplication instructions directly.
672 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
673 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
674 (@aarch64_frecps<mode>): New expanders.
676 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
678 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
679 on and produce uint64_ts rather than ints.
680 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
681 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
683 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
685 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
686 an unused xmsk register when handling approximate rsqrt.
688 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
690 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
691 flag_finite_math_only condition.
693 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
696 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
697 to destination operand for shufps alternative.
698 (*vec_extractv2si_1): Ditto.
700 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
703 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
706 2020-02-20 Martin Liska <mliska@suse.cz>
709 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
711 2020-02-20 Martin Liska <mliska@suse.cz>
714 * common/config/avr/avr-common.c: Remote trailing "|".
716 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
718 * collect2.c (maybe_run_lto_and_relink): Fix typo in
721 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
723 PR tree-optimization/93767
724 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
725 access-size bias from the offset calculations for negative strides.
727 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
729 * collect2.c (c_file, o_file): Make const again.
730 (ldout,lderrout, dump_ld_file): Remove.
731 (tool_cleanup): Avoid calling not signal-safe functions.
732 (maybe_run_lto_and_relink): Avoid possible signal handler
733 access to unintialzed memory (lto_o_files).
734 (main): Avoid leaking temp files in $TMPDIR.
735 Initialize c_file/o_file with concat, which avoids exposing
736 uninitialized memory to signal handler, which calls unlink(!).
737 Avoid calling maybe_unlink when the main function returns,
738 since the atexit handler is already doing this.
739 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
741 2020-02-19 Martin Jambor <mjambor@suse.cz>
743 PR tree-optimization/93776
744 * tree-sra.c (create_access): Do not create zero size accesses.
745 (get_access_for_expr): Do not search for zero sized accesses.
747 2020-02-19 Martin Jambor <mjambor@suse.cz>
749 PR tree-optimization/93667
750 * tree-sra.c (scalarizable_type_p): Return false if record fields
751 do not follow wach other.
753 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
755 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
756 rather than fmv.x.s/fmv.s.x.
758 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
760 * config/aarch64/aarch64-simd-builtins.def
761 (intrinsic_vec_smult_lo_): New.
762 (intrinsic_vec_umult_lo_): Likewise.
763 (vec_widen_smult_hi_): Likewise.
764 (vec_widen_umult_hi_): Likewise.
765 * config/aarch64/aarch64-simd.md
766 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
767 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
768 (vmull_high_s16): Likewise.
769 (vmull_high_s32): Likewise.
770 (vmull_high_u8): Likewise.
771 (vmull_high_u16): Likewise.
772 (vmull_high_u32): Likewise.
773 (vmull_s8): Likewise.
774 (vmull_s16): Likewise.
775 (vmull_s32): Likewise.
776 (vmull_u8): Likewise.
777 (vmull_u16): Likewise.
778 (vmull_u32): Likewise.
780 2020-02-18 Martin Liska <mliska@suse.cz>
782 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
783 bootstrap by missing removal of invalid sanity check.
785 2020-02-18 Martin Liska <mliska@suse.cz>
788 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
789 Always compare LHS of gimple_assign.
791 2020-02-18 Martin Liska <mliska@suse.cz>
794 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
795 and return type of functions.
796 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
797 Drop MALLOC attribute for void functions.
798 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
799 malloc_state for a new VOID clone.
801 2020-02-18 Martin Liska <mliska@suse.cz>
804 * common.opt: Add -fprofile-reproducibility.
805 * doc/invoke.texi: Document it.
806 * value-prof.c (dump_histogram_value):
807 Document and support behavior for counters[0]
808 being a negative value.
809 (get_nth_most_common_value): Handle negative
810 counters[0] in respect to flag_profile_reproducible.
812 2020-02-18 Jakub Jelinek <jakub@redhat.com>
815 * cgraph.c (verify_speculative_call): Use speculative_id instead of
816 speculative_uid in messages. Remove trailing whitespace from error
817 message. Use num_speculative_call_targets instead of
818 num_speculative_targets in a message.
819 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
820 edge messages and stmt instead of cal_stmt in reference message.
822 PR tree-optimization/93780
823 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
824 before calling build_vector_type.
825 (execute_update_addresses_taken): Likewise.
828 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
829 typo, functoin -> function.
830 * tree.c (free_lang_data_in_decl): Fix comment typo,
831 functoin -> function.
832 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
834 2020-02-17 David Malcolm <dmalcolm@redhat.com>
836 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
838 (print_option_information): Don't call get_option_url if URLs
841 2020-02-17 Alexandre Oliva <oliva@adacore.com>
843 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
844 handling of register_common-less targets.
846 2020-02-17 Martin Liska <mliska@suse.cz>
849 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
851 2020-02-17 Martin Liska <mliska@suse.cz>
854 * config/rs6000/rs6000.c (rs6000_option_override_internal):
857 2020-02-17 Martin Liska <mliska@suse.cz>
860 * config/rx/elf.opt: Fix typo.
862 2020-02-17 Richard Biener <rguenther@suse.de>
865 * opts-global.c (print_ignored_options): Use inform and
868 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
871 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
873 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
876 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
877 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
879 2020-02-15 Jason Merrill <jason@redhat.com>
881 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
883 2020-02-15 Jakub Jelinek <jakub@redhat.com>
885 PR tree-optimization/93744
886 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
887 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
888 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
889 sure @2 in the first and @1 in the other patterns has no side-effects.
891 2020-02-15 David Malcolm <dmalcolm@redhat.com>
892 Bernd Edlinger <bernd.edlinger@hotmail.de>
896 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
897 * configure.ac (--with-diagnostics-urls): New configuration
898 option, based on --with-diagnostics-color.
899 (DIAGNOSTICS_URLS_DEFAULT): New define.
900 * config.h: Regenerate.
901 * configure: Regenerate.
902 * diagnostic.c (diagnostic_urls_init): Handle -1 for
903 DIAGNOSTICS_URLS_DEFAULT from configure-time
904 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
905 and TERM_URLS environment variable.
906 * diagnostic-url.h (diagnostic_url_format): New enum type.
907 (diagnostic_urls_enabled_p): rename to...
908 (determine_url_format): ... this, and change return type.
909 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
910 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
911 the linux console, and mingw.
912 (diagnostic_urls_enabled_p): rename to...
913 (determine_url_format): ... this, and adjust.
914 * pretty-print.h (pretty_printer::show_urls): rename to...
915 (pretty_printer::url_format): ... this, and change to enum.
916 * pretty-print.c (pretty_printer::pretty_printer,
917 pp_begin_url, pp_end_url, test_urls): Adjust.
918 * doc/install.texi (--with-diagnostics-urls): Document the new
919 configuration option.
920 (--with-diagnostics-color): Document the existing interaction
921 with GCC_COLORS better.
922 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
923 vindex reference. Update description of defaults based on the above.
924 (-fdiagnostics-color): Update description of how -fdiagnostics-color
925 interacts with GCC_COLORS.
927 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
930 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
931 conjunction with TARGET_GNU_TLS in early return.
933 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
935 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
936 the mode is not wider than UNITS_PER_WORD.
938 2020-02-14 Martin Jambor <mjambor@suse.cz>
940 PR tree-optimization/93516
941 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
942 access of the same type as the parent.
943 (propagate_subaccesses_from_lhs): Likewise.
945 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
948 * config/i386/avx512vbmi2intrin.h
949 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
950 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
951 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
952 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
953 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
954 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
955 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
956 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
957 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
958 of lacking a closing parenthesis.
959 * config/i386/avx512vbmi2vlintrin.h
960 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
961 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
962 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
963 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
964 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
965 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
966 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
967 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
968 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
969 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
970 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
971 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
972 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
973 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
974 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
975 _mm_shldi_epi32, _mm_mask_shldi_epi32,
976 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
977 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
979 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
982 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
983 the target function entry.
985 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
987 * common/config/arc/arc-common.c (arc_option_optimization_table):
988 Disable if-conversion step when optimized for size.
990 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
992 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
993 R12-R15 are always in ARCOMPACT16_REGS register class.
994 * config/arc/arc.opt (mq-class): Deprecate.
995 * config/arc/constraint.md ("q"): Remove dependency on mq-class
997 * doc/invoke.texi (mq-class): Update text.
998 * common/config/arc/arc-common.c (arc_option_optimization_table):
1001 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1003 * config/arc/arc.c (arc_insn_cost): New function.
1004 (TARGET_INSN_COST): Define.
1005 * config/arc/arc.md (cost): New attribute.
1006 (add_n): Use arc_nonmemory_operand.
1007 (ashlsi3_insn): Likewise, also update constraints.
1008 (ashrsi3_insn): Likewise.
1009 (rotrsi3): Likewise.
1010 (add_shift): Likewise.
1011 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
1013 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
1015 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
1017 (umulsidi_600): Likewise.
1019 2020-02-13 Jakub Jelinek <jakub@redhat.com>
1022 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
1023 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
1024 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
1025 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
1026 pass __A to the builtin followed by __W instead of __A followed by
1028 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
1029 _mm512_mask_popcnt_epi64): Likewise.
1030 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
1031 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
1032 _mm256_mask_popcnt_epi64): Likewise.
1034 PR tree-optimization/93582
1035 * fold-const.h (shift_bytes_in_array_left,
1036 shift_bytes_in_array_right): Declare.
1037 * fold-const.c (shift_bytes_in_array_left,
1038 shift_bytes_in_array_right): New function, moved from
1039 gimple-ssa-store-merging.c, no longer static.
1040 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
1041 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
1042 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
1043 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
1044 shift_bytes_in_array.
1045 (verify_shift_bytes_in_array): Rename to ...
1046 (verify_shift_bytes_in_array_left): ... this. Use
1047 shift_bytes_in_array_left instead of shift_bytes_in_array.
1048 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
1049 instead of verify_shift_bytes_in_array.
1050 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
1051 / native_interpret_expr where the store covers all needed bits,
1052 punt on PDP-endian, otherwise allow all involved offsets and sizes
1053 not to be byte-aligned.
1056 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
1057 use const_0_to_255_operand predicate instead of immediate_operand.
1058 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
1059 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
1060 vgf2p8affineinvqb_<mode><mask_name>,
1061 vgf2p8affineqb_<mode><mask_name>): Drop mode from
1062 const_0_to_255_operand predicated operands.
1064 2020-02-12 Jeff Law <law@redhat.com>
1066 * config/h8300/h8300.md (comparison shortening peepholes): Use
1067 a mode iterator to merge the HImode and SImode peepholes.
1069 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1072 * real.c (is_even): Make static. Function comment fix.
1073 (is_halfway_below): Make static, don't assert R is not inf/nan,
1074 instead return false for those. Small formatting fixes.
1076 2020-02-12 Martin Sebor <msebor@redhat.com>
1079 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
1080 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
1081 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
1082 (strlen_check_and_optimize_call): Adjust callee name.
1084 2020-02-12 Jeff Law <law@redhat.com>
1086 * config/h8300/h8300.md (comparison shortening peepholes): Drop
1087 (and (xor)) variant. Combine other two into single peephole.
1089 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1091 PR rtl-optimization/93565
1092 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
1094 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
1096 * config/aarch64/aarch64-simd.md
1097 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
1098 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
1099 generating separate ADDV and zero_extend patterns.
1100 * config/aarch64/iterators.md (VDQV_E): New iterator.
1102 2020-02-12 Jeff Law <law@redhat.com>
1104 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
1105 expanders, splits, etc.
1106 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
1107 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
1108 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
1109 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
1110 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
1111 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
1113 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
1115 2020-02-12 Jakub Jelinek <jakub@redhat.com>
1118 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
1119 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
1120 TARGET_AVX512DQ from condition.
1121 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
1122 instead of <mask_mode512bit_condition> in condition. If
1123 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
1125 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
1128 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
1131 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
1133 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
1135 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
1136 where strlen is more legible.
1137 (rs6000_builtin_vectorized_libmass): Ditto.
1138 (rs6000_print_options_internal): Ditto.
1140 2020-02-11 Martin Sebor <msebor@redhat.com>
1142 PR tree-optimization/93683
1143 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
1145 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
1147 * config/rs6000/predicates.md (cint34_operand): Rename the
1148 -mprefixed-addr option to be -mprefixed.
1149 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
1150 the -mprefixed-addr option to be -mprefixed.
1151 (OTHER_FUTURE_MASKS): Likewise.
1152 (POWERPC_MASKS): Likewise.
1153 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
1154 the -mprefixed-addr option to be -mprefixed. Change error
1155 messages to refer to -mprefixed.
1156 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
1158 (rs6000_legitimate_offset_address_p): Likewise.
1159 (rs6000_mode_dependent_address): Likewise.
1160 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
1161 "-mprefixed" for target attributes and pragmas.
1162 (address_to_insn_form): Rename the -mprefixed-addr option to be
1164 (rs6000_adjust_insn_length): Likewise.
1165 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
1166 -mprefixed-addr option to be -mprefixed.
1167 (ASM_OUTPUT_OPCODE): Likewise.
1168 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
1169 -mprefixed-addr option to be -mprefixed.
1170 * config/rs6000/rs6000.opt (-mprefixed): Rename the
1171 -mprefixed-addr option to be prefixed. Change the option from
1172 being undocumented to being documented.
1173 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
1174 -mprefixed option. Update the -mpcrel documentation to mention
1177 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
1179 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
1180 including FIRST_PSEUDO_REGISTER - 1.
1181 * ira-color.c (print_hard_reg_set): Ditto.
1183 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1185 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1186 (USTERNOP_QUALIFIERS): New define.
1187 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1188 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1189 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1190 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1191 * config/arm/arm_neon.h (vusdot_s32): New.
1192 (vusdot_lane_s32): New.
1193 (vusdotq_lane_s32): New.
1194 (vsudot_lane_s32): New.
1195 (vsudotq_lane_s32): New.
1196 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1197 * config/arm/iterators.md (DOTPROD_I8MM): New.
1198 (sup, opsuffix): Add <us/su>.
1199 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1200 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1202 2020-02-11 Richard Biener <rguenther@suse.de>
1204 PR tree-optimization/93661
1205 PR tree-optimization/93662
1206 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1208 * tree-sra.c (get_access_for_expr): Likewise.
1210 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1213 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1214 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1215 Change condition from TARGET_AVX2 to TARGET_AVX.
1217 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1220 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1221 argument of strncmp.
1223 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1225 Try to generate zero-based comparisons.
1226 * config/cris/cris.c (cris_reduce_compare): New function.
1227 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1228 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1229 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1231 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1234 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1235 in Thumb state and also as a destination in Arm state. Add T16
1238 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1240 * md.texi (Define Subst): Match closing paren in example.
1242 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1246 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1247 arguments of strncmp.
1249 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1252 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1253 but different source value.
1254 (adjust_callers_for_value_intersection): New function.
1255 (gather_edges_for_value): Adjust order of callers to let a
1256 non-self-recursive caller be the first element.
1257 (self_recursive_pass_through_p): Add a new parameter "simple", and
1258 check generalized self-recursive pass-through jump function.
1259 (self_recursive_agg_pass_through_p): Likewise.
1260 (find_more_scalar_values_for_callers_subset): Compute value from
1261 pass-through jump function for self-recursive.
1262 (intersect_with_plats): Cleanup previous implementation code for value
1263 itersection with self-recursive call edge.
1264 (intersect_with_agg_replacements): Likewise.
1265 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1266 function for self-recursive call edge. Cleanup previous implementation
1267 code for value intersection with self-recursive call edge.
1268 (decide_whether_version_node): Remove dead callers and adjust order
1269 to let a non-self-recursive caller be the first element.
1271 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1273 * recog.c: Move pass_split_before_sched2 code in front of
1274 pass_split_before_regstack.
1275 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1276 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1277 (rest_of_handle_split_before_sched2): Remove.
1278 (pass_split_before_sched2::execute): Unconditionally call
1280 (enable_split_before_sched2): New function.
1281 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1282 (pass_split_before_regstack::gate): Ditto.
1283 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1284 Update name check for renamed split4 pass.
1285 * config/sh/sh.c (register_sh_passes): Update pass insertion
1286 point for renamed split4 pass.
1288 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1290 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1291 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1292 copying them around between host and target.
1294 2020-02-08 Andrew Pinski <apinski@marvell.com>
1297 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1298 STRICT_ALIGNMENT also.
1300 2020-02-08 Jim Wilson <jimw@sifive.com>
1303 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1305 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1306 Jakub Jelinek <jakub@redhat.com>
1309 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1310 xmm16-xmm31 call-used even in 64-bit ms-abi.
1312 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1314 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1315 (simd_ummla, simd_usmmla): Likewise.
1316 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1317 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1318 (vusmmlaq_s32): New.
1320 2020-02-07 Richard Biener <rguenther@suse.de>
1323 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1324 skipping unreachable regions.
1325 (optimize_inline_calls): Skip folding stmts when we didn't
1328 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1331 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1332 Don't return aggregates with only SFmode and DFmode in SSE
1334 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1336 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1339 * config/rs6000/rs6000-logue.c
1340 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1341 if it fails, move rs into end_addr and retry. Add
1342 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1343 the insn pattern doesn't describe well what exactly happens to
1347 * config/i386/predicates.md (avx_identity_operand): Remove.
1348 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1349 (avx_<castmode><avxsizesuffix>_<castmode>,
1350 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1351 a VEC_CONCAT of the operand and UNSPEC_CAST.
1352 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1353 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1357 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1358 recog_data.insn if distance_non_agu_define changed it.
1360 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1363 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1364 we only had X-FORM (reg+reg) addressing for vectors. Also before
1365 ISA 3.0, we only had X-FORM addressing for scalars in the
1366 traditional Altivec registers.
1368 2020-02-06 <zhongyunde@huawei.com>
1369 Vladimir Makarov <vmakarov@redhat.com>
1371 PR rtl-optimization/93561
1372 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1373 hard register range.
1375 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1377 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1380 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1382 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1383 where the low and the high 32 bits are equal to each other specially,
1384 with an rldimi instruction.
1386 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1388 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1390 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1392 * config/arm/arm-tables.opt: Regenerate.
1394 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1397 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1398 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1399 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1401 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1403 PR rtl-optimization/87763
1404 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1406 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1408 * config/aarch64/aarch64-simd-builtins.def
1409 (bfmlaq): New built-in function.
1410 (bfmlalb): New built-in function.
1411 (bfmlalt): New built-in function.
1412 (bfmlalb_lane): New built-in function.
1413 (bfmlalt_lane): New built-in function.
1414 * config/aarch64/aarch64-simd.md
1415 (aarch64_bfmmlaqv4sf): New pattern.
1416 (aarch64_bfmlal<bt>v4sf): New pattern.
1417 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1418 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1419 (vbfmlalbq_f32): New intrinsic.
1420 (vbfmlaltq_f32): New intrinsic.
1421 (vbfmlalbq_lane_f32): New intrinsic.
1422 (vbfmlaltq_lane_f32): New intrinsic.
1423 (vbfmlalbq_laneq_f32): New intrinsic.
1424 (vbfmlaltq_laneq_f32): New intrinsic.
1425 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1426 (bt): New int attribute.
1428 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1430 * config/i386/i386.md (*pushtf): Emit "#" instead of
1431 calling gcc_unreachable in insn output.
1434 (*pushsf_rex64): Ditto for alternatives other than 1.
1435 (*pushsf): Ditto for alternatives other than 1.
1437 2020-02-06 Martin Liska <mliska@suse.cz>
1439 PR gcov-profile/91971
1440 PR gcov-profile/93466
1441 * coverage.c (coverage_init): Revert mangling of
1442 path into filename. It can lead to huge filename length.
1443 Creation of subfolders seem more natural.
1445 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1448 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1449 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1450 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1452 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1455 * config/i386/predicates.md (avx_identity_operand): New predicate.
1456 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1457 define_insn_and_split.
1460 * omp-low.c (use_pointer_for_field): For nested constructs, also
1461 look for map clauses on target construct.
1462 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1463 taskreg_nesting_level.
1466 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1467 shared clause, call omp_notice_variable on outer context if any.
1469 2020-02-05 Jason Merrill <jason@redhat.com>
1472 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1473 non-zero address even if weak and not yet defined.
1475 2020-02-05 Martin Sebor <msebor@redhat.com>
1477 PR tree-optimization/92765
1478 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1479 * tree-ssa-strlen.c (compute_string_length): Remove.
1480 (determine_min_objsize): Remove.
1481 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1482 Avoid using type size as the upper bound on string length.
1483 (handle_builtin_string_cmp): Add an argument. Adjust.
1484 (strlen_check_and_optimize_call): Pass additional argument to
1485 handle_builtin_string_cmp.
1487 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1489 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1490 (*pushdi2_rex64 peephole2): Unconditionally split after
1492 (*ashl<mode>3_doubleword): Ditto.
1493 (*<shift_insn><mode>3_doubleword): Ditto.
1495 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1498 * config/rs6000/rs6000.c (get_vector_offset): Fix
1500 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1502 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1504 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1507 (Special Functions for Debugging the Analyzer): Update description
1508 of __analyzer_dump_exploded_nodes.
1510 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1513 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1514 include sets and not clobbers in the vzeroupper pattern.
1515 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1516 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1517 (*avx_vzeroupper_1): New define_insn_and_split.
1520 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1521 don't run when !optimize.
1522 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1525 2020-02-05 Richard Biener <rguenther@suse.de>
1528 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1529 checks before matching calls.
1531 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1533 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1534 function comment typo.
1537 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1538 simd_clone_create failed when i == 0, adjust clone->nargs by
1541 2020-02-05 Martin Liska <mliska@suse.cz>
1544 * doc/invoke.texi: Document that one should
1545 not combine ASLR and -fpch.
1547 2020-02-04 Richard Biener <rguenther@suse.de>
1549 PR tree-optimization/93538
1550 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1552 2020-02-04 Richard Biener <rguenther@suse.de>
1554 PR tree-optimization/91123
1555 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1556 (vn_walk_cb_data::last_vuse): New member.
1557 (vn_walk_cb_data::saved_operands): Likewsie.
1558 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1559 (vn_walk_cb_data::push_partial_def): Use finish.
1560 (vn_reference_lookup_2): Update last_vuse and use finish if
1561 we've saved operands.
1562 (vn_reference_lookup_3): Use finish and update calls to
1563 push_partial_defs everywhere. When translating through
1564 memcpy or aggregate copies save off operands and alias-set.
1565 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1566 operation for redundant store removal.
1568 2020-02-04 Richard Biener <rguenther@suse.de>
1570 PR tree-optimization/92819
1571 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1572 generating more stmts than before.
1574 2020-02-04 Martin Liska <mliska@suse.cz>
1576 * config/arm/arm.c (arm_gen_far_branch): Move the function
1577 outside of selftests.
1579 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1581 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1582 function to adjust PC-relative vector addresses.
1583 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1584 handle vectors with PC-relative addresses.
1586 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1588 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1590 (hard_reg_and_mode_to_addr_mask): Delete.
1591 (rs6000_adjust_vec_address): If the original vector address
1592 was REG+REG or REG+OFFSET and the element is not zero, do the add
1593 of the elements in the original address before adding the offset
1594 for the vector element. Use address_to_insn_form to validate the
1595 address using the register being loaded, rather than guessing
1596 whether the address is a DS-FORM or DQ-FORM address.
1598 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1600 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1601 to calculate the offset in memory from the start of a vector of a
1602 particular element. Add code to keep the element number in
1603 bounds if the element number is variable.
1604 (rs6000_adjust_vec_address): Move calculation of offset of the
1605 vector element to get_vector_offset.
1606 (rs6000_split_vec_extract_var): Do not do the initial AND of
1607 element here, move the code to get_vector_offset.
1609 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1611 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1614 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1616 * config/rs6000/constraints.md: Improve documentation.
1618 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1621 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1622 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1624 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1626 * config.gcc: Remove "carrizo" support.
1627 * config/gcn/gcn-opts.h (processor_type): Likewise.
1628 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1629 * config/gcn/gcn.opt (gpu_type): Likewise.
1630 * config/gcn/t-omp-device: Likewise.
1632 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1635 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1636 * config/arm/arm.c (arm_gen_far_branch): New function
1638 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1640 2020-02-03 Julian Brown <julian@codesourcery.com>
1641 Tobias Burnus <tobias@codesourcery.com>
1643 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1645 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1648 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1649 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1652 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1654 PR rtl-optimization/91333
1655 * ira-color.c (struct allocno_color_data): Add member
1657 (init_allocno_threads): Set the member up.
1658 (bucket_allocno_compare_func): Add compare hard reg
1661 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1663 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1665 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1666 * config.in: Regenerated.
1667 * configure: Regenerated.
1668 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1669 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1670 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1672 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1674 * configure: Regenerate.
1676 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1678 PR rtl-optimization/91333
1679 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1680 reg preferences comparison up.
1682 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1684 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1685 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1686 aarch64-sve-builtins-base.h.
1687 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1688 aarch64-sve-builtins-base.cc.
1689 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1690 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1692 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1693 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1694 (svcvtnt): New functions.
1695 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1696 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1697 (svcvtnt): New functions.
1698 (svcvt): Add a form that converts f32 to bf16.
1699 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1700 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1702 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1703 Treat B as bfloat16_t.
1704 (ternary_bfloat_lane_base): New class.
1705 (ternary_bfloat_def): Likewise.
1706 (ternary_bfloat): New shape.
1707 (ternary_bfloat_lane_def): New class.
1708 (ternary_bfloat_lane): New shape.
1709 (ternary_bfloat_lanex2_def): New class.
1710 (ternary_bfloat_lanex2): New shape.
1711 (ternary_bfloat_opt_n_def): New class.
1712 (ternary_bfloat_opt_n): New shape.
1713 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1714 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1715 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1716 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1717 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1718 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1719 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1720 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1721 the pattern off the narrow mode instead of the wider one.
1722 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1723 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1724 (sve_fp_op): Handle them.
1725 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1726 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1728 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1730 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1731 * config/aarch64/aarch64-modes.def (BF): Move definition before
1732 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1733 (SVE_MODES): Handle BF modes.
1734 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1736 (aarch64_full_sve_mode): Likewise.
1737 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1739 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1740 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1741 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1742 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1744 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1746 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1747 (TYPES_all_data): Add bf16.
1748 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1749 (register_tuple_type): Increase buffer size.
1750 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1751 (bf16): New type suffix.
1752 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1753 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1754 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1755 Change type from all_data to all_arith.
1756 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1759 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1760 Matthew Malcomson <matthew.malcomson@arm.com>
1761 Richard Sandiford <richard.sandiford@arm.com>
1763 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1764 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1765 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1766 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1767 __ARM_FEATURE_MATMUL_FP64.
1768 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1769 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1770 be disabled at the same time.
1771 (f32mm): New extension.
1772 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1773 (AARCH64_FL_F64MM): Bump to the next bit up.
1774 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1775 (TARGET_SVE_F64MM): New macros.
1776 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1777 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1778 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1779 (UNSPEC_ZIP2Q): New unspeccs.
1780 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1781 (optab, sur, perm_insn): Handle the new unspecs.
1782 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1783 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1784 TARGET_SVE_F64MM instead of separate tests.
1785 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1786 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1787 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1788 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1789 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1790 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1791 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1792 (TYPES_s_signed): New macro.
1793 (TYPES_s_integer): Use it.
1794 (TYPES_d_float): New macro.
1795 (TYPES_d_data): Use it.
1796 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1797 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1798 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1799 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1800 (svmmla): New shape.
1801 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1802 template parameters.
1803 (ternary_resize2_lane_base): Likewise.
1804 (ternary_resize2_base): New class.
1805 (ternary_qq_lane_base): Likewise.
1806 (ternary_intq_uintq_lane_def): Likewise.
1807 (ternary_intq_uintq_lane): New shape.
1808 (ternary_intq_uintq_opt_n_def): New class
1809 (ternary_intq_uintq_opt_n): New shape.
1810 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1811 (ternary_uintq_intq_def): New class.
1812 (ternary_uintq_intq): New shape.
1813 (ternary_uintq_intq_lane_def): New class.
1814 (ternary_uintq_intq_lane): New shape.
1815 (ternary_uintq_intq_opt_n_def): New class.
1816 (ternary_uintq_intq_opt_n): New shape.
1817 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1818 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1819 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1820 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1822 (svdotprod_lane_impl): ...this new class.
1823 (svmmla_impl, svusdot_impl): New classes.
1824 (svdot_lane): Update to use svdotprod_lane_impl.
1825 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1826 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1828 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1829 function, with no types defined.
1830 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1831 AARCH64_FL_I8MM functions.
1832 (svmmla): New AARCH64_FL_F32MM function.
1833 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1834 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1835 AARCH64_FL_F64MM function.
1836 (REQUIRED_EXTENSIONS):
1838 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1840 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1843 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1845 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1846 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1847 (*movti_internal): Do not check for
1848 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1849 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1850 just after check for TARGET_AVX.
1851 (*movdf_internal): Ditto.
1852 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1853 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1854 * config/i386/sse.md (mov<mode>_internal): Only check
1855 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1856 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1857 (<sse>_andnot<mode>3<mask_name>): Move check for
1858 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1859 (<code><mode>3<mask_name>): Ditto.
1860 (*andnot<mode>3): Ditto.
1861 (*andnottf3): Ditto.
1862 (*<code><mode>3): Ditto.
1863 (*<code>tf3): Ditto.
1864 (*andnot<VI:mode>3): Remove
1865 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1866 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1867 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1868 (sse4_1_blendv<ssemodesuffix>): Ditto.
1869 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1870 Explain that tune applies to 128bit instructions only.
1872 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1874 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1875 to definition of hsa_kernel_description. Parse assembly to find SGPR
1876 and VGPR count of kernel and store in hsa_kernel_description.
1878 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1880 PR rtl-optimization/91838
1881 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1882 to truncate if allowed or reject combination.
1884 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1886 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1887 (find_inv_vars_cb): Likewise.
1889 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1891 * calls.c (special_function_p): Split out the check for DECL_NAME
1892 being non-NULL and fndecl being extern at file scope into a
1893 new maybe_special_function_p and call it. Drop check for fndecl
1894 being non-NULL that was after a usage of DECL_NAME (fndecl).
1895 * tree.h (maybe_special_function_p): New inline function.
1897 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1899 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1900 (mask_gather_load<mode>): ... here, and zero-initialize the
1902 (maskload<mode>di): Zero-initialize the destination.
1905 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1908 * doc/analyzer.texi (Limitations): Note that constraints on
1909 floating-point values are currently ignored.
1911 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1914 * symtab.c (symtab_node::noninterposable_alias): If localalias
1915 already exists, but is not usable, append numbers after it until
1916 a unique name is found. Formatting fix.
1919 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1922 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1924 * config/gcn/gcn.c (print_operand): Handle LTGT.
1925 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1927 2020-01-30 Richard Biener <rguenther@suse.de>
1929 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1930 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1932 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1934 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1935 without a DECL in .data.rel.ro.local.
1937 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1940 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1944 * config/i386/sse.md
1945 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1946 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1947 any_extend code iterator instead of always zero_extend.
1948 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1949 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1950 Use any_extend code iterator instead of always zero_extend.
1951 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1952 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1953 Use any_extend code iterator instead of always zero_extend.
1954 (*sse2_pmovmskb_ext): New define_insn.
1955 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1958 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1959 (*popcountsi2_zext_falsedep): New define_insn.
1961 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1963 * config.in: Regenerated.
1964 * configure: Regenerated.
1966 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1969 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1970 LLVM's assembler changed the default in version 9.
1972 2020-01-24 Jeff Law <law@redhat.com>
1974 PR tree-optimization/89689
1975 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1977 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1981 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1983 PR rtl-optimization/87763
1984 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1985 simplification to handle subregs as well as bare regs.
1986 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1988 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1991 * ira.c (ira): Revert use of simplified LRA algorithm.
1993 2020-01-29 Martin Jambor <mjambor@suse.cz>
1995 PR tree-optimization/92706
1996 * tree-sra.c (struct access): Fields first_link, last_link,
1997 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1998 next_rhs_queued and grp_rhs_queued respectively, new fields
1999 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
2000 (struct assign_link): Field next renamed to next_rhs, new field
2001 next_lhs. Updated comment.
2002 (work_queue_head): Renamed to rhs_work_queue_head.
2003 (lhs_work_queue_head): New variable.
2004 (add_link_to_lhs): New function.
2005 (relink_to_new_repr): Also relink LHS lists.
2006 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
2007 (add_access_to_lhs_work_queue): New function.
2008 (pop_access_from_work_queue): Renamed to
2009 pop_access_from_rhs_work_queue.
2010 (pop_access_from_lhs_work_queue): New function.
2011 (build_accesses_from_assign): Also add links to LHS lists and to LHS
2013 (child_would_conflict_in_lacc): Renamed to
2014 child_would_conflict_in_acc. Adjusted parameter names.
2015 (create_artificial_child_access): New parameter set_grp_read, use it.
2016 (subtree_mark_written_and_enqueue): Renamed to
2017 subtree_mark_written_and_rhs_enqueue.
2018 (propagate_subaccesses_across_link): Renamed to
2019 propagate_subaccesses_from_rhs.
2020 (propagate_subaccesses_from_lhs): New function.
2021 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
2024 2020-01-29 Martin Jambor <mjambor@suse.cz>
2026 PR tree-optimization/92706
2027 * tree-sra.c (struct access): Adjust comment of
2028 grp_total_scalarization.
2029 (find_access_in_subtree): Look for single children spanning an entire
2031 (scalarizable_type_p): Allow register accesses, adjust callers.
2032 (completely_scalarize): Remove function.
2033 (scalarize_elem): Likewise.
2034 (create_total_scalarization_access): Likewise.
2035 (sort_and_splice_var_accesses): Do not track total scalarization
2037 (analyze_access_subtree): New parameter totally, adjust to new meaning
2038 of grp_total_scalarization.
2039 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
2040 (can_totally_scalarize_forest_p): New function.
2041 (create_total_scalarization_access): Likewise.
2042 (create_total_access_and_reshape): Likewise.
2043 (total_should_skip_creating_access): Likewise.
2044 (totally_scalarize_subtree): Likewise.
2045 (analyze_all_variable_accesses): Perform total scalarization after
2046 subaccess propagation using the new functions above.
2047 (initialize_constant_pool_replacements): Output initializers by
2048 traversing the access tree.
2050 2020-01-29 Martin Jambor <mjambor@suse.cz>
2052 * tree-sra.c (verify_sra_access_forest): New function.
2053 (verify_all_sra_access_forests): Likewise.
2054 (create_artificial_child_access): Set parent.
2055 (analyze_all_variable_accesses): Call the verifier.
2057 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2059 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
2060 if called on indirect edge.
2061 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
2062 speculative call if needed.
2064 2020-01-29 Richard Biener <rguenther@suse.de>
2066 PR tree-optimization/93428
2067 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
2068 permutation when the load node is created.
2069 (vect_analyze_slp_instance): Re-use it here.
2071 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2073 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
2075 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
2077 PR rtl-optimization/93272
2078 * ira-lives.c (process_out_of_region_eh_regs): New function.
2079 (process_bb_node_lives): Call it.
2081 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2083 * coverage.c (read_counts_file): Make error message lowercase.
2085 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2087 * profile-count.c (profile_quality_display_names): Fix ordering.
2089 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
2092 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
2093 hash only when edge is first within the sequence.
2094 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
2095 (symbol_table::create_edge): Do not set target_prob.
2096 (cgraph_edge::remove_caller): Watch for speculative calls when updating
2098 (cgraph_edge::make_speculative): Drop target_prob parameter.
2099 (cgraph_edge::speculative_call_info): Remove.
2100 (cgraph_edge::first_speculative_call_target): New member function.
2101 (update_call_stmt_hash_for_removing_direct_edge): New function.
2102 (cgraph_edge::resolve_speculation): Rewrite to new API.
2103 (cgraph_edge::speculative_call_for_target): New member function.
2104 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
2105 multiple speculation targets.
2106 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
2108 (verify_speculative_call): Verify that targets form an interval.
2109 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
2110 (cgraph_edge::first_speculative_call_target): New member function.
2111 (cgraph_edge::next_speculative_call_target): New member function.
2112 (cgraph_edge::speculative_call_target_ref): New member function.
2113 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
2114 (cgraph_edge): Remove target_prob.
2115 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
2116 Fix handling of speculative calls.
2117 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
2118 * ipa-fnsummary.c (analyze_function_body): Likewise.
2119 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
2120 * ipa-profile.c (dump_histogram): Fix formating.
2121 (ipa_profile_generate_summary): Watch for overflows.
2122 (ipa_profile): Do not require probablity to be 1/2; update to new API.
2123 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
2124 (update_indirect_edges_after_inlining): Update to new API.
2125 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
2127 * profile-count.h: (profile_probability::adjusted): New.
2128 * tree-inline.c (copy_bb): Update to new speculative call API; fix
2129 updating of profile.
2130 * value-prof.c (gimple_ic_transform): Rename to ...
2131 (dump_ic_profile): ... this one; update dumping.
2132 (stream_in_histogram_value): Fix formating.
2133 (gimple_value_profile_transformations): Update.
2135 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2138 * config/i386/i386.md (*movoi_internal_avx): Remove
2139 TARGET_SSE_TYPELESS_STORES check.
2140 (*movti_internal): Prefer TARGET_AVX over
2141 TARGET_SSE_TYPELESS_STORES.
2142 (*movtf_internal): Likewise.
2143 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
2144 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
2145 from TARGET_SSE_TYPELESS_STORES.
2147 2020-01-28 David Malcolm <dmalcolm@redhat.com>
2149 * diagnostic-core.h (warning_at): Rename overload to...
2150 (warning_meta): ...this.
2151 (emit_diagnostic_valist): Delete decl of overload taking
2152 diagnostic_metadata.
2153 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
2154 (warning_at): Rename overload taking diagnostic_metadata to...
2155 (warning_meta): ...this.
2157 2020-01-28 Richard Biener <rguenther@suse.de>
2159 PR tree-optimization/93439
2160 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
2161 * tree-cfg.c (move_sese_region_to_fn): ... here.
2162 (verify_types_in_gimple_reference): Verify used cliques are
2165 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
2168 * config/i386/i386-options.c (set_ix86_tune_features): Add an
2169 argument of a pointer to struct gcc_options and pass it to
2170 parse_mtune_ctrl_str.
2171 (ix86_function_specific_restore): Pass opts to
2172 set_ix86_tune_features.
2173 (ix86_option_override_internal): Likewise.
2174 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
2175 gcc_options and use it for x_ix86_tune_ctrl_string.
2177 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2179 PR rtl-optimization/87763
2180 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
2181 simplification to handle subregs as well as bare regs.
2182 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
2184 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2186 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2187 for reduction chains that (now) include a call.
2189 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2191 PR tree-optimization/92822
2192 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2193 out the don't-care elements of a vector whose significant elements
2194 are duplicates, make the don't-care elements duplicates too.
2196 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2198 PR tree-optimization/93434
2199 * tree-predcom.c (split_data_refs_to_components): Record which
2200 components have had aliasing loads removed. Prevent store-store
2201 commoning for all such components.
2203 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2206 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2207 -1 or is_vshift is true, use new_vector with number of elts npatterns
2208 rather than new_unary_operation.
2210 PR tree-optimization/93454
2211 * gimple-fold.c (fold_array_ctor_reference): Perform
2212 elt_size.to_uhwi () just once, instead of calling it in every
2213 iteration. Punt if that value is above size of the temporary
2214 buffer. Decrease third native_encode_expr argument when
2215 bufoff + elt_sz is above size of buf.
2217 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2219 * config/mips/mips.c (mips_declare_object_name)
2220 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2222 2020-01-27 Martin Liska <mliska@suse.cz>
2224 PR gcov-profile/93403
2225 * tree-profile.c (gimple_init_gcov_profiler): Generate
2226 both __gcov_indirect_call_profiler_v4 and
2227 __gcov_indirect_call_profiler_v4_atomic.
2229 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2232 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2234 (@aarch64_split_simd_mov<mode>): Use it.
2235 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2236 Leave the vec_extract patterns to handle 2-element vectors.
2237 (aarch64_simd_mov_from_<mode>high): Likewise.
2238 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2239 (vec_extractv2dfv1df): Likewise.
2241 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2243 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2244 jump conditions for *compare_condjump<GPI:mode>.
2246 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2249 * digraph.cc (test_edge::test_edge): Specify template for base
2252 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2254 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2256 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2258 * config/arc/arc-protos.h (gen_mlo): Remove.
2259 (gen_mhi): Likewise.
2260 * config/arc/arc.c (AUX_MULHI): Define.
2261 (arc_must_save_reister): Special handling for r58/59.
2262 (arc_compute_frame_size): Consider mlo/mhi registers.
2263 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2265 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2266 mlo/mhi name selection.
2267 (arc_restore_callee_saves): Don't early restore blink when ISR.
2268 (arc_expand_prologue): Add mlo/mhi saving.
2269 (arc_expand_epilogue): Add mlo/mhi restoring.
2272 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2273 numbering when MUL64 option is used.
2274 (DWARF2_FRAME_REG_OUT): Define.
2275 * config/arc/arc.md (arc600_stall): New pattern.
2276 (VUNSPEC_ARC_ARC600_STALL): Define.
2277 (mulsi64): Use correct mlo/mhi registers.
2278 (mulsi_600): Clean it up.
2279 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2281 (mhi_operand): Likewise.
2283 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2284 Petro Karashchenko <petro.karashchenko@ring.com>
2286 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2287 attributes if needed.
2288 (prepare_move_operands): Generate special unspec instruction for
2290 (arc_isuncached_mem_p): Propagate uncached attribute to each
2292 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2293 (VUNSPEC_ARC_STDI): Likewise.
2294 (ALLI): New mode iterator.
2295 (mALLI): New mode attribute.
2296 (lddi): New instruction pattern.
2298 (stdidi_split): Split instruction for architectures which are not
2299 supporting ll64 option.
2300 (lddidi_split): Likewise.
2302 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2304 PR rtl-optimization/92989
2305 * lra-lives.c (process_bb_lives): Update the live-in set before
2306 processing additional clobbers.
2308 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2310 PR rtl-optimization/93170
2311 * cselib.c (cselib_invalidate_regno_val): New function, split out
2313 (cselib_invalidate_regno): ...here.
2314 (cselib_invalidated_by_call_p): New function.
2315 (cselib_process_insn): Iterate over all the hard-register entries in
2316 REG_VALUES and invalidate any that cross call-clobbered registers.
2318 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2320 * dojump.c (split_comparison): Use HONOR_NANS rather than
2321 HONOR_SNANS when splitting LTGT.
2323 2020-01-27 Martin Liska <mliska@suse.cz>
2326 * opts.c (print_filtered_help): Exclude language-specific
2327 options from --help=common unless enabled in all FEs.
2329 2020-01-27 Martin Liska <mliska@suse.cz>
2331 * opts.c (print_help): Exclude params from
2332 all except --help=param.
2334 2020-01-27 Martin Liska <mliska@suse.cz>
2337 * config/i386/i386-features.c (make_resolver_func):
2338 Align the code with ppc64 target implementation.
2339 Do not generate a unique name for resolver function.
2341 2020-01-27 Richard Biener <rguenther@suse.de>
2343 PR tree-optimization/93397
2344 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2345 converted reduction chain SLP graph adjustment.
2347 2020-01-26 Marek Polacek <polacek@redhat.com>
2350 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2353 2020-01-26 Jason Merrill <jason@redhat.com>
2356 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2359 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2361 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2362 (rx_setmem): Likewise.
2364 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2367 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2368 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2369 drop <di> from constraint of last operand.
2372 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2373 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2374 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2376 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2379 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2382 2020-01-24 Jeff Law <law@redhat.com>
2384 PR tree-optimization/92788
2385 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2388 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2391 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2392 *avx_vperm_broadcast_<mode>,
2393 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2394 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2395 Move before avx2_perm<mode>/avx512f_perm<mode>.
2398 * simplify-rtx.c (simplify_const_unary_operation,
2399 simplify_const_binary_operation): Punt for mode precision above
2400 MAX_BITSIZE_MODE_ANY_INT.
2402 2020-01-24 Andrew Pinski <apinski@marvell.com>
2404 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2407 2020-01-24 Jeff Law <law@redhat.com>
2410 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2411 for REGs. Call output_operand_lossage to get more reasonable
2414 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2416 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2417 gcn_fp_compare_operator.
2418 (vec_cmpu<mode>di): Use gcn_compare_operator.
2419 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2420 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2421 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2422 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2423 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2424 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2425 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2426 gcn_fp_compare_operator.
2427 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2428 gcn_fp_compare_operator.
2429 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2430 gcn_fp_compare_operator.
2431 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2432 gcn_fp_compare_operator.
2434 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2436 * doc/install.texi (Cross-Compiler-Specific Options): Document
2437 `--with-toolexeclibdir' option.
2439 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2441 * target.def (flags_regnum): Also mention effect on delay slot filling.
2442 * doc/tm.texi: Regenerate.
2444 2020-01-23 Jeff Law <law@redhat.com>
2446 PR translation/90162
2447 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2449 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2452 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2455 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2457 PR rtl-optimization/93402
2458 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2461 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2463 * config.in: Regenerated.
2464 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2465 for TARGET_LIBC_GNUSTACK.
2466 * configure: Regenerated.
2467 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2468 found to be 2.31 or greater.
2470 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2472 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2474 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2475 (mips_asm_file_end): New function. Delegate to
2476 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2477 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2479 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2482 * config/i386/i386-modes.def (POImode): New mode.
2483 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2484 * config/i386/i386.md (DPWI): New mode attribute.
2485 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2487 (QPWI): ... this. Use POI instead of OI for TImode.
2488 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2489 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2492 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2495 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2497 (speculation_tracker_rev): New pattern.
2498 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2499 Use speculation_tracker_rev to track the inverse condition.
2501 2020-01-23 Richard Biener <rguenther@suse.de>
2503 PR tree-optimization/93381
2504 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2505 alias-set of the def as argument and record the first one.
2506 (vn_walk_cb_data::first_set): New member.
2507 (vn_reference_lookup_3): Pass the alias-set of the current def
2508 to push_partial_def. Fix alias-set used in the aggregate copy
2510 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2511 * real.c (clear_significand_below): Fix out-of-bound access.
2513 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2516 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2517 New define_insn patterns.
2519 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2521 * doc/sourcebuild.texi (check-function-bodies): Add an
2522 optional target/xfail selector.
2524 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2526 PR rtl-optimization/93124
2527 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2528 bare USE and CLOBBER insns.
2530 2020-01-22 Andrew Pinski <apinski@marvell.com>
2532 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2534 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2537 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2538 diagnostic_manager into "ana" namespace.
2539 * selftest-run-tests.c (selftest::run_tests): Update for move of
2540 selftest::run_analyzer_selftests to
2541 ana::selftest::run_analyzer_selftests.
2543 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2545 * cfgexpand.c (union_stack_vars): Update the size.
2547 2020-01-22 Richard Biener <rguenther@suse.de>
2549 PR tree-optimization/93381
2550 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2551 throughout, handle all conversions the same.
2553 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2556 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2557 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2558 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2559 Call force_reg on high_in2 unconditionally.
2561 2020-01-22 Martin Liska <mliska@suse.cz>
2563 PR tree-optimization/92924
2564 * profile.c (compute_value_histograms): Divide
2567 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2570 * output.h (assemble_name_resolve): Declare.
2571 * varasm.c (assemble_name_resolve): New function.
2572 (assemble_name): Use it.
2573 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2575 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2577 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2578 update_web_docs_git instead of update_web_docs_svn.
2580 2020-01-21 Andrew Pinski <apinski@marvell.com>
2583 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2584 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2585 (*tlsgd_small_<mode>): Likewise.
2586 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2587 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2588 register. Convert that register back to dest using convert_mode.
2590 2020-01-21 Jim Wilson <jimw@sifive.com>
2592 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2595 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2596 Uros Bizjak <ubizjak@gmail.com>
2599 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2601 (legitimize_tls_address): Do GNU2 TLS address computation in
2602 ptr_mode and zero-extend result to Pmode.
2603 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2604 :P with :PTR and Pmode with ptr_mode.
2605 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2606 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2607 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2609 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2612 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2613 the last two operands are CONST_INT_P before using them as such.
2615 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2617 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2618 to get the integer element types.
2620 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2622 * config/aarch64/aarch64-sve-builtins.h
2623 (function_expander::convert_to_pmode): Declare.
2624 * config/aarch64/aarch64-sve-builtins.cc
2625 (function_expander::convert_to_pmode): New function.
2626 (function_expander::get_contiguous_base): Use it.
2627 (function_expander::prepare_gather_address_operands): Likewise.
2628 * config/aarch64/aarch64-sve-builtins-sve2.cc
2629 (svwhilerw_svwhilewr_impl::expand): Likewise.
2631 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2634 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2635 cfun->machine->label_is_assembled.
2636 (aarch64_print_patchable_function_entry): New.
2637 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2638 * config/aarch64/aarch64.h (struct machine_function): New field,
2641 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2644 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2647 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2650 * cgraph.c (cgraph_edge::resolve_speculation,
2651 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2652 call_stmt_site_hash.
2654 2020-01-21 Martin Liska <mliska@suse.cz>
2656 * config/rs6000/rs6000.c (common_mode_defined): Remove
2659 2020-01-21 Richard Biener <rguenther@suse.de>
2661 PR tree-optimization/92328
2662 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2663 type when value-numbering same-sized store by inserting a
2665 (eliminate_dom_walker::eliminate_stmt): When eliminating
2666 a redundant store handle bit-reinterpretation of the same value.
2668 2020-01-21 Andrew Pinski <apinski@marvel.com>
2671 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2673 (prepare_block_for_update): This. Use a worklist instead of
2676 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2678 * gcc/config/arm/arm.c (clear_operation_p):
2679 Initialise last_regno, skip first iteration
2680 based on the first_set value and use ints instead
2681 of the unnecessary HOST_WIDE_INTs.
2683 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2686 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2687 compare_mode other than SFmode or DFmode.
2689 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2692 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2693 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2694 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2696 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2698 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2700 2020-01-20 Andrew Pinski <apinski@marvell.com>
2703 * targhooks.c (default_print_patchable_function_entry): Use
2704 output_asm_insn to emit the nop instruction.
2706 2020-01-20 Fangrui Song <maskray@google.com>
2709 * targhooks.c (default_print_patchable_function_entry): Align to
2712 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2715 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2716 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2717 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2718 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2719 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2720 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2721 Remove the {q} suffix from lea.
2722 (*tls_dynamic_gnu2_call_64): Renamed to ...
2723 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2724 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2725 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2726 Pass Pmode to gen_tls_dynamic_gnu2_64.
2728 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2730 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2732 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2734 * config/aarch64/aarch64-sve-builtins-base.cc
2735 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2737 2020-01-20 Richard Biener <rguenther@suse.de>
2740 * dwarf2out.c (prune_unused_types): Unconditionally mark
2741 called function DIEs.
2743 2020-01-20 Martin Liska <mliska@suse.cz>
2745 PR tree-optimization/93199
2746 * tree-eh.c (struct leh_state): Add
2747 new field outer_non_cleanup.
2748 (cleanup_is_dead_in): Pass leh_state instead
2749 of eh_region. Add a checking that state->outer_non_cleanup
2750 points to outer non-clean up region.
2751 (lower_try_finally): Record outer_non_cleanup
2753 (lower_catch): Likewise.
2754 (lower_eh_filter): Likewise.
2755 (lower_eh_must_not_throw): Likewise.
2756 (lower_cleanup): Likewise.
2758 2020-01-20 Richard Biener <rguenther@suse.de>
2760 PR tree-optimization/93094
2761 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2762 (vect_transform_loop): Likewise.
2763 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2764 loop_vectorized_call to vect_transform_loop.
2765 * tree-vect-loop.c (vect_transform_loop): Pass down
2766 loop_vectorized_call to vect_loop_versioning.
2767 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2768 the earlier discovered loop_vectorized_call.
2770 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2772 * doc/contribute.texi: Update for SVN -> Git transition.
2773 * doc/install.texi: Likewise.
2775 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2778 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2779 speculative targets.
2780 (verify_speculative_call): New function
2781 (cgraph_node::verify_node): Use it.
2782 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2785 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2788 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2789 (cgraph_edge::make_direct): Remove all indirect targets.
2790 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2791 (cgraph_node::verify_node): Verify that only one call_stmt or
2792 lto_stmt_uid is set.
2793 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2795 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2796 (lto_output_ref): Simplify streaming of stmt.
2797 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2799 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2801 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2802 Mark parameter unused.
2804 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2806 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2808 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2810 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2812 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2814 * Makefile.in: Add coroutine-passes.o.
2815 * builtin-types.def (BT_CONST_SIZE): New.
2816 (BT_FN_BOOL_PTR): New.
2817 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2818 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2819 * coroutine-builtins.def: New file.
2820 * coroutine-passes.cc: New file.
2821 * function.h (struct GTY function): Add a bit to indicate that the
2822 function is a coroutine component.
2823 * internal-fn.c (expand_CO_FRAME): New.
2824 (expand_CO_YIELD): New.
2825 (expand_CO_SUSPN): New.
2826 (expand_CO_ACTOR): New.
2827 * internal-fn.def (CO_ACTOR): New.
2831 * passes.def: Add pass_coroutine_lower_builtins,
2832 pass_coroutine_early_expand_ifns.
2833 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2834 (make_pass_coroutine_early_expand_ifns): New.
2835 * doc/invoke.texi: Document the fcoroutines command line
2838 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2840 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2843 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2844 after checking the argument is a REG. Don't use REGNO (reg)
2845 again to set last_regno, reuse regno variable instead.
2847 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2849 * doc/analyzer.texi (Limitations): Add note about NaN.
2851 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2852 Sudakshina Das <sudi.das@arm.com>
2854 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2855 and valid immediate.
2856 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2857 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2858 * config/arm/constraints.md (Pg): New.
2859 * config/arm/predicates.md (long_shift_imm): New.
2860 (arm_reg_or_long_shift_imm): Likewise.
2861 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2862 (thumb2_lsll): Likewise.
2865 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2866 Sudakshina Das <sudi.das@arm.com>
2868 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2869 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2870 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2871 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2872 * config/arm/thumb2.md (thumb2_asrl): New.
2873 (thumb2_lsll): Likewise.
2875 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2877 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2880 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2882 * gdbinit.in (help-gcc-hooks): New command.
2883 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2884 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2887 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2889 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2890 correct target macro.
2892 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2894 * config/aarch64/aarch64-protos.h
2895 (aarch64_sve_ld1ro_operand_p): New.
2896 * config/aarch64/aarch64-sve-builtins-base.cc
2897 (class load_replicate): New.
2898 (class svld1ro_impl): New.
2899 (class svld1rq_impl): Change to inherit from load_replicate.
2900 (svld1ro): New sve intrinsic function base.
2901 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2902 New DEF_SVE_FUNCTION.
2903 * config/aarch64/aarch64-sve-builtins-base.h
2904 (svld1ro): New decl.
2905 * config/aarch64/aarch64-sve-builtins.cc
2906 (function_expander::add_mem_operand): Modify assert to allow
2908 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2910 * config/aarch64/aarch64.c
2911 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2912 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2913 (aarch64_sve_ld1ro_operand_p): New.
2914 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2915 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2916 * config/aarch64/predicates.md
2917 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2919 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2921 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2922 Introduce this ACLE specified predefined macro.
2923 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2924 (fp): Disabling this disables f64mm.
2925 (simd): Disabling this disables f64mm.
2926 (fp16): Disabling this disables f64mm.
2927 (sve): Disabling this disables f64mm.
2928 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2929 (AARCH64_ISA_F64MM): New.
2930 (TARGET_F64MM): New.
2931 * doc/invoke.texi (f64mm): Document new option.
2933 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2935 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2936 (neoversen1_tunings): Likewise.
2938 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2941 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2942 Add assert to ensure prolog has been emitted.
2943 (aarch64_split_atomic_op): Likewise.
2944 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2945 Use epilogue_completed rather than reload_completed.
2946 (aarch64_atomic_exchange<mode>): Likewise.
2947 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2948 (atomic_nand<mode>): Likewise.
2949 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2950 (atomic_fetch_nand<mode>): Likewise.
2951 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2952 (atomic_nand_fetch<mode>): Likewise.
2954 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2957 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2959 (REVERSE_CONDITION): Delete.
2960 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2961 (CCFP_CCFPE): Likewise.
2962 (e): New mode attribute.
2963 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2964 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2965 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2966 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2967 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2968 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2969 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2970 name of generator from gen_ccmpdi to gen_ccmpccdi.
2971 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2972 the previous comparison but aren't able to, use the new ccmp_rev
2975 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2977 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2978 than testing directly for INTEGER_CST.
2979 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2981 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2983 PR tree-optimization/93292
2984 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2985 get_vectype_for_scalar_type returns NULL.
2987 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2989 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2990 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2992 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2994 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2996 * params.opt: (max-predicted-iterations): Set bounds.
2997 * predict.c (real_almost_one, real_br_prob_base,
2998 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2999 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
3000 probabilities; do not truncate to reg_br_prob_bases.
3001 (estimate_loops_at_level): Pass max_cyclic_prob.
3002 (estimate_loops): Compute max_cyclic_prob.
3003 (estimate_bb_frequencies): Do not initialize real_*; update calculation
3005 * profile-count.c (profile_probability::to_sreal): New.
3006 * profile-count.h (class sreal): Move up in file.
3007 (profile_probability::to_sreal): Declare.
3009 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3012 (arm_invalid_conversion): New function for target hook.
3013 (arm_invalid_unary_op): New function for target hook.
3014 (arm_invalid_binary_op): New function for target hook.
3016 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3018 * config.gcc: Add arm_bf16.h.
3019 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
3020 (arm_simd_builtin_std_type): Add BFmode.
3021 (arm_init_simd_builtin_types): Define element types for vector types.
3022 (arm_init_bf16_types): New function.
3023 (arm_init_builtins): Add arm_init_bf16_types function call.
3024 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
3025 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
3026 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
3027 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
3028 (arm_vector_mode_supported_p): Add V4BF, V8BF.
3029 (arm_mangle_type): Add __bf16.
3030 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
3031 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
3032 arm_bf16_ptr_type_node.
3033 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
3034 define_split between ARM registers.
3035 * config/arm/arm_bf16.h: New file.
3036 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3037 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
3039 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
3040 * config/arm/vfp.md: Add BFmode to movhf patterns.
3042 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
3043 Andre Vieira <andre.simoesdiasvieira@arm.com>
3045 * config/arm/arm-cpus.in (mve, mve_float): New features.
3046 (dsp, mve, mve.fp): New options.
3047 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
3048 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
3049 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
3051 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3052 Thomas Preud'homme <thomas.preudhomme@arm.com>
3054 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
3056 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
3057 error for using -mcmse when targeting Armv8.1-M Mainline.
3059 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3060 Thomas Preud'homme <thomas.preudhomme@arm.com>
3062 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
3063 address in r4 when targeting Armv8.1-M Mainline.
3064 (nonsecure_call_value_internal): Likewise.
3065 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
3066 a register match_operand again. Emit BLXNS when targeting
3068 (nonsecure_call_value_reg_thumb2): Likewise.
3070 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3071 Thomas Preud'homme <thomas.preudhomme@arm.com>
3073 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
3074 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
3075 variable as true when floating-point ABI is not hard. Replace
3076 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
3077 Generate VLSTM and VLLDM instruction respectively before and
3078 after a function call to cmse_nonsecure_call function.
3079 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
3080 (VUNSPEC_VLLDM): Likewise.
3081 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
3082 (lazy_load_multiple_insn): Likewise.
3084 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3085 Thomas Preud'homme <thomas.preudhomme@arm.com>
3087 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
3088 (arm_emit_vfp_multi_reg_pop): Likewise.
3089 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
3090 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
3091 restore callee-saved VFP registers.
3093 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3094 Thomas Preud'homme <thomas.preudhomme@arm.com>
3096 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
3097 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
3098 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
3099 callee-saved GPRs as well as clear ip register before doing a nonsecure
3100 call then restore callee-saved GPRs after it when targeting
3102 (arm_reorg): Adapt to function rename.
3104 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3105 Thomas Preud'homme <thomas.preudhomme@arm.com>
3107 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
3108 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
3109 clear_vfp_multiple pattern based on a new vfp parameter.
3110 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
3111 targeting Armv8.1-M Mainline.
3112 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
3113 unconditionally when targeting Armv8.1-M Mainline architecture. Check
3114 whether VFP registers are available before looking call_used_regs for a
3116 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
3117 of prototype of clear_operation_p.
3118 (clear_vfp_multiple_operation): New predicate.
3119 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
3120 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
3122 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3123 Thomas Preud'homme <thomas.preudhomme@arm.com>
3125 * config/arm/arm-protos.h (clear_operation_p): Declare.
3126 * config/arm/arm.c (clear_operation_p): New function.
3127 (cmse_clear_registers): Generate clear_multiple instruction pattern if
3128 targeting Armv8.1-M Mainline or successor.
3129 (output_return_instruction): Only output APSR register clearing if
3130 Armv8.1-M Mainline instructions not available.
3131 (thumb_exit): Likewise.
3132 * config/arm/predicates.md (clear_multiple_operation): New predicate.
3133 * config/arm/thumb2.md (clear_apsr): New define_insn.
3134 (clear_multiple): Likewise.
3135 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
3137 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3138 Thomas Preud'homme <thomas.preudhomme@arm.com>
3140 * config/arm/arm.c (fp_sysreg_names): Declare and define.
3141 (use_return_insn): Also return false for Armv8.1-M Mainline.
3142 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
3143 Mainline instructions are available.
3144 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
3145 when targeting Armv8.1-M Mainline Security Extensions.
3146 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
3147 Mainline entry function.
3148 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
3149 targeting Armv8.1-M Mainline or successor.
3150 (arm_expand_epilogue): Fix indentation of caller-saved register
3151 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
3153 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
3154 (FP_SYSREGS): Likewise.
3155 (enum vfp_sysregs_encoding): Define enum.
3156 (fp_sysreg_names): Declare.
3157 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
3158 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
3159 (pop_fpsysreg_insn): Likewise.
3161 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
3162 Thomas Preud'homme <thomas.preudhomme@arm.com>
3164 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
3165 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
3166 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
3167 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
3168 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
3169 (ARMv8_1m_main): New feature group.
3170 (armv8.1-m.main): New architecture.
3171 * config/arm/arm-tables.opt: Regenerate.
3172 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
3173 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
3174 (arm_options_perform_arch_sanity_checks): Error out when targeting
3175 Armv8.1-M Mainline Security Extensions.
3176 * config/arm/arm.h (arm_arch8_1m_main): Declare.
3178 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3180 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
3181 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
3182 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
3183 aarch64_bfdot_laneq): New.
3184 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3185 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3186 vbfdotq_laneq_f32): New.
3187 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3188 VBFMLA_W, VBF): New.
3189 (isquadop): Add V4BF, V8BF.
3191 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3193 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3194 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3195 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3196 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3197 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3198 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3199 usdot_laneq, sudot_lane,sudot_laneq): New.
3200 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3201 (aarch64_<sur>dot_lane): New.
3202 * config/aarch64/arm_neon.h (vusdot_s32): New.
3204 (vusdot_lane_s32): New.
3205 (vsudot_lane_s32): New.
3206 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3207 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3209 2020-01-16 Martin Liska <mliska@suse.cz>
3211 * value-prof.c (dump_histogram_value): Fix
3212 obvious spacing issue.
3214 2020-01-16 Andrew Pinski <apinski@marvell.com>
3216 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3217 !storage_order_barrier_p.
3219 2020-01-16 Andrew Pinski <apinski@marvell.com>
3221 * sched-int.h (_dep): Add unused bit-field field for the padding.
3222 * sched-deps.c (init_dep_1): Init unused field.
3224 2020-01-16 Andrew Pinski <apinski@marvell.com>
3226 * optabs.h (create_expand_operand): Initialize target field also.
3228 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3230 PR tree-optimization/92429
3231 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3232 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3234 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3237 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3239 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3240 aarch64_sve_int_mode to each mode.
3242 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3244 * doc/analyzer.texi (Overview): Add note about
3245 -fdump-ipa-analyzer.
3247 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3249 PR tree-optimization/93231
3250 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3251 input_type is unsigned. Use tree_to_shwi for shift constant.
3252 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3253 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3254 inputs more efficiently.
3256 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3258 * config/i386/i386.md (*movsf_internal): Do not require
3259 SSE2 ISA for alternatives 14 and 15.
3261 2020-01-15 Richard Biener <rguenther@suse.de>
3264 * tree-eh.c (sink_clobbers): If we already visited the destination
3265 block do not defer insertion.
3266 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3267 the purpose of defered insertion.
3269 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3271 * BASE-VER: Bump to 10.0.1.
3273 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3275 PR tree-optimization/93247
3276 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3277 type of the stmt that we're going to vectorize.
3279 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3281 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3282 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3285 2020-01-15 Martin Liska <mliska@suse.cz>
3287 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3288 2 calls of streamer_read_hwi in a function call.
3290 2020-01-15 Richard Biener <rguenther@suse.de>
3292 * alias.c (record_alias_subset): Avoid redundant work when
3293 subset is already recorded.
3295 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3297 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3298 the analyzer options provide CWE identifiers.
3300 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3302 * tree-diagnostic-path.cc (path_summary::event_range::print):
3303 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3304 using get_pure_location.
3306 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3308 PR tree-optimization/93262
3309 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3310 perform head trimming only if the last argument is constant,
3311 either all ones, or larger or equal to head trim, in the latter
3312 case decrease the last argument by head_trim.
3314 PR tree-optimization/93249
3315 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3316 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3317 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3318 perform head trim unless we can prove there are no '\0' chars
3319 from the source among the first head_trim chars.
3321 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3323 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3325 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3328 * config/i386/sse.md
3329 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3330 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3331 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3332 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3333 just a single alternative instead of two, make operands 1 and 2
3336 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3339 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3342 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3344 * Makefile.in (lang_opt_files): Add analyzer.opt.
3345 (ANALYZER_OBJS): New.
3346 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3347 tristate.o and ANALYZER_OBJS.
3348 (TEXI_GCCINT_FILES): Add analyzer.texi.
3349 * common.opt (-fanalyzer): New driver option.
3350 * config.in: Regenerate.
3351 * configure: Regenerate.
3352 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3353 (gccdepdir): Also create depdir for "analyzer" subdir.
3354 * digraph.cc: New file.
3355 * digraph.h: New file.
3356 * doc/analyzer.texi: New file.
3357 * doc/gccint.texi ("Static Analyzer") New menu item.
3358 (analyzer.texi): Include it.
3359 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3360 ("Warning Options"): Add static analysis warnings to the list.
3361 (-Wno-analyzer-double-fclose): New option.
3362 (-Wno-analyzer-double-free): New option.
3363 (-Wno-analyzer-exposure-through-output-file): New option.
3364 (-Wno-analyzer-file-leak): New option.
3365 (-Wno-analyzer-free-of-non-heap): New option.
3366 (-Wno-analyzer-malloc-leak): New option.
3367 (-Wno-analyzer-possible-null-argument): New option.
3368 (-Wno-analyzer-possible-null-dereference): New option.
3369 (-Wno-analyzer-null-argument): New option.
3370 (-Wno-analyzer-null-dereference): New option.
3371 (-Wno-analyzer-stale-setjmp-buffer): New option.
3372 (-Wno-analyzer-tainted-array-index): New option.
3373 (-Wno-analyzer-use-after-free): New option.
3374 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3375 (-Wno-analyzer-use-of-uninitialized-value): New option.
3376 (-Wanalyzer-too-complex): New option.
3377 (-fanalyzer-call-summaries): New warning.
3378 (-fanalyzer-checker=): New warning.
3379 (-fanalyzer-fine-grained): New warning.
3380 (-fno-analyzer-state-merge): New warning.
3381 (-fno-analyzer-state-purge): New warning.
3382 (-fanalyzer-transitivity): New warning.
3383 (-fanalyzer-verbose-edges): New warning.
3384 (-fanalyzer-verbose-state-changes): New warning.
3385 (-fanalyzer-verbosity=): New warning.
3386 (-fdump-analyzer): New warning.
3387 (-fdump-analyzer-callgraph): New warning.
3388 (-fdump-analyzer-exploded-graph): New warning.
3389 (-fdump-analyzer-exploded-nodes): New warning.
3390 (-fdump-analyzer-exploded-nodes-2): New warning.
3391 (-fdump-analyzer-exploded-nodes-3): New warning.
3392 (-fdump-analyzer-supergraph): New warning.
3393 * doc/sourcebuild.texi (dg-require-dot): New.
3394 (dg-check-dot): New.
3395 * gdbinit.in (break-on-saved-diagnostic): New command.
3396 * graphviz.cc: New file.
3397 * graphviz.h: New file.
3398 * ordered-hash-map-tests.cc: New file.
3399 * ordered-hash-map.h: New file.
3400 * passes.def (pass_analyzer): Add before
3401 pass_ipa_whole_program_visibility.
3402 * selftest-run-tests.c (selftest::run_tests): Call
3403 selftest::ordered_hash_map_tests_cc_tests.
3404 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3406 * shortest-paths.h: New file.
3407 * timevar.def (TV_ANALYZER): New timevar.
3408 (TV_ANALYZER_SUPERGRAPH): Likewise.
3409 (TV_ANALYZER_STATE_PURGE): Likewise.
3410 (TV_ANALYZER_PLAN): Likewise.
3411 (TV_ANALYZER_SCC): Likewise.
3412 (TV_ANALYZER_WORKLIST): Likewise.
3413 (TV_ANALYZER_DUMP): Likewise.
3414 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3415 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3416 * tree-pass.h (make_pass_analyzer): New decl.
3417 * tristate.cc: New file.
3418 * tristate.h: New file.
3420 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3423 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3424 alternatives 9 and 10.
3426 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3428 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3429 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3430 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3431 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3432 (selftest::hash_map_tests_c_tests): Call it.
3433 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3434 New static constant, using the value of = H::empty_zero_p.
3435 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3436 from default_hash_traits <Value>.
3437 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3439 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3440 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3441 calls to mark_empty with !Descriptor::empty_zero_p.
3442 (hash_table::empty_slow): Conditionalize the memset call with a
3443 check that Descriptor::empty_zero_p; otherwise, loop through the
3444 entries calling mark_empty on them.
3445 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3446 (pointer_hash::empty_zero_p): Likewise.
3447 (pair_hash::empty_zero_p): Likewise.
3448 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3450 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3451 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3452 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3453 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3454 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3455 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3456 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3457 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3459 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3462 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3464 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3465 fix typo on return value.
3467 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3470 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3472 (cgraph_edge::make_speculative): Add param for setting speculative_id
3474 (cgraph_edge::speculative_call_info): Update comments and find reference
3475 by speculative_id for multiple indirect targets.
3476 (cgraph_edge::resolve_speculation): Decrease the speculations
3477 for indirect edge, drop it's speculative if not direct target
3478 left. Update comments.
3479 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3480 (cgraph_node::dump): Print num_speculative_call_targets.
3481 (cgraph_node::verify_node): Don't report error if speculative
3482 edge not include statement.
3483 (cgraph_edge::num_speculative_call_targets_p): New function.
3484 * cgraph.h (int common_target_id): Remove.
3485 (int common_target_probability): Remove.
3486 (num_speculative_call_targets): New variable.
3487 (make_speculative): Add param for setting speculative_id.
3488 (cgraph_edge::num_speculative_call_targets_p): New declare.
3489 (target_prob): New variable.
3490 (speculative_id): New variable.
3491 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3492 call summaries for multiple speculative call targets.
3493 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3494 * ipa-profile.c (struct speculative_call_target): New struct.
3495 (class speculative_call_summary): New class.
3496 (class speculative_call_summaries): New class.
3497 (call_sums): New variable.
3498 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3499 (ipa_profile_write_edge_summary): New function.
3500 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3501 (ipa_profile_dump_all_summaries): New function.
3502 (ipa_profile_read_edge_summary): New function.
3503 (ipa_profile_read_summary_section): New function.
3504 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3505 (ipa_profile): Generate num_speculative_call_targets from
3507 * ipa-ref.h (speculative_id): New variable.
3508 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3509 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3510 common_target_probability. Stream out speculative_id and
3511 num_speculative_call_targets.
3512 (input_edge): Likewise.
3513 * predict.c (dump_prediction): Remove edges count assert to be
3515 * symtab.c (symtab_node::create_reference): Init speculative_id.
3516 (symtab_node::clone_references): Clone speculative_id.
3517 (symtab_node::clone_referring): Clone speculative_id.
3518 (symtab_node::clone_reference): Clone speculative_id.
3519 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3520 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3521 if indirect call contains multiple speculative targets.
3522 * value-prof.h (check_ic_target): Remove.
3523 * value-prof.c (gimple_value_profile_transformations):
3524 Use void function gimple_ic_transform.
3525 * value-prof.c (gimple_ic_transform): Handle topn case.
3526 Fix comment typos. Change it to a void function.
3528 2020-01-13 Andrew Pinski <apinski@marvell.com>
3530 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3531 (octeontx2t98): New define.
3532 (octeontx2t96): New define.
3533 (octeontx2t93): New define.
3534 (octeontx2f95): New define.
3535 (octeontx2f95n): New define.
3536 (octeontx2f95mm): New define.
3537 * config/aarch64/aarch64-tune.md: Regenerate.
3538 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3540 2020-01-13 Jason Merrill <jason@redhat.com>
3542 PR c++/33799 - destroy return value if local cleanup throws.
3543 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3545 2020-01-13 Martin Liska <mliska@suse.cz>
3547 * ipa-cp.c (get_max_overall_size): Use newly
3548 renamed param param_ipa_cp_unit_growth.
3549 * params.opt: Remove legacy param name.
3551 2020-01-13 Martin Sebor <msebor@redhat.com>
3553 PR tree-optimization/93213
3554 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3555 stores to be eliminated.
3557 2020-01-13 Martin Liska <mliska@suse.cz>
3559 * opts.c (print_help): Do not print CL_PARAM
3560 and CL_WARNING for CL_OPTIMIZATION.
3562 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3565 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3566 depending on optimization settings.
3568 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3570 PR tree-optimization/90838
3571 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3572 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3573 argument rather than to initialize temporary for targets that
3574 don't use the mode argument at all. Initialize ctzval to avoid
3577 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3579 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3580 * tree-core.h: Document it.
3581 * gimplify.c (gimplify_omp_workshare): Set it.
3582 * omp-low.c (lower_omp_target): Use it.
3583 * tree-pretty-print.c (dump_omp_clause): Print it.
3585 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3586 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3588 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3590 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3591 * common.opt (fdiagnostics-path-format=): New option.
3592 (diagnostic_path_format): New enum.
3593 (fdiagnostics-show-path-depths): New option.
3594 * coretypes.h (diagnostic_event_id_t): New forward decl.
3595 * diagnostic-color.c (color_dict): Add "path".
3596 * diagnostic-event-id.h: New file.
3597 * diagnostic-format-json.cc (json_from_expanded_location): Make
3599 (json_end_diagnostic): Call context->make_json_for_path if it
3600 exists and the diagnostic has a path.
3601 (diagnostic_output_format_init): Clear context->print_path.
3602 * diagnostic-path.h: New file.
3603 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3604 when printing a run of events in a diagnostic_path so that they
3605 all get the same color.
3606 (layout::m_diagnostic_path_p): New field.
3607 (layout::layout): Initialize it.
3608 (layout::print_any_labels): Don't colorize the label text for an
3609 event in a diagnostic_path.
3610 (gcc_rich_location::add_location_if_nearby): Add
3611 "restrict_to_current_line_spans" and "label" params. Pass the
3612 former to layout.maybe_add_location_range; pass the latter
3613 when calling add_range.
3614 * diagnostic.c: Include "diagnostic-path.h".
3615 (diagnostic_initialize): Initialize context->path_format and
3616 context->show_path_depths.
3617 (diagnostic_show_any_path): New function.
3618 (diagnostic_path::interprocedural_p): New function.
3619 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3620 (simple_diagnostic_path::num_events): New function.
3621 (simple_diagnostic_path::get_event): New function.
3622 (simple_diagnostic_path::add_event): New function.
3623 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3624 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3625 (debug): New overload taking a diagnostic_path *.
3626 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3627 * diagnostic.h (enum diagnostic_path_format): New enum.
3628 (json::value): New forward decl.
3629 (diagnostic_context::path_format): New field.
3630 (diagnostic_context::show_path_depths): New field.
3631 (diagnostic_context::print_path): New callback field.
3632 (diagnostic_context::make_json_for_path): New callback field.
3633 (diagnostic_show_any_path): New decl.
3634 (json_from_expanded_location): New decl.
3635 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3636 (-fdiagnostics-show-path-depths): New option.
3637 (-fdiagnostics-color): Add "path" to description of default
3638 GCC_COLORS; describe it.
3639 (-fdiagnostics-format=json): Document how diagnostic paths are
3640 represented in the JSON output format.
3641 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3642 Add optional params "restrict_to_current_line_spans" and "label".
3643 * opts.c (common_handle_option): Handle
3644 OPT_fdiagnostics_path_format_ and
3645 OPT_fdiagnostics_show_path_depths.
3646 * pretty-print.c: Include "diagnostic-event-id.h".
3647 (pp_format): Implement "%@" format code for printing
3648 diagnostic_event_id_t *.
3649 (selftest::test_pp_format): Add tests for "%@".
3650 * selftest-run-tests.c (selftest::run_tests): Call
3651 selftest::tree_diagnostic_path_cc_tests.
3652 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3653 * toplev.c (general_init): Initialize global_dc->path_format and
3654 global_dc->show_path_depths.
3655 * tree-diagnostic-path.cc: New file.
3656 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3657 non-static. Drop "diagnostic" param in favor of storing the
3658 original value of "where" and re-using it.
3659 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3660 maybe_unwind_expanded_macro_loc.
3661 (tree_diagnostics_defaults): Initialize context->print_path and
3662 context->make_json_for_path.
3663 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3665 (default_tree_make_json_for_path): New decl.
3666 (maybe_unwind_expanded_macro_loc): New decl.
3668 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3670 PR tree-optimization/93210
3671 * fold-const.h (native_encode_initializer,
3672 can_native_interpret_type_p): Declare.
3673 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3675 (native_encode_initializer): New function, moved from dwarf2out.c.
3676 Adjust to native_encode_expr compatible arguments, including dry-run
3677 and partial extraction modes. Don't handle STRING_CST.
3678 (can_native_interpret_type_p): No longer static.
3679 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3680 offset / BITS_PER_UNIT fits into int and don't call it if
3681 can_native_interpret_type_p fails. If suboff is NULL and for
3682 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3683 native_encode_initializer.
3684 (fold_const_aggregate_ref_1): Formatting fix.
3685 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3686 (tree_add_const_value_attribute): Adjust caller.
3688 PR tree-optimization/90838
3689 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3690 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3691 CTZ_DEFINED_VALUE_AT_ZERO.
3693 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3696 * lra-constraints.c (match_reload): Permit input operands have the
3697 same mode as output while other input operands have a different
3700 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3702 PR tree-optimization/90838
3703 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3704 (check_ctz_string): Likewise.
3705 (optimize_count_trailing_zeroes): Likewise.
3706 (simplify_count_trailing_zeroes): Likewise.
3707 (pass_forwprop::execute): Try ctz simplification.
3708 * match.pd: Add matching for ctz idioms.
3710 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3712 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3714 (aarch64_invalid_unary_op): New function for target hook.
3715 (aarch64_invalid_binary_op): New function for target hook.
3717 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3719 * config.gcc: Add arm_bf16.h.
3720 * config/aarch64/aarch64-builtins.c
3721 (aarch64_simd_builtin_std_type): Add BFmode.
3722 (aarch64_init_simd_builtin_types): Define element types for vector
3724 (aarch64_init_bf16_types): New function.
3725 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3726 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3728 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3729 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3731 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3732 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3733 * config/aarch64/aarch64.c
3734 (aarch64_classify_vector_mode): Add support for BF types.
3735 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3736 (aarch64_vq_mode): Add support for BF types.
3737 (aarch64_simd_container_mode): Add support for BF types.
3738 (aarch64_mangle_type): Add support for BF scalar type.
3739 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3740 * config/aarch64/arm_bf16.h: New file.
3741 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3742 * config/aarch64/iterators.md: Add BF types to mode attributes.
3743 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3745 2020-01-10 Jason Merrill <jason@redhat.com>
3747 PR c++/93173 - incorrect tree sharing.
3748 * gimplify.c (copy_if_shared): No longer static.
3749 * gimplify.h: Declare it.
3751 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3753 * doc/invoke.texi (-msve-vector-bits=): Document that
3754 -msve-vector-bits=128 now generates VL-specific code for
3755 little-endian targets.
3756 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3757 build_vector_type_for_mode to construct the data vector types.
3758 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3759 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3760 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3761 for 128-bit vectors.
3763 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3765 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3768 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3770 * config/aarch64/aarch64-builtins.c
3771 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3772 rather than checking the number of elements and the element mode.
3774 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3776 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3777 get_related_vectype_for_scalar_type rather than build_vector_type
3778 to create the index type for a conditional reduction.
3780 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3782 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3783 for any type of gather or scatter, including strided accesses.
3785 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3787 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3790 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3792 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3794 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3795 parameter and its use to reset DR_OFFSET's.
3796 (vect_transform_loop): Remove orig_drs_init argument.
3797 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3798 member of dr_vec_info rather than the offset of the associated
3799 data_reference's innermost_loop_behavior.
3800 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3801 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3802 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3803 get_dr_vinfo_offset.
3804 (vectorizable_store): Likewise.
3805 (vectorizable_load): Likewise.
3807 2020-01-10 Richard Biener <rguenther@suse.de>
3809 * gimple-ssa-store-merging
3810 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3812 2020-01-10 Martin Liska <mliska@suse.cz>
3815 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3816 encapsulation that was there before r280040.
3818 2020-01-10 Richard Biener <rguenther@suse.de>
3821 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3822 sequences to avoid walking them again for secondary opportunities.
3823 (pass_lower_eh_dispatch::execute): Instead actually insert
3826 2020-01-10 Richard Biener <rguenther@suse.de>
3829 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3830 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3831 avoid quadraticness.
3833 2020-01-10 Martin Jambor <mjambor@suse.cz>
3835 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3836 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3837 to get param_ipa_sra_max_replacements.
3838 (param_splitting_across_edge): Pass the caller to
3839 pull_accesses_from_callee.
3841 2020-01-10 Martin Jambor <mjambor@suse.cz>
3843 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3844 * ipa-cp.c (max_new_size): Removed.
3845 (orig_overall_size): New variable.
3846 (get_max_overall_size): New function.
3847 (estimate_local_effects): Use it. Adjust dump.
3848 (decide_about_value): Likewise.
3849 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3850 orig_overall_size. Adjust dump.
3851 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3853 2020-01-10 Martin Jambor <mjambor@suse.cz>
3855 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3856 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3857 instead of param_ipa_max_agg_items.
3858 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3859 optimization info for the callee.
3861 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3863 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3864 markers if debug_inline_points is false.
3866 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3868 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3870 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3871 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3872 aarch64-sve-builtins-sve2.h.
3873 (aarch64-sve-builtins-sve2.o): New rule.
3874 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3875 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3876 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3877 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3878 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3879 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3881 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3882 instructions that are handled here.
3883 (@cond_asrd<mode>): Generalize to...
3884 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3885 (*cond_asrd<mode>_2): Generalize to...
3886 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3887 (*cond_asrd<mode>_z): Generalize to...
3888 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3889 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3890 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3891 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3892 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3894 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3895 (@aarch64_scatter_stnt<mode>): Likewise.
3896 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3897 (@aarch64_mul_lane_<mode>): Likewise.
3898 (@aarch64_sve_suqadd<mode>_const): Likewise.
3899 (*<sur>h<addsub><mode>): Generalize to...
3900 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3902 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3903 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3904 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3905 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3906 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3907 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3908 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3909 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3910 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3911 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3912 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3913 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3914 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3915 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3916 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3917 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3918 (@aarch64_sve2_xar<mode>): Likewise.
3919 (@aarch64_sve2_bcax<mode>): Likewise.
3920 (*aarch64_sve2_eor3<mode>): Rename to...
3921 (@aarch64_sve2_eor3<mode>): ...this.
3922 (@aarch64_sve2_bsl<mode>): New expander.
3923 (@aarch64_sve2_nbsl<mode>): Likewise.
3924 (@aarch64_sve2_bsl1n<mode>): Likewise.
3925 (@aarch64_sve2_bsl2n<mode>): Likewise.
3926 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3927 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3928 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3929 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3930 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3931 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3932 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3933 (<su>mull<bt><Vwide>): Generalize to...
3934 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3936 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3937 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3938 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3939 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3940 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3941 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3942 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3943 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3944 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3945 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3946 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3947 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3948 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3949 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3950 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3951 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3952 (<SHRNB:r>shrnb<mode>): Generalize to...
3953 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3955 (<SHRNT:r>shrnt<mode>): Generalize to...
3956 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3958 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3959 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3960 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3961 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3962 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3963 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3964 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3965 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3966 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3967 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3968 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3969 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3970 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3971 (@aarch64_sve2_cvtnt<mode>): Likewise.
3972 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3973 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3974 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3975 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3976 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3977 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3978 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3979 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3980 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3981 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3982 (@aarch64_sve2_pmul<mode>): Likewise.
3983 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3984 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3985 (@aarch64_sve2_tbl2<mode>): Likewise.
3986 (@aarch64_sve2_tbx<mode>): Likewise.
3987 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3988 (@aarch64_sve2_histcnt<mode>): Likewise.
3989 (@aarch64_sve2_histseg<mode>): Likewise.
3990 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3991 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3992 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3993 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3994 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3995 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3996 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3997 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3998 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3999 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
4000 (SVE2_PMULL_PAIR_I): New mode iterators.
4001 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
4002 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
4003 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
4004 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
4005 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
4006 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
4007 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
4008 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
4009 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
4010 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
4011 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
4012 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
4013 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
4014 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
4015 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
4016 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
4017 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
4018 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
4019 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
4020 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
4021 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
4022 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
4023 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
4024 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
4025 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
4026 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
4027 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
4028 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
4029 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
4030 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
4031 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
4033 (VNARROW, Ventype): New mode attributes.
4034 (Vewtype): Handle VNx2DI. Fix typo in comment.
4035 (VDOUBLE): New mode attribute.
4036 (sve_lane_con): Handle VNx8HI.
4037 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
4038 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
4039 (sve_int_op, sve_int_op_rev): Handle the above codes.
4040 (sve_pred_int_rhs2_operand): Likewise.
4041 (MULLBT, SHRNB, SHRNT): Delete.
4042 (SVE_INT_SHIFT_IMM): New int iterator.
4043 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
4044 and UNSPEC_WHILEHS for TARGET_SVE2.
4045 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
4046 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
4047 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
4048 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
4049 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
4050 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
4051 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
4052 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
4053 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
4054 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
4055 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
4056 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
4057 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
4058 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
4059 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
4060 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
4061 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
4062 (optab): Handle the new unspecs.
4063 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
4065 (lr): Handle the new unspecs.
4067 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
4068 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
4069 (sve_int_qsub_op): New int attributes.
4070 (sve_fp_op, rot): Handle the new unspecs.
4071 * config/aarch64/aarch64-sve-builtins.h
4072 (function_resolver::require_matching_pointer_type): Declare.
4073 (function_resolver::resolve_unary): Add an optional boolean argument.
4074 (function_resolver::finish_opt_n_resolution): Add an optional
4075 type_suffix_index argument.
4076 (gimple_folder::redirect_call): Declare.
4077 (gimple_expander::prepare_gather_address_operands): Add an optional
4079 * config/aarch64/aarch64-sve-builtins.cc: Include
4080 aarch64-sve-builtins-sve2.h.
4081 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
4082 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
4083 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
4084 (TYPES_hsd_integer): Use TYPES_hsd_signed.
4085 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
4086 (TYPES_s_unsigned): Likewise.
4087 (TYPES_s_integer): Use TYPES_s_unsigned.
4088 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
4089 (TYPES_sd_integer): Use them.
4090 (TYPES_d_unsigned): New macro.
4091 (TYPES_d_integer): Use it.
4092 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
4093 (TYPES_cvt_narrow): Likewise.
4094 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
4095 (preds_mx): New variable.
4096 (function_builder::add_overloaded_function): Allow the new feature
4097 set to be more restrictive than the original one.
4098 (function_resolver::infer_pointer_type): Remove qualifiers from
4099 the pointer type before printing it.
4100 (function_resolver::require_matching_pointer_type): New function.
4101 (function_resolver::resolve_sv_displacement): Handle functions
4102 that don't support 32-bit vector indices or svint32_t vector offsets.
4103 (function_resolver::finish_opt_n_resolution): Take the inferred type
4104 as a separate argument.
4105 (function_resolver::resolve_unary): Optionally treat all forms in
4106 the same way as normal merging functions.
4107 (gimple_folder::redirect_call): New function.
4108 (function_expander::prepare_gather_address_operands): Add an argument
4109 that says whether scaled forms are available. If they aren't,
4110 handle scaling of vector indices and don't add the extension and
4112 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
4113 fall back to using cond_* instead.
4114 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
4115 Split out the member variables into...
4116 (rtx_code_function_base): ...this new base class.
4117 (rtx_code_function_rotated): Inherit rtx_code_function_base.
4118 (unspec_based_function): Split out the member variables into...
4119 (unspec_based_function_base): ...this new base class.
4120 (unspec_based_function_rotated): Inherit unspec_based_function_base.
4121 (unspec_based_function_exact_insn): New class.
4122 (unspec_based_add_function, unspec_based_add_lane_function)
4123 (unspec_based_lane_function, unspec_based_pred_function)
4124 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
4125 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
4126 (unspec_based_sub_function, unspec_based_sub_lane_function): New
4128 (unspec_based_fused_function): New class.
4129 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
4130 (unspec_based_fused_lane_function): New class.
4131 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
4133 (CODE_FOR_MODE1): New macro.
4134 (fixed_insn_function): New class.
4135 (while_comparison): Likewise.
4136 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
4137 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
4138 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
4139 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
4140 (load_gather_sv_restricted, shift_left_imm_long): Declare.
4141 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
4142 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
4143 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
4144 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
4145 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
4146 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
4147 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
4148 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
4149 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
4150 Also add an initial argument for unary_convert_narrowt, regardless
4151 of the predication type.
4152 (build_32_64): Allow loads and stores to specify MODE_none.
4153 (build_sv_index64, build_sv_uint_offset): New functions.
4154 (long_type_suffix): New function.
4155 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
4156 (binary_imm_long_base, load_gather_sv_base): Likewise.
4157 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
4158 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
4159 (unary_narrowb_base, unary_narrowt_base): Likewise.
4160 (binary_long_lane_def, binary_long_lane): New shape.
4161 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
4162 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
4163 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
4164 (binary_to_uint_def, binary_to_uint): Likewise.
4165 (binary_wide_def, binary_wide): Likewise.
4166 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
4167 (compare_def, compare): Likewise.
4168 (compare_ptr_def, compare_ptr): Likewise.
4169 (load_ext_gather_index_restricted_def,
4170 load_ext_gather_index_restricted): Likewise.
4171 (load_ext_gather_offset_restricted_def,
4172 load_ext_gather_offset_restricted): Likewise.
4173 (load_gather_sv_def): Inherit from load_gather_sv_base.
4174 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
4175 (shift_left_imm_def, shift_left_imm): Likewise.
4176 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
4177 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
4178 (store_scatter_index_restricted_def,
4179 store_scatter_index_restricted): Likewise.
4180 (store_scatter_offset_restricted_def,
4181 store_scatter_offset_restricted): Likewise.
4182 (tbl_tuple_def, tbl_tuple): Likewise.
4183 (ternary_long_lane_def, ternary_long_lane): Likewise.
4184 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4185 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4186 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4187 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4188 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4189 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4190 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4191 (ternary_uint_def, ternary_uint): Likewise.
4192 (unary_convert): Fix typo in comment.
4193 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4194 (unary_long_def, unary_long): Likewise.
4195 (unary_narrowb_def, unary_narrowb): Likewise.
4196 (unary_narrowt_def, unary_narrowt): Likewise.
4197 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4198 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4199 (unary_to_int_def, unary_to_int): Likewise.
4200 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4201 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4202 (svasrd_impl): Delete.
4203 (svcadd_impl::expand): Handle integer operations too.
4204 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4205 new functions to derive the unspec numbers.
4206 (svmla_svmls_lane_impl): Replace with...
4207 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4208 integer operations too.
4209 (svwhile_impl): Rename to...
4210 (svwhilelx_impl): ...this and inherit from while_comparison.
4211 (svasrd): Use unspec_based_function.
4212 (svmla_lane): Use svmla_lane_impl.
4213 (svmls_lane): Use svmls_lane_impl.
4214 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4215 (svwhilele, svwhilelt): Use svwhilelx_impl.
4216 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4217 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4218 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4219 * config/aarch64/aarch64-sve-builtins.def: Include
4220 aarch64-sve-builtins-sve2.def.
4222 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4224 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4225 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4226 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4227 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4228 immediates as well as vector ones.
4229 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4230 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4231 (aarch64_sve_qsub_immediate): Update calls accordingly.
4233 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4235 * config/aarch64/aarch64-sve2.md: Add banner comments.
4236 (<su>mulh<r>s<mode>3): Move further up file.
4237 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4238 (*aarch64_sve2_sra<mode>): Move further down file.
4239 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4241 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4243 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4245 (while_optab_cmp): Handle them.
4246 * config/aarch64/aarch64-sve.md
4247 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4248 and add a "@" marker.
4249 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4250 instead of gen_aarch64_sve2_while_ptest.
4251 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4253 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4255 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4256 (UNSPEC_WHILELE): ...this.
4257 (UNSPEC_WHILE_LO): Rename to...
4258 (UNSPEC_WHILELO): ...this.
4259 (UNSPEC_WHILE_LS): Rename to...
4260 (UNSPEC_WHILELS): ...this.
4261 (UNSPEC_WHILE_LT): Rename to...
4262 (UNSPEC_WHILELT): ...this.
4263 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4264 (cmp_op, while_optab_cmp): Likewise.
4265 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4266 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4267 (svwhilelt): Likewise.
4269 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4271 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4272 (unary_to_uint): Define.
4273 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4274 (unary_count): Rename to...
4275 (unary_to_uint_def, unary_to_uint): ...this.
4276 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4278 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4280 * config/aarch64/aarch64-sve-builtins-functions.h
4281 (code_for_mode_function): New class.
4282 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4283 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4284 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4285 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4286 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4288 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4290 * config/aarch64/iterators.md (addsub): New code attribute.
4291 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4293 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4294 in the asm string and attributes. Fix indentation.
4295 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4297 (@aarch64_sve_<optab><mode>): ...this.
4298 * config/aarch64/aarch64-sve-builtins.h
4299 (function_expander::expand_signed_unpred_op): Delete.
4300 * config/aarch64/aarch64-sve-builtins.cc
4301 (function_expander::expand_signed_unpred_op): Likewise.
4302 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4303 try using code_for_aarch64_sve instead.
4304 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4305 (svqsub_impl): Likewise.
4306 (svqadd, svqsub): Use rtx_code_function instead.
4308 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4310 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4311 (HADDSUB, sur, addsub): Remove them.
4313 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4315 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4316 functions the same way, rather than singling out those that
4317 aren't mapped directly to optabs.
4319 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4321 * target.def (compatible_vector_types_p): New target hook.
4322 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4323 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4324 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4325 * doc/tm.texi: Regenerate.
4326 * gimple-expr.c: Include target.h.
4327 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4328 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4330 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4331 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4332 Use the original predicate if it already has a suitable type.
4334 2020-01-09 Martin Jambor <mjambor@suse.cz>
4336 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4337 resolve_speculation and redirect_call_stmt_to_callee static. Change
4338 return type of set_call_stmt to cgraph_edge *.
4339 * auto-profile.c (afdo_indirect_call): Adjust call to
4340 redirect_call_stmt_to_callee.
4341 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4342 make the this pointer explicit, adjust self-recursive calls and the
4343 call top make_direct. Return the resulting edge.
4344 (cgraph_edge::remove): Make this pointer explicit.
4345 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4346 (cgraph_edge::make_direct): Likewise, adjust call to
4347 resolve_speculation.
4348 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4349 call to set_call_stmt.
4350 (cgraph_update_edges_for_call_stmt_node): Update call to
4351 set_call_stmt and remove.
4352 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4353 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4354 (cgraph_node::create_edge_including_clones): Moved "first" definition
4355 of edge to the block where it was used. Adjusted calls to
4357 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4358 cgraph_edge::remove.
4359 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4360 make_direct and redirect_call_stmt_to_callee.
4361 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4362 resolve_speculation and make_direct.
4363 * ipa-inline-transform.c (inline_transform): Adjust call to
4364 redirect_call_stmt_to_callee.
4365 (check_speculations_1):: Adjust call to resolve_speculation.
4366 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4367 resolve-speculation.
4368 (inline_small_functions): Adjust call to resolve_speculation.
4369 (ipa_inline): Likewise.
4370 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4372 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4373 safe with regards to edge removal, adjust calls to
4374 redirect_call_stmt_to_callee.
4375 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4376 and redirect_call_stmt_to_callee.
4377 * multiple_target.c (create_dispatcher_calls): Adjust call to
4378 redirect_call_stmt_to_callee
4379 (redirect_to_specific_clone): Likewise.
4380 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4381 Adjust calls to cgraph_edge::remove.
4382 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4383 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4384 (expand_call_inline): Adjust call to cgraph_edge::remove.
4386 2020-01-09 Martin Liska <mliska@suse.cz>
4388 * params.opt: Set Optimization for
4389 param_max_speculative_devirt_maydefs.
4391 2020-01-09 Martin Sebor <msebor@redhat.com>
4395 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4397 2020-01-09 Martin Liska <mliska@suse.cz>
4399 * auto-profile.c (auto_profile): Use opt_for_fn
4401 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4402 (propagate_vals_across_arith_jfunc): Likewise.
4403 (hint_time_bonus): Likewise.
4404 (incorporate_penalties): Likewise.
4405 (good_cloning_opportunity_p): Likewise.
4406 (perform_estimation_of_a_value): Likewise.
4407 (estimate_local_effects): Likewise.
4408 (ipcp_propagate_stage): Likewise.
4409 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4410 (set_switch_stmt_execution_predicate): Likewise.
4411 (analyze_function_body): Likewise.
4412 * ipa-inline-analysis.c (offline_size): Likewise.
4413 * ipa-inline.c (early_inliner): Likewise.
4414 * ipa-prop.c (ipa_analyze_node): Likewise.
4415 (ipcp_transform_function): Likewise.
4416 * ipa-sra.c (process_scan_results): Likewise.
4417 (ipa_sra_summarize_function): Likewise.
4418 * params.opt: Rename ipcp-unit-growth to
4419 ipa-cp-unit-growth. Add Optimization for various
4420 IPA-related parameters.
4422 2020-01-09 Richard Biener <rguenther@suse.de>
4425 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4427 2020-01-09 Richard Biener <rguenther@suse.de>
4429 PR tree-optimization/93040
4430 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4432 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4434 * common/config/avr/avr-common.c (avr_option_optimization_table)
4435 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4437 2020-01-09 Martin Liska <mliska@suse.cz>
4439 * cgraphclones.c (symbol_table::materialize_all_clones):
4440 Use cgraph_node::dump_name.
4442 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4445 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4446 output_operand_lossage instead of gcc_unreachable.
4447 * doc/md.texi (riscv f constraint): Fix typo.
4450 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4451 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4452 CONST_SCALAR_INT_P instead of CONST_INT_P.
4453 (*subv<mode>4_1): Rename to ...
4454 (subv<mode>4_1): ... this.
4455 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4456 define_insn_and_split patterns.
4457 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4460 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4462 * vec.c (class selftest::count_dtor): New class.
4463 (selftest::test_auto_delete_vec): New test.
4464 (selftest::vec_c_tests): Call it.
4465 * vec.h (class auto_delete_vec): New class template.
4466 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4468 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4470 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4472 2020-01-08 Jim Wilson <jimw@sifive.com>
4474 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4475 use of TLS_MODEL_LOCAL_EXEC when not pic.
4477 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4479 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4482 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4485 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4486 *stack_protect_set_3 peephole2): Also check that the second
4487 insns source is general_operand.
4490 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4491 predicate for output operand instead of register_operand.
4492 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4493 memory destination and non-memory operands[2].
4495 2020-01-08 Martin Liska <mliska@suse.cz>
4497 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4498 ::dump_asm_name instead of (::name or ::asm_name).
4499 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4500 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4501 (analyze_functions): Likewise.
4502 (expand_all_functions): Likewise.
4503 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4504 (propagate_bits_across_jump_function): Likewise.
4505 (dump_profile_updates): Likewise.
4506 (ipcp_store_bits_results): Likewise.
4507 (ipcp_store_vr_results): Likewise.
4508 * ipa-devirt.c (dump_targets): Likewise.
4509 * ipa-fnsummary.c (analyze_function_body): Likewise.
4510 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4511 (process_hsa_functions): Likewise.
4512 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4513 (set_alias_uids): Likewise.
4514 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4515 * ipa-inline.c (recursive_inlining): Likewise.
4516 (inline_to_all_callers_1): Likewise.
4517 (ipa_inline): Likewise.
4518 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4519 (ipa_propagate_frequency): Likewise.
4520 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4521 (remove_described_reference): Likewise.
4522 * ipa-pure-const.c (worse_state): Likewise.
4523 (check_retval_uses): Likewise.
4524 (analyze_function): Likewise.
4525 (propagate_pure_const): Likewise.
4526 (propagate_nothrow): Likewise.
4527 (dump_malloc_lattice): Likewise.
4528 (propagate_malloc): Likewise.
4529 (pass_local_pure_const::execute): Likewise.
4530 * ipa-visibility.c (optimize_weakref): Likewise.
4531 (function_and_variable_visibility): Likewise.
4532 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4533 (ipa_discover_variable_flags): Likewise.
4534 * lto-streamer-out.c (output_function): Likewise.
4535 (output_constructor): Likewise.
4536 * tree-inline.c (copy_bb): Likewise.
4537 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4538 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4540 2020-01-08 Richard Biener <rguenther@suse.de>
4543 * tree-eh.c (sink_clobbers): Update virtual operands for
4544 the first and last stmt only. Add a dry-run capability.
4545 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4546 after CFG manipulations and in RPO order to catch all
4547 secondary opportunities reliably.
4549 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4552 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4554 2019-01-08 Richard Biener <rguenther@suse.de>
4557 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4558 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4559 virtual operand, also updating SSA use.
4560 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4561 Update stmt after resetting virtual operand.
4562 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4563 * gimple-iterator.c (gsi_remove): When not removing the stmt
4564 permanently do not delink immediate uses or mark the stmt modified.
4566 2020-01-08 Martin Liska <mliska@suse.cz>
4568 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4569 (ipa_call_context::estimate_size_and_time): Likewise.
4570 (inline_analyze_function): Likewise.
4572 2020-01-08 Martin Liska <mliska@suse.cz>
4574 * cgraph.c (cgraph_node::dump): Use systematically
4577 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4579 Add -nodevicespecs option for avr.
4582 * config/avr/avr.opt (-nodevicespecs): New driver option.
4583 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4584 "-specs=device-specs/..." if that option is not set.
4585 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4587 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4589 Implement 64-bit double functions for avr.
4592 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4593 --with-double-comparison.
4594 * doc/install.texi: Document them.
4595 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4596 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4597 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4598 * doc/invoke.texi (AVR Built-in Macros): Document them.
4599 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4600 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4601 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4603 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4606 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4607 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4608 when only building rm-profile multilibs.
4610 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4613 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4614 lattice for a value to check.
4615 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4616 finite propagation in self-recursive scc.
4618 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4620 * ipa-inline.c (caller_growth_limits): Restore the AND.
4622 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4624 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4625 (VEC_ALLREG_ALT): New iterator.
4626 (VEC_ALLREG_INT_MODE): New iterator.
4627 (VCMP_MODE): New iterator.
4628 (VCMP_MODE_INT): New iterator.
4629 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4630 (vec_cmp<u>v64qidi): New define_expand.
4631 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4632 (vec_cmpu<mode>di_exec): New define_expand.
4633 (vec_cmp<u>v64qidi_exec): New define_expand.
4634 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4635 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4636 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4637 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4638 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4639 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4640 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4641 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4642 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4643 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4645 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4646 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4648 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4650 * config/gcn/constraints.md (DA): Update description and match.
4652 (Db): New constraint.
4653 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4655 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4656 Implement 'Db' mixed immediate type.
4657 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4658 (addcv64si3_dup<exec_vcc>): Delete.
4659 (subcv64si3<exec_vcc>): Rework constraints.
4660 (addv64di3): Rework constraints.
4661 (addv64di3_exec): Rework constraints.
4662 (subv64di3): Rework constraints.
4663 (addv64di3_dup): Delete.
4664 (addv64di3_dup_exec): Delete.
4665 (addv64di3_zext): Rework constraints.
4666 (addv64di3_zext_exec): Rework constraints.
4667 (addv64di3_zext_dup): Rework constraints.
4668 (addv64di3_zext_dup_exec): Rework constraints.
4669 (addv64di3_zext_dup2): Rework constraints.
4670 (addv64di3_zext_dup2_exec): Rework constraints.
4671 (addv64di3_sext_dup2): Rework constraints.
4672 (addv64di3_sext_dup2_exec): Rework constraints.
4674 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4676 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4677 existing target checks.
4679 2020-01-07 Richard Biener <rguenther@suse.de>
4681 * doc/install.texi: Bump minimal supported MPC version.
4683 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4685 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4686 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4687 * langhooks.c: Include stor-layout.h.
4688 (lhd_simulate_enum_decl): New function.
4689 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4690 handle_arm_sve_h for the LTO frontend.
4691 (register_vector_type): Cope with null returns from pushdecl.
4693 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4695 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4696 (aarch64_sve::nvectors_if_data_type): Replace with...
4697 (aarch64_sve::builtin_type_p): ...this.
4698 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4699 (find_vector_type): Delete.
4700 (add_sve_type_attribute): New function.
4701 (lookup_sve_type_attribute): Likewise.
4702 (register_builtin_types): Add an "SVE type" attribute to each type.
4703 (register_tuple_type): Likewise.
4704 (svbool_type_p, nvectors_if_data_type): Delete.
4705 (mangle_builtin_type): Use lookup_sve_type_attribute.
4706 (builtin_type_p): Likewise. Add an overload that returns the
4707 number of constituent vector and predicate registers.
4708 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4709 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4710 instead of aarch64_sve_argument_p.
4711 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4712 (aarch64_pass_by_reference): Likewise.
4713 (aarch64_function_value_1): Likewise.
4714 (aarch64_return_in_memory): Likewise.
4715 (aarch64_layout_arg): Likewise.
4717 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4719 PR tree-optimization/93156
4720 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4721 least significant bit is always clear.
4723 PR tree-optimization/93118
4724 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4725 simplifier with two intermediate conversions.
4727 2020-01-07 Martin Liska <mliska@suse.cz>
4729 * params.opt: Add Optimization for various parameters.
4731 2020-01-07 Martin Liska <mliska@suse.cz>
4734 * doc/extend.texi: Explain cloning for target_clone
4737 2020-01-07 Martin Liska <mliska@suse.cz>
4739 PR tree-optimization/92860
4740 * common.opt: Make in Optimization option
4741 as it is affected by -O0, which is an Optimization
4743 * tree-inline.c (tree_inlinable_function_p):
4744 Use opt_for_fn for warn_inline.
4745 (expand_call_inline): Likewise.
4747 2020-01-07 Martin Liska <mliska@suse.cz>
4749 PR tree-optimization/92860
4750 * common.opt: Make flag_ree as optimization
4753 2020-01-07 Martin Liska <mliska@suse.cz>
4755 PR optimization/92860
4756 * params.opt: Mark param_min_crossjump_insns with Optimization
4759 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4761 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4762 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4764 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4766 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4767 helper function to return the valid addressing formats for a given
4768 hard register and mode.
4769 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4771 * config/rs6000/constraints.md (Q constraint): Update
4773 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4776 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4777 Use 'Q' for doing vector extract from memory.
4778 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4780 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4781 doing vector extract from memory.
4782 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4783 extract from memory.
4785 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4786 for the offset being 34-bits when -mcpu=future is used.
4788 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4790 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4791 instead of cmpib_comparison_operator in cmpib patterns.
4792 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4793 of cmpib_comparison_operator. Revise comment.
4795 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4797 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4798 in an IFN_DIV_POW2 node to be equal.
4800 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4802 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4803 (vect_check_scalar_mask): ...this.
4804 (vectorizable_store, vectorizable_load): Update call accordingly.
4805 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4806 argument in calls to conditional internal functions.
4808 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4810 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4811 '0' matching inputs.
4812 (subv64di3_exec): Likewise.
4814 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4816 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4817 * doc/md.texi (movstr): Likewise.
4819 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4821 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4824 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4826 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4828 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4829 to a temporary file and use move-if-change to update the real
4830 file where necessary.
4832 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4834 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4835 rather than Upa for CPY /M.
4837 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4839 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4842 2020-01-06 Martin Liska <mliska@suse.cz>
4844 PR tree-optimization/92860
4845 * params.opt: Mark param_max_combine_insns with Optimization
4848 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4851 * config/i386/i386.md (SWIDWI): New mode iterator.
4852 (DWI, dwi): Add TImode variants.
4853 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4854 <general_hilo_operand> instead of <general_operand>. Use
4855 CONST_SCALAR_INT_P instead of CONST_INT_P.
4856 (*addv<mode>4_1): Rename to ...
4857 (addv<mode>4_1): ... this.
4858 (QWI): New mode attribute.
4859 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4860 define_insn_and_split patterns.
4861 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4863 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4864 <general_hilo_operand> instead of <general_operand>.
4865 (*addcarry<mode>_1): New define_insn.
4866 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4868 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4870 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4871 Use "call" instead of "set".
4873 2020-01-03 Martin Jambor <mjambor@suse.cz>
4876 * ipa-cp.c (print_all_lattices): Skip functions without info.
4878 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4881 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4882 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4883 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4884 for 'e' simd clones.
4887 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4889 (mprefer-vector-width=): Add Save.
4890 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4891 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4892 (ix86_debug_options, ix86_function_specific_print): Adjust
4893 ix86_target_string callers.
4894 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4895 (ix86_valid_target_attribute_tree): Likewise.
4896 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4897 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4898 ix86_target_string caller.
4901 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4902 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4903 instead of gen_int_shift_amount + convert_modes.
4905 PR rtl-optimization/93088
4906 * loop-iv.c (find_single_def_src): Punt after looking through
4907 128 reg copies for regs with single definitions. Move definitions
4910 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4912 * config/arm/arm-c.c (arm_cpu_builtins): Define
4913 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4914 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4915 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4916 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4917 * config/arm/arm-tables.opt: Regenerated.
4918 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4919 arm_arch_i8mm and arm_arch_bf16 when enabled.
4920 * config/arm/arm.h (TARGET_I8MM): New macro.
4921 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4922 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4923 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4924 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4925 (v8_6_a_simd_variants): New.
4926 (v8_*_a_simd_variants): Add i8mm and bf16.
4927 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4929 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4932 * predict.c (compute_function_frequency): Don't call
4933 warn_function_cold on functions that already have cold attribute.
4935 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4938 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4939 COMDAT group function labels in .data.rel.ro.local section.
4940 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4943 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4944 comparison_operator in B and S integer comparisons. Likewise, use
4945 ordered_comparison_operator instead of cmpib_comparison_operator in
4947 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4949 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4951 Update copyright years.
4953 * gcc.c (process_command): Update copyright notice dates.
4954 * gcov-dump.c (print_version): Ditto.
4955 * gcov.c (print_version): Ditto.
4956 * gcov-tool.c (print_version): Ditto.
4957 * gengtype.c (create_file): Ditto.
4958 * doc/cpp.texi: Bump @copying's copyright year.
4959 * doc/cppinternals.texi: Ditto.
4960 * doc/gcc.texi: Ditto.
4961 * doc/gccint.texi: Ditto.
4962 * doc/gcov.texi: Ditto.
4963 * doc/install.texi: Ditto.
4964 * doc/invoke.texi: Ditto.
4966 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4968 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4971 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4973 PR tree-optimization/93098
4974 * match.pd (popcount): For shift amounts, use integer_onep
4975 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4976 tests. Make sure that precision is power of two larger than or equal
4977 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4978 instead of ULL suffixed constants. Formatting fixes.
4980 Copyright (C) 2020 Free Software Foundation, Inc.
4982 Copying and distribution of this file, with or without modification,
4983 are permitted in any medium without royalty provided the copyright
4984 notice and this notice are preserved.