1 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
3 * config/gcn/gcn-valu.md (VEC_SUBDWORD_MODE): New mode iterator.
4 (<expander><mode>2<exec>): Change modes to VEC_ALL1REG_INT_MODE.
5 (<expander><mode>3<exec>): Likewise.
6 (<expander><mode>3): New.
7 (v<expander><mode>3): New.
8 (<expander><mode>3): New.
9 (<expander><mode>3<exec>): Rename to ...
10 (<expander>v64si3<exec>): ... this, and change modes to V64SI.
11 * config/gcn/gcn.md (mnemonic): Use '%B' for not.
13 2020-02-27 Alexandre Oliva <oliva@adacore.com>
15 * config/vx-common.h (NO_DOLLAR_IN_LABEL, NO_DOT_IN_LABEL): Leave
18 2020-02-27 Richard Biener <rguenther@suse.de>
20 PR tree-optimization/93508
21 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle _CHK like
22 non-_CHK variants. Valueize their length arguments.
24 2020-02-27 Richard Biener <rguenther@suse.de>
26 PR tree-optimization/93953
27 * tree-vect-slp.c (slp_copy_subtree): Avoid keeping a reference
28 to the hash-map entry.
30 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
32 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
34 2020-02-27 Mark Williams <mwilliams@fb.com>
36 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
37 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
38 -ffile-prefix-map and -fmacro-prefix-map.
39 * lto-streamer-out.c: Include file-prefix-map.h.
40 (lto_output_location): Remap the file part of locations.
42 2020-02-27 Jakub Jelinek <jakub@redhat.com>
45 * gimplify.c (gimplify_init_constructor): Don't promote readonly
46 DECL_REGISTER variables to TREE_STATIC.
48 PR tree-optimization/93582
49 PR tree-optimization/93945
50 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
51 non-zero INTEGER_CST second argument and ref->offset or ref->size
52 not a multiple of BITS_PER_UNIT.
54 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
56 * doc/install.texi (Binaries): Update description of BullFreeware.
58 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
62 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
63 C++ Language Options, Warning Options, and Static Analyzer
64 Options lists. Document negative form of options enabled by
65 default. Move some things around to more accurately sort
67 (C++ Dialect Options, Warning Options, Static Analyzer
68 Options): Document negative form of options when enabled by
69 default. Move some things around to more accurately sort
70 warnings by category. Add some missing index entries.
73 2020-02-26 Carl Love <cel@us.ibm.com>
76 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
77 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
78 for the vector unsigned short arguments. It is also listed as the
79 name of the built-in for arguments vector unsigned short,
80 vector unsigned int and vector unsigned long long built-ins. The
81 name of the builtins for these arguments should be:
82 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
83 __builtin_crypto_vpmsumd respectively.
85 2020-02-26 Richard Biener <rguenther@suse.de>
87 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
90 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
93 * optabs-tree.c (supportable_convert_operation): Reject types with
96 2020-02-26 David Malcolm <dmalcolm@redhat.com>
98 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
100 2020-02-26 Jakub Jelinek <jakub@redhat.com>
102 PR tree-optimization/93820
103 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
104 argument to ALL_INTEGER_CST_P boolean.
105 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
106 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
107 adjacent INTEGER_CST store into merged_store->only_constants like
110 2020-02-25 Jakub Jelinek <jakub@redhat.com>
113 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
115 * cfghooks.c (verify_flow_info): Likewise.
116 * predict.c (combine_predictions_for_bb): Likewise.
117 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
118 sucessor -> successor.
119 (find_traces_1_round): Fix comment typo, destinarion -> destination.
120 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
122 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
123 message typo, sucessors -> successors.
125 2020-02-25 Martin Sebor <msebor@redhat.com>
127 * doc/extend.texi (attribute access): Correct an example.
129 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
131 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
133 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
135 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
136 (VD): Enable for V4BF.
138 (VQ): Enable for V8BF.
141 (VDBL, Vdbl): Add V4BF.
142 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
143 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
144 (bfloat16x8x2_t): Likewise.
145 (bfloat16x4x3_t): Likewise.
146 (bfloat16x8x3_t): Likewise.
147 (bfloat16x4x4_t): Likewise.
148 (bfloat16x8x4_t): Likewise.
149 (vcombine_bf16): New.
150 (vld1_bf16, vld1_bf16_x2): New.
151 (vld1_bf16_x3, vld1_bf16_x4): New.
152 (vld1q_bf16, vld1q_bf16_x2): New.
153 (vld1q_bf16_x3, vld1q_bf16_x4): New.
154 (vld1_lane_bf16): New.
155 (vld1q_lane_bf16): New.
156 (vld1_dup_bf16): New.
157 (vld1q_dup_bf16): New.
160 (vld2_dup_bf16): New.
161 (vld2q_dup_bf16): New.
164 (vld3_dup_bf16): New.
165 (vld3q_dup_bf16): New.
168 (vld4_dup_bf16): New.
169 (vld4q_dup_bf16): New.
170 (vst1_bf16, vst1_bf16_x2): New.
171 (vst1_bf16_x3, vst1_bf16_x4): New.
172 (vst1q_bf16, vst1q_bf16_x2): New.
173 (vst1q_bf16_x3, vst1q_bf16_x4): New.
174 (vst1_lane_bf16): New.
175 (vst1q_lane_bf16): New.
183 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
185 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
186 (VALL_F16): Likewise.
187 (VALLDI_F16): Likewise.
190 (vswap_width_name): Likewise.
191 (VSWAP_WIDTH): Likewise.
195 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
196 (vget_lane_bf16, vgetq_lane_bf16): New.
198 (vdup_n_bf16, vdupq_n_bf16): New.
199 (vdup_lane_bf16, vdup_laneq_bf16): New.
200 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
201 (vduph_lane_bf16, vduph_laneq_bf16): New.
202 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
203 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
204 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
205 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
206 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
207 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
208 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
209 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
210 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
211 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
212 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
213 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
214 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
215 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
216 (vreinterpretq_bf16_p128): New.
217 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
218 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
219 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
220 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
221 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
222 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
223 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
224 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
225 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
226 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
227 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
228 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
229 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
230 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
231 (vreinterpretq_p128_bf16): New.
233 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
235 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
236 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
237 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
238 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
239 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
240 * config/arm/iterators.md (VSF2BF): New attribute.
241 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
242 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
243 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
245 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
247 * config/arm/arm.md (required_for_purecode): New attribute.
248 (enabled): Handle required_for_purecode.
249 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
250 work with -mpure-code.
252 2020-02-25 Jakub Jelinek <jakub@redhat.com>
254 PR rtl-optimization/93908
255 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
258 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
260 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
262 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
264 * doc/install.texi (--enable-checking): Adjust wording.
266 2020-02-25 Richard Biener <rguenther@suse.de>
268 PR tree-optimization/93868
269 * tree-vect-slp.c (slp_copy_subtree): New function.
270 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
271 re-arranging stmts in it.
273 2020-02-25 Jakub Jelinek <jakub@redhat.com>
276 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
277 dummy function and remove it at the end.
280 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
281 paramter -> parameter.
282 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
283 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
285 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
287 * doc/install.texi (--enable-checking): Properly document current
289 (--enable-stage1-checking): Minor clarification about bootstrap.
291 2020-02-24 David Malcolm <dmalcolm@redhat.com>
294 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
295 -fanalyzer-checker=taint is also required.
296 (-fanalyzer-checker=): Note that providing this option enables the
297 given checker, and doing so may be required for checkers that are
300 2020-02-24 David Malcolm <dmalcolm@redhat.com>
302 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
303 significant control flow events; add a "3" which shows all
304 control flow events; the old "3" becomes "4".
306 2020-02-24 Jakub Jelinek <jakub@redhat.com>
308 PR tree-optimization/93582
309 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
310 pd.offset and pd.size to be counted in bits rather than bytes, add
311 support for maxsizei that is not a multiple of BITS_PER_UNIT and
312 handle bitfield stores and loads.
313 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
314 uncomparable quantities - bytes vs. bits. Allow push_partial_def
315 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
316 pd.offset/pd.size to be counted in bits rather than bytes.
317 Formatting fix. Rename shadowed len variable to buflen.
319 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
320 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
323 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
324 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
325 * opts-common.c (parse_options_from_collect_gcc_options): New function.
326 (prepend_xassembler_to_collect_as_options): Likewise.
327 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
328 (prepend_xassembler_to_collect_as_options): Likewise.
329 * lto-opts.c (lto_write_options): Stream assembler options
330 in COLLECT_AS_OPTIONS.
331 * lto-wrapper.c (xassembler_options_error): New static variable.
332 (get_options_from_collect_gcc_options): Move parsing options code to
333 parse_options_from_collect_gcc_options and call it.
334 (merge_and_complain): Validate -Xassembler options.
335 (append_compiler_options): Handle OPT_Xassembler.
336 (run_gcc): Append command line -Xassembler options to
338 * doc/invoke.texi: Add documentation about using Xassembler
341 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
343 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
345 (riscv_rtx_costs): Update cost model for LTGT.
347 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
349 PR rtl-optimization/93564
350 * ira-color.c (struct update_cost_queue_elem): New member start.
351 (queue_update_cost, get_next_update_cost): Add new arg start.
352 (allocnos_conflict_p): New function.
353 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
354 Add checking conflicts with allocnos_conflict_p.
355 (update_costs_from_prefs, restore_costs_from_copies): Adjust
356 update_costs_from_allocno calls.
357 (update_conflict_hard_regno_costs): Add checking conflicts with
358 allocnos_conflict_p. Adjust calls of queue_update_cost and
359 get_next_update_cost.
360 (assign_hard_reg): Adjust calls of queue_update_cost. Add
362 (bucket_allocno_compare_func): Restore previous version.
364 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
366 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
367 double-word size when handling aggregate return values.
368 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
369 that homogeneous SFmode and DFmode aggregates are passed and returned
370 in general registers.
372 2020-02-21 Jakub Jelinek <jakub@redhat.com>
375 * opts.c (print_filtered_help): Translate help before appending
376 messages to it rather than after that.
378 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
380 PR rtl-optimization/PR92989
381 * lra-lives.c (process_bb_lives): Restore the original order
382 of the bb liveness update. Call make_hard_regno_dead for each
383 register clobbered at the start of an EH receiver.
385 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
388 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
389 self-recursively generated.
391 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
394 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
397 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
399 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
400 Document new target supports option.
402 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
404 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
405 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
406 * config/arm/iterators.md (MATMUL): New iterator.
407 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
408 (mmla_sfx): New attribute.
409 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
410 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
411 (UNSPEC_MATMUL_US): New.
413 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
415 * config/arm/arm.md: Prevent scalar shifts from being used when big
418 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
419 Richard Biener <rguenther@suse.de>
421 PR tree-optimization/93586
422 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
423 after mismatched array refs; do not sure type size information to
424 recover from unmatched referneces with !flag_strict_aliasing_p.
426 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
428 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
429 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
430 (scatter_store<mode>): Rename to ...
431 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
432 (scatter<mode>_exec): Delete. Move contents ...
433 (mask_scatter_store<mode>): ... here, and rename that to ...
434 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
435 Remove mode conversion.
436 (mask_gather_load<mode>): Rename to ...
437 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
438 Remove mode conversion.
439 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
441 2020-02-21 Martin Jambor <mjambor@suse.cz>
443 PR tree-optimization/93845
444 * tree-sra.c (verify_sra_access_forest): Only test access size of
447 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
449 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
450 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
451 (addv64di3_exec): Likewise.
452 (subv64di3): Likewise.
453 (subv64di3_exec): Likewise.
454 (addv64di3_zext): Likewise.
455 (addv64di3_zext_exec): Likewise.
456 (addv64di3_zext_dup): Likewise.
457 (addv64di3_zext_dup_exec): Likewise.
458 (addv64di3_zext_dup2): Likewise.
459 (addv64di3_zext_dup2_exec): Likewise.
460 (addv64di3_sext_dup2): Likewise.
461 (addv64di3_sext_dup2_exec): Likewise.
462 (<expander>v64di3): Likewise.
463 (<expander>v64di3_exec): Likewise.
464 (*<reduc_op>_dpp_shr_v64di): Likewise.
465 (*plus_carry_dpp_shr_v64di): Likewise.
466 * config/gcn/gcn.md (adddi3): Likewise.
467 (addptrdi3): Likewise.
468 (<expander>di3): Likewise.
470 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
472 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
474 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
476 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
477 support. Use aarch64_emit_mult instead of emitting multiplication
478 instructions directly.
479 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
480 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
482 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
484 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
485 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
486 instead of emitting multiplication instructions directly.
487 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
488 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
489 (@aarch64_frecps<mode>): New expanders.
491 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
493 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
494 on and produce uint64_ts rather than ints.
495 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
496 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
498 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
500 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
501 an unused xmsk register when handling approximate rsqrt.
503 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
505 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
506 flag_finite_math_only condition.
508 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
511 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
512 to destination operand for shufps alternative.
513 (*vec_extractv2si_1): Ditto.
515 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
518 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
521 2020-02-20 Martin Liska <mliska@suse.cz>
524 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
526 2020-02-20 Martin Liska <mliska@suse.cz>
529 * common/config/avr/avr-common.c: Remote trailing "|".
531 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
533 * collect2.c (maybe_run_lto_and_relink): Fix typo in
536 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
538 PR tree-optimization/93767
539 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
540 access-size bias from the offset calculations for negative strides.
542 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
544 * collect2.c (c_file, o_file): Make const again.
545 (ldout,lderrout, dump_ld_file): Remove.
546 (tool_cleanup): Avoid calling not signal-safe functions.
547 (maybe_run_lto_and_relink): Avoid possible signal handler
548 access to unintialzed memory (lto_o_files).
549 (main): Avoid leaking temp files in $TMPDIR.
550 Initialize c_file/o_file with concat, which avoids exposing
551 uninitialized memory to signal handler, which calls unlink(!).
552 Avoid calling maybe_unlink when the main function returns,
553 since the atexit handler is already doing this.
554 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
556 2020-02-19 Martin Jambor <mjambor@suse.cz>
558 PR tree-optimization/93776
559 * tree-sra.c (create_access): Do not create zero size accesses.
560 (get_access_for_expr): Do not search for zero sized accesses.
562 2020-02-19 Martin Jambor <mjambor@suse.cz>
564 PR tree-optimization/93667
565 * tree-sra.c (scalarizable_type_p): Return false if record fields
566 do not follow wach other.
568 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
570 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
571 rather than fmv.x.s/fmv.s.x.
573 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
575 * config/aarch64/aarch64-simd-builtins.def
576 (intrinsic_vec_smult_lo_): New.
577 (intrinsic_vec_umult_lo_): Likewise.
578 (vec_widen_smult_hi_): Likewise.
579 (vec_widen_umult_hi_): Likewise.
580 * config/aarch64/aarch64-simd.md
581 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
582 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
583 (vmull_high_s16): Likewise.
584 (vmull_high_s32): Likewise.
585 (vmull_high_u8): Likewise.
586 (vmull_high_u16): Likewise.
587 (vmull_high_u32): Likewise.
588 (vmull_s8): Likewise.
589 (vmull_s16): Likewise.
590 (vmull_s32): Likewise.
591 (vmull_u8): Likewise.
592 (vmull_u16): Likewise.
593 (vmull_u32): Likewise.
595 2020-02-18 Martin Liska <mliska@suse.cz>
597 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
598 bootstrap by missing removal of invalid sanity check.
600 2020-02-18 Martin Liska <mliska@suse.cz>
603 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
604 Always compare LHS of gimple_assign.
606 2020-02-18 Martin Liska <mliska@suse.cz>
609 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
610 and return type of functions.
611 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
612 Drop MALLOC attribute for void functions.
613 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
614 malloc_state for a new VOID clone.
616 2020-02-18 Martin Liska <mliska@suse.cz>
619 * common.opt: Add -fprofile-reproducibility.
620 * doc/invoke.texi: Document it.
621 * value-prof.c (dump_histogram_value):
622 Document and support behavior for counters[0]
623 being a negative value.
624 (get_nth_most_common_value): Handle negative
625 counters[0] in respect to flag_profile_reproducible.
627 2020-02-18 Jakub Jelinek <jakub@redhat.com>
630 * cgraph.c (verify_speculative_call): Use speculative_id instead of
631 speculative_uid in messages. Remove trailing whitespace from error
632 message. Use num_speculative_call_targets instead of
633 num_speculative_targets in a message.
634 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
635 edge messages and stmt instead of cal_stmt in reference message.
637 PR tree-optimization/93780
638 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
639 before calling build_vector_type.
640 (execute_update_addresses_taken): Likewise.
643 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
644 typo, functoin -> function.
645 * tree.c (free_lang_data_in_decl): Fix comment typo,
646 functoin -> function.
647 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
649 2020-02-17 David Malcolm <dmalcolm@redhat.com>
651 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
653 (print_option_information): Don't call get_option_url if URLs
656 2020-02-17 Alexandre Oliva <oliva@adacore.com>
658 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
659 handling of register_common-less targets.
661 2020-02-17 Martin Liska <mliska@suse.cz>
664 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
666 2020-02-17 Martin Liska <mliska@suse.cz>
669 * config/rs6000/rs6000.c (rs6000_option_override_internal):
672 2020-02-17 Martin Liska <mliska@suse.cz>
675 * config/rx/elf.opt: Fix typo.
677 2020-02-17 Richard Biener <rguenther@suse.de>
680 * opts-global.c (print_ignored_options): Use inform and
683 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
686 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
688 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
691 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
692 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
694 2020-02-15 Jason Merrill <jason@redhat.com>
696 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
698 2020-02-15 Jakub Jelinek <jakub@redhat.com>
700 PR tree-optimization/93744
701 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
702 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
703 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
704 sure @2 in the first and @1 in the other patterns has no side-effects.
706 2020-02-15 David Malcolm <dmalcolm@redhat.com>
707 Bernd Edlinger <bernd.edlinger@hotmail.de>
711 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
712 * configure.ac (--with-diagnostics-urls): New configuration
713 option, based on --with-diagnostics-color.
714 (DIAGNOSTICS_URLS_DEFAULT): New define.
715 * config.h: Regenerate.
716 * configure: Regenerate.
717 * diagnostic.c (diagnostic_urls_init): Handle -1 for
718 DIAGNOSTICS_URLS_DEFAULT from configure-time
719 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
720 and TERM_URLS environment variable.
721 * diagnostic-url.h (diagnostic_url_format): New enum type.
722 (diagnostic_urls_enabled_p): rename to...
723 (determine_url_format): ... this, and change return type.
724 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
725 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
726 the linux console, and mingw.
727 (diagnostic_urls_enabled_p): rename to...
728 (determine_url_format): ... this, and adjust.
729 * pretty-print.h (pretty_printer::show_urls): rename to...
730 (pretty_printer::url_format): ... this, and change to enum.
731 * pretty-print.c (pretty_printer::pretty_printer,
732 pp_begin_url, pp_end_url, test_urls): Adjust.
733 * doc/install.texi (--with-diagnostics-urls): Document the new
734 configuration option.
735 (--with-diagnostics-color): Document the existing interaction
736 with GCC_COLORS better.
737 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
738 vindex reference. Update description of defaults based on the above.
739 (-fdiagnostics-color): Update description of how -fdiagnostics-color
740 interacts with GCC_COLORS.
742 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
745 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
746 conjunction with TARGET_GNU_TLS in early return.
748 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
750 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
751 the mode is not wider than UNITS_PER_WORD.
753 2020-02-14 Martin Jambor <mjambor@suse.cz>
755 PR tree-optimization/93516
756 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
757 access of the same type as the parent.
758 (propagate_subaccesses_from_lhs): Likewise.
760 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
763 * config/i386/avx512vbmi2intrin.h
764 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
765 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
766 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
767 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
768 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
769 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
770 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
771 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
772 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
773 of lacking a closing parenthesis.
774 * config/i386/avx512vbmi2vlintrin.h
775 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
776 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
777 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
778 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
779 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
780 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
781 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
782 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
783 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
784 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
785 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
786 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
787 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
788 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
789 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
790 _mm_shldi_epi32, _mm_mask_shldi_epi32,
791 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
792 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
794 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
797 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
798 the target function entry.
800 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
802 * common/config/arc/arc-common.c (arc_option_optimization_table):
803 Disable if-conversion step when optimized for size.
805 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
807 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
808 R12-R15 are always in ARCOMPACT16_REGS register class.
809 * config/arc/arc.opt (mq-class): Deprecate.
810 * config/arc/constraint.md ("q"): Remove dependency on mq-class
812 * doc/invoke.texi (mq-class): Update text.
813 * common/config/arc/arc-common.c (arc_option_optimization_table):
816 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
818 * config/arc/arc.c (arc_insn_cost): New function.
819 (TARGET_INSN_COST): Define.
820 * config/arc/arc.md (cost): New attribute.
821 (add_n): Use arc_nonmemory_operand.
822 (ashlsi3_insn): Likewise, also update constraints.
823 (ashrsi3_insn): Likewise.
825 (add_shift): Likewise.
826 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
828 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
830 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
832 (umulsidi_600): Likewise.
834 2020-02-13 Jakub Jelinek <jakub@redhat.com>
837 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
838 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
839 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
840 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
841 pass __A to the builtin followed by __W instead of __A followed by
843 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
844 _mm512_mask_popcnt_epi64): Likewise.
845 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
846 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
847 _mm256_mask_popcnt_epi64): Likewise.
849 PR tree-optimization/93582
850 * fold-const.h (shift_bytes_in_array_left,
851 shift_bytes_in_array_right): Declare.
852 * fold-const.c (shift_bytes_in_array_left,
853 shift_bytes_in_array_right): New function, moved from
854 gimple-ssa-store-merging.c, no longer static.
855 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
856 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
857 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
858 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
859 shift_bytes_in_array.
860 (verify_shift_bytes_in_array): Rename to ...
861 (verify_shift_bytes_in_array_left): ... this. Use
862 shift_bytes_in_array_left instead of shift_bytes_in_array.
863 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
864 instead of verify_shift_bytes_in_array.
865 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
866 / native_interpret_expr where the store covers all needed bits,
867 punt on PDP-endian, otherwise allow all involved offsets and sizes
868 not to be byte-aligned.
871 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
872 use const_0_to_255_operand predicate instead of immediate_operand.
873 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
874 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
875 vgf2p8affineinvqb_<mode><mask_name>,
876 vgf2p8affineqb_<mode><mask_name>): Drop mode from
877 const_0_to_255_operand predicated operands.
879 2020-02-12 Jeff Law <law@redhat.com>
881 * config/h8300/h8300.md (comparison shortening peepholes): Use
882 a mode iterator to merge the HImode and SImode peepholes.
884 2020-02-12 Jakub Jelinek <jakub@redhat.com>
887 * real.c (is_even): Make static. Function comment fix.
888 (is_halfway_below): Make static, don't assert R is not inf/nan,
889 instead return false for those. Small formatting fixes.
891 2020-02-12 Martin Sebor <msebor@redhat.com>
894 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
895 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
896 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
897 (strlen_check_and_optimize_call): Adjust callee name.
899 2020-02-12 Jeff Law <law@redhat.com>
901 * config/h8300/h8300.md (comparison shortening peepholes): Drop
902 (and (xor)) variant. Combine other two into single peephole.
904 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
906 PR rtl-optimization/93565
907 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
909 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
911 * config/aarch64/aarch64-simd.md
912 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
913 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
914 generating separate ADDV and zero_extend patterns.
915 * config/aarch64/iterators.md (VDQV_E): New iterator.
917 2020-02-12 Jeff Law <law@redhat.com>
919 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
920 expanders, splits, etc.
921 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
922 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
923 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
924 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
925 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
926 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
928 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
930 2020-02-12 Jakub Jelinek <jakub@redhat.com>
933 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
934 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
935 TARGET_AVX512DQ from condition.
936 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
937 instead of <mask_mode512bit_condition> in condition. If
938 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
940 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
943 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
946 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
948 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
950 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
951 where strlen is more legible.
952 (rs6000_builtin_vectorized_libmass): Ditto.
953 (rs6000_print_options_internal): Ditto.
955 2020-02-11 Martin Sebor <msebor@redhat.com>
957 PR tree-optimization/93683
958 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
960 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
962 * config/rs6000/predicates.md (cint34_operand): Rename the
963 -mprefixed-addr option to be -mprefixed.
964 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
965 the -mprefixed-addr option to be -mprefixed.
966 (OTHER_FUTURE_MASKS): Likewise.
967 (POWERPC_MASKS): Likewise.
968 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
969 the -mprefixed-addr option to be -mprefixed. Change error
970 messages to refer to -mprefixed.
971 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
973 (rs6000_legitimate_offset_address_p): Likewise.
974 (rs6000_mode_dependent_address): Likewise.
975 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
976 "-mprefixed" for target attributes and pragmas.
977 (address_to_insn_form): Rename the -mprefixed-addr option to be
979 (rs6000_adjust_insn_length): Likewise.
980 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
981 -mprefixed-addr option to be -mprefixed.
982 (ASM_OUTPUT_OPCODE): Likewise.
983 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
984 -mprefixed-addr option to be -mprefixed.
985 * config/rs6000/rs6000.opt (-mprefixed): Rename the
986 -mprefixed-addr option to be prefixed. Change the option from
987 being undocumented to being documented.
988 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
989 -mprefixed option. Update the -mpcrel documentation to mention
992 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
994 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
995 including FIRST_PSEUDO_REGISTER - 1.
996 * ira-color.c (print_hard_reg_set): Ditto.
998 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1000 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
1001 (USTERNOP_QUALIFIERS): New define.
1002 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
1003 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
1004 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
1005 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
1006 * config/arm/arm_neon.h (vusdot_s32): New.
1007 (vusdot_lane_s32): New.
1008 (vusdotq_lane_s32): New.
1009 (vsudot_lane_s32): New.
1010 (vsudotq_lane_s32): New.
1011 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
1012 * config/arm/iterators.md (DOTPROD_I8MM): New.
1013 (sup, opsuffix): Add <us/su>.
1014 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
1015 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
1017 2020-02-11 Richard Biener <rguenther@suse.de>
1019 PR tree-optimization/93661
1020 PR tree-optimization/93662
1021 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
1023 * tree-sra.c (get_access_for_expr): Likewise.
1025 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1028 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1029 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1030 Change condition from TARGET_AVX2 to TARGET_AVX.
1032 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1035 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1036 argument of strncmp.
1038 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1040 Try to generate zero-based comparisons.
1041 * config/cris/cris.c (cris_reduce_compare): New function.
1042 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1043 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1044 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1046 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1049 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1050 in Thumb state and also as a destination in Arm state. Add T16
1053 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1055 * md.texi (Define Subst): Match closing paren in example.
1057 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1061 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1062 arguments of strncmp.
1064 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1067 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1068 but different source value.
1069 (adjust_callers_for_value_intersection): New function.
1070 (gather_edges_for_value): Adjust order of callers to let a
1071 non-self-recursive caller be the first element.
1072 (self_recursive_pass_through_p): Add a new parameter "simple", and
1073 check generalized self-recursive pass-through jump function.
1074 (self_recursive_agg_pass_through_p): Likewise.
1075 (find_more_scalar_values_for_callers_subset): Compute value from
1076 pass-through jump function for self-recursive.
1077 (intersect_with_plats): Cleanup previous implementation code for value
1078 itersection with self-recursive call edge.
1079 (intersect_with_agg_replacements): Likewise.
1080 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1081 function for self-recursive call edge. Cleanup previous implementation
1082 code for value intersection with self-recursive call edge.
1083 (decide_whether_version_node): Remove dead callers and adjust order
1084 to let a non-self-recursive caller be the first element.
1086 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1088 * recog.c: Move pass_split_before_sched2 code in front of
1089 pass_split_before_regstack.
1090 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1091 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1092 (rest_of_handle_split_before_sched2): Remove.
1093 (pass_split_before_sched2::execute): Unconditionally call
1095 (enable_split_before_sched2): New function.
1096 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1097 (pass_split_before_regstack::gate): Ditto.
1098 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1099 Update name check for renamed split4 pass.
1100 * config/sh/sh.c (register_sh_passes): Update pass insertion
1101 point for renamed split4 pass.
1103 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1105 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1106 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1107 copying them around between host and target.
1109 2020-02-08 Andrew Pinski <apinski@marvell.com>
1112 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1113 STRICT_ALIGNMENT also.
1115 2020-02-08 Jim Wilson <jimw@sifive.com>
1118 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1120 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1121 Jakub Jelinek <jakub@redhat.com>
1124 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1125 xmm16-xmm31 call-used even in 64-bit ms-abi.
1127 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1129 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1130 (simd_ummla, simd_usmmla): Likewise.
1131 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1132 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1133 (vusmmlaq_s32): New.
1135 2020-02-07 Richard Biener <rguenther@suse.de>
1138 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1139 skipping unreachable regions.
1140 (optimize_inline_calls): Skip folding stmts when we didn't
1143 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1146 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1147 Don't return aggregates with only SFmode and DFmode in SSE
1149 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1151 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1154 * config/rs6000/rs6000-logue.c
1155 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1156 if it fails, move rs into end_addr and retry. Add
1157 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1158 the insn pattern doesn't describe well what exactly happens to
1162 * config/i386/predicates.md (avx_identity_operand): Remove.
1163 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1164 (avx_<castmode><avxsizesuffix>_<castmode>,
1165 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1166 a VEC_CONCAT of the operand and UNSPEC_CAST.
1167 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1168 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1172 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1173 recog_data.insn if distance_non_agu_define changed it.
1175 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1178 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1179 we only had X-FORM (reg+reg) addressing for vectors. Also before
1180 ISA 3.0, we only had X-FORM addressing for scalars in the
1181 traditional Altivec registers.
1183 2020-02-06 <zhongyunde@huawei.com>
1184 Vladimir Makarov <vmakarov@redhat.com>
1186 PR rtl-optimization/93561
1187 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1188 hard register range.
1190 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1192 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1195 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1197 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1198 where the low and the high 32 bits are equal to each other specially,
1199 with an rldimi instruction.
1201 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1203 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1205 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1207 * config/arm/arm-tables.opt: Regenerate.
1209 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1212 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1213 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1214 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1216 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1218 PR rtl-optimization/87763
1219 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1221 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1223 * config/aarch64/aarch64-simd-builtins.def
1224 (bfmlaq): New built-in function.
1225 (bfmlalb): New built-in function.
1226 (bfmlalt): New built-in function.
1227 (bfmlalb_lane): New built-in function.
1228 (bfmlalt_lane): New built-in function.
1229 * config/aarch64/aarch64-simd.md
1230 (aarch64_bfmmlaqv4sf): New pattern.
1231 (aarch64_bfmlal<bt>v4sf): New pattern.
1232 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1233 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1234 (vbfmlalbq_f32): New intrinsic.
1235 (vbfmlaltq_f32): New intrinsic.
1236 (vbfmlalbq_lane_f32): New intrinsic.
1237 (vbfmlaltq_lane_f32): New intrinsic.
1238 (vbfmlalbq_laneq_f32): New intrinsic.
1239 (vbfmlaltq_laneq_f32): New intrinsic.
1240 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1241 (bt): New int attribute.
1243 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1245 * config/i386/i386.md (*pushtf): Emit "#" instead of
1246 calling gcc_unreachable in insn output.
1249 (*pushsf_rex64): Ditto for alternatives other than 1.
1250 (*pushsf): Ditto for alternatives other than 1.
1252 2020-02-06 Martin Liska <mliska@suse.cz>
1254 PR gcov-profile/91971
1255 PR gcov-profile/93466
1256 * coverage.c (coverage_init): Revert mangling of
1257 path into filename. It can lead to huge filename length.
1258 Creation of subfolders seem more natural.
1260 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1263 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1264 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1265 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1267 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1270 * config/i386/predicates.md (avx_identity_operand): New predicate.
1271 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1272 define_insn_and_split.
1275 * omp-low.c (use_pointer_for_field): For nested constructs, also
1276 look for map clauses on target construct.
1277 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1278 taskreg_nesting_level.
1281 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1282 shared clause, call omp_notice_variable on outer context if any.
1284 2020-02-05 Jason Merrill <jason@redhat.com>
1287 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1288 non-zero address even if weak and not yet defined.
1290 2020-02-05 Martin Sebor <msebor@redhat.com>
1292 PR tree-optimization/92765
1293 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1294 * tree-ssa-strlen.c (compute_string_length): Remove.
1295 (determine_min_objsize): Remove.
1296 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1297 Avoid using type size as the upper bound on string length.
1298 (handle_builtin_string_cmp): Add an argument. Adjust.
1299 (strlen_check_and_optimize_call): Pass additional argument to
1300 handle_builtin_string_cmp.
1302 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1304 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1305 (*pushdi2_rex64 peephole2): Unconditionally split after
1307 (*ashl<mode>3_doubleword): Ditto.
1308 (*<shift_insn><mode>3_doubleword): Ditto.
1310 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1313 * config/rs6000/rs6000.c (get_vector_offset): Fix
1315 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1317 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1319 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1322 (Special Functions for Debugging the Analyzer): Update description
1323 of __analyzer_dump_exploded_nodes.
1325 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1328 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1329 include sets and not clobbers in the vzeroupper pattern.
1330 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1331 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1332 (*avx_vzeroupper_1): New define_insn_and_split.
1335 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1336 don't run when !optimize.
1337 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1340 2020-02-05 Richard Biener <rguenther@suse.de>
1343 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1344 checks before matching calls.
1346 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1348 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1349 function comment typo.
1352 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1353 simd_clone_create failed when i == 0, adjust clone->nargs by
1356 2020-02-05 Martin Liska <mliska@suse.cz>
1359 * doc/invoke.texi: Document that one should
1360 not combine ASLR and -fpch.
1362 2020-02-04 Richard Biener <rguenther@suse.de>
1364 PR tree-optimization/93538
1365 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1367 2020-02-04 Richard Biener <rguenther@suse.de>
1369 PR tree-optimization/91123
1370 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1371 (vn_walk_cb_data::last_vuse): New member.
1372 (vn_walk_cb_data::saved_operands): Likewsie.
1373 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1374 (vn_walk_cb_data::push_partial_def): Use finish.
1375 (vn_reference_lookup_2): Update last_vuse and use finish if
1376 we've saved operands.
1377 (vn_reference_lookup_3): Use finish and update calls to
1378 push_partial_defs everywhere. When translating through
1379 memcpy or aggregate copies save off operands and alias-set.
1380 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1381 operation for redundant store removal.
1383 2020-02-04 Richard Biener <rguenther@suse.de>
1385 PR tree-optimization/92819
1386 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1387 generating more stmts than before.
1389 2020-02-04 Martin Liska <mliska@suse.cz>
1391 * config/arm/arm.c (arm_gen_far_branch): Move the function
1392 outside of selftests.
1394 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1396 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1397 function to adjust PC-relative vector addresses.
1398 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1399 handle vectors with PC-relative addresses.
1401 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1403 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1405 (hard_reg_and_mode_to_addr_mask): Delete.
1406 (rs6000_adjust_vec_address): If the original vector address
1407 was REG+REG or REG+OFFSET and the element is not zero, do the add
1408 of the elements in the original address before adding the offset
1409 for the vector element. Use address_to_insn_form to validate the
1410 address using the register being loaded, rather than guessing
1411 whether the address is a DS-FORM or DQ-FORM address.
1413 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1415 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1416 to calculate the offset in memory from the start of a vector of a
1417 particular element. Add code to keep the element number in
1418 bounds if the element number is variable.
1419 (rs6000_adjust_vec_address): Move calculation of offset of the
1420 vector element to get_vector_offset.
1421 (rs6000_split_vec_extract_var): Do not do the initial AND of
1422 element here, move the code to get_vector_offset.
1424 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1426 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1429 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1431 * config/rs6000/constraints.md: Improve documentation.
1433 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1436 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1437 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1439 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1441 * config.gcc: Remove "carrizo" support.
1442 * config/gcn/gcn-opts.h (processor_type): Likewise.
1443 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1444 * config/gcn/gcn.opt (gpu_type): Likewise.
1445 * config/gcn/t-omp-device: Likewise.
1447 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1450 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1451 * config/arm/arm.c (arm_gen_far_branch): New function
1453 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1455 2020-02-03 Julian Brown <julian@codesourcery.com>
1456 Tobias Burnus <tobias@codesourcery.com>
1458 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1460 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1463 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1464 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1467 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1469 PR rtl-optimization/91333
1470 * ira-color.c (struct allocno_color_data): Add member
1472 (init_allocno_threads): Set the member up.
1473 (bucket_allocno_compare_func): Add compare hard reg
1476 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1478 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1480 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1481 * config.in: Regenerated.
1482 * configure: Regenerated.
1483 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1484 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1485 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1487 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1489 * configure: Regenerate.
1491 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1493 PR rtl-optimization/91333
1494 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1495 reg preferences comparison up.
1497 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1499 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1500 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1501 aarch64-sve-builtins-base.h.
1502 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1503 aarch64-sve-builtins-base.cc.
1504 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1505 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1507 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1508 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1509 (svcvtnt): New functions.
1510 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1511 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1512 (svcvtnt): New functions.
1513 (svcvt): Add a form that converts f32 to bf16.
1514 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1515 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1517 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1518 Treat B as bfloat16_t.
1519 (ternary_bfloat_lane_base): New class.
1520 (ternary_bfloat_def): Likewise.
1521 (ternary_bfloat): New shape.
1522 (ternary_bfloat_lane_def): New class.
1523 (ternary_bfloat_lane): New shape.
1524 (ternary_bfloat_lanex2_def): New class.
1525 (ternary_bfloat_lanex2): New shape.
1526 (ternary_bfloat_opt_n_def): New class.
1527 (ternary_bfloat_opt_n): New shape.
1528 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1529 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1530 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1531 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1532 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1533 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1534 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1535 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1536 the pattern off the narrow mode instead of the wider one.
1537 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1538 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1539 (sve_fp_op): Handle them.
1540 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1541 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1543 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1545 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1546 * config/aarch64/aarch64-modes.def (BF): Move definition before
1547 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1548 (SVE_MODES): Handle BF modes.
1549 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1551 (aarch64_full_sve_mode): Likewise.
1552 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1554 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1555 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1556 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1557 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1559 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1561 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1562 (TYPES_all_data): Add bf16.
1563 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1564 (register_tuple_type): Increase buffer size.
1565 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1566 (bf16): New type suffix.
1567 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1568 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1569 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1570 Change type from all_data to all_arith.
1571 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1574 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1575 Matthew Malcomson <matthew.malcomson@arm.com>
1576 Richard Sandiford <richard.sandiford@arm.com>
1578 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1579 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1580 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1581 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1582 __ARM_FEATURE_MATMUL_FP64.
1583 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1584 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1585 be disabled at the same time.
1586 (f32mm): New extension.
1587 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1588 (AARCH64_FL_F64MM): Bump to the next bit up.
1589 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1590 (TARGET_SVE_F64MM): New macros.
1591 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1592 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1593 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1594 (UNSPEC_ZIP2Q): New unspeccs.
1595 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1596 (optab, sur, perm_insn): Handle the new unspecs.
1597 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1598 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1599 TARGET_SVE_F64MM instead of separate tests.
1600 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1601 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1602 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1603 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1604 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1605 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1606 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1607 (TYPES_s_signed): New macro.
1608 (TYPES_s_integer): Use it.
1609 (TYPES_d_float): New macro.
1610 (TYPES_d_data): Use it.
1611 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1612 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1613 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1614 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1615 (svmmla): New shape.
1616 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1617 template parameters.
1618 (ternary_resize2_lane_base): Likewise.
1619 (ternary_resize2_base): New class.
1620 (ternary_qq_lane_base): Likewise.
1621 (ternary_intq_uintq_lane_def): Likewise.
1622 (ternary_intq_uintq_lane): New shape.
1623 (ternary_intq_uintq_opt_n_def): New class
1624 (ternary_intq_uintq_opt_n): New shape.
1625 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1626 (ternary_uintq_intq_def): New class.
1627 (ternary_uintq_intq): New shape.
1628 (ternary_uintq_intq_lane_def): New class.
1629 (ternary_uintq_intq_lane): New shape.
1630 (ternary_uintq_intq_opt_n_def): New class.
1631 (ternary_uintq_intq_opt_n): New shape.
1632 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1633 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1634 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1635 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1637 (svdotprod_lane_impl): ...this new class.
1638 (svmmla_impl, svusdot_impl): New classes.
1639 (svdot_lane): Update to use svdotprod_lane_impl.
1640 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1641 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1643 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1644 function, with no types defined.
1645 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1646 AARCH64_FL_I8MM functions.
1647 (svmmla): New AARCH64_FL_F32MM function.
1648 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1649 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1650 AARCH64_FL_F64MM function.
1651 (REQUIRED_EXTENSIONS):
1653 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1655 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1658 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1660 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1661 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1662 (*movti_internal): Do not check for
1663 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1664 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1665 just after check for TARGET_AVX.
1666 (*movdf_internal): Ditto.
1667 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1668 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1669 * config/i386/sse.md (mov<mode>_internal): Only check
1670 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1671 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1672 (<sse>_andnot<mode>3<mask_name>): Move check for
1673 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1674 (<code><mode>3<mask_name>): Ditto.
1675 (*andnot<mode>3): Ditto.
1676 (*andnottf3): Ditto.
1677 (*<code><mode>3): Ditto.
1678 (*<code>tf3): Ditto.
1679 (*andnot<VI:mode>3): Remove
1680 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1681 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1682 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1683 (sse4_1_blendv<ssemodesuffix>): Ditto.
1684 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1685 Explain that tune applies to 128bit instructions only.
1687 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1689 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1690 to definition of hsa_kernel_description. Parse assembly to find SGPR
1691 and VGPR count of kernel and store in hsa_kernel_description.
1693 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1695 PR rtl-optimization/91838
1696 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1697 to truncate if allowed or reject combination.
1699 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1701 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1702 (find_inv_vars_cb): Likewise.
1704 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1706 * calls.c (special_function_p): Split out the check for DECL_NAME
1707 being non-NULL and fndecl being extern at file scope into a
1708 new maybe_special_function_p and call it. Drop check for fndecl
1709 being non-NULL that was after a usage of DECL_NAME (fndecl).
1710 * tree.h (maybe_special_function_p): New inline function.
1712 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1714 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1715 (mask_gather_load<mode>): ... here, and zero-initialize the
1717 (maskload<mode>di): Zero-initialize the destination.
1720 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1723 * doc/analyzer.texi (Limitations): Note that constraints on
1724 floating-point values are currently ignored.
1726 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1729 * symtab.c (symtab_node::noninterposable_alias): If localalias
1730 already exists, but is not usable, append numbers after it until
1731 a unique name is found. Formatting fix.
1734 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1737 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1739 * config/gcn/gcn.c (print_operand): Handle LTGT.
1740 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1742 2020-01-30 Richard Biener <rguenther@suse.de>
1744 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1745 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1747 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1749 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1750 without a DECL in .data.rel.ro.local.
1752 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1755 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1759 * config/i386/sse.md
1760 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1761 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1762 any_extend code iterator instead of always zero_extend.
1763 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1764 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1765 Use any_extend code iterator instead of always zero_extend.
1766 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1767 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1768 Use any_extend code iterator instead of always zero_extend.
1769 (*sse2_pmovmskb_ext): New define_insn.
1770 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1773 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1774 (*popcountsi2_zext_falsedep): New define_insn.
1776 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1778 * config.in: Regenerated.
1779 * configure: Regenerated.
1781 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1784 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1785 LLVM's assembler changed the default in version 9.
1787 2020-01-24 Jeff Law <law@redhat.com>
1789 PR tree-optimization/89689
1790 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1792 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1796 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1798 PR rtl-optimization/87763
1799 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1800 simplification to handle subregs as well as bare regs.
1801 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1803 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1806 * ira.c (ira): Revert use of simplified LRA algorithm.
1808 2020-01-29 Martin Jambor <mjambor@suse.cz>
1810 PR tree-optimization/92706
1811 * tree-sra.c (struct access): Fields first_link, last_link,
1812 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1813 next_rhs_queued and grp_rhs_queued respectively, new fields
1814 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1815 (struct assign_link): Field next renamed to next_rhs, new field
1816 next_lhs. Updated comment.
1817 (work_queue_head): Renamed to rhs_work_queue_head.
1818 (lhs_work_queue_head): New variable.
1819 (add_link_to_lhs): New function.
1820 (relink_to_new_repr): Also relink LHS lists.
1821 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1822 (add_access_to_lhs_work_queue): New function.
1823 (pop_access_from_work_queue): Renamed to
1824 pop_access_from_rhs_work_queue.
1825 (pop_access_from_lhs_work_queue): New function.
1826 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1828 (child_would_conflict_in_lacc): Renamed to
1829 child_would_conflict_in_acc. Adjusted parameter names.
1830 (create_artificial_child_access): New parameter set_grp_read, use it.
1831 (subtree_mark_written_and_enqueue): Renamed to
1832 subtree_mark_written_and_rhs_enqueue.
1833 (propagate_subaccesses_across_link): Renamed to
1834 propagate_subaccesses_from_rhs.
1835 (propagate_subaccesses_from_lhs): New function.
1836 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1839 2020-01-29 Martin Jambor <mjambor@suse.cz>
1841 PR tree-optimization/92706
1842 * tree-sra.c (struct access): Adjust comment of
1843 grp_total_scalarization.
1844 (find_access_in_subtree): Look for single children spanning an entire
1846 (scalarizable_type_p): Allow register accesses, adjust callers.
1847 (completely_scalarize): Remove function.
1848 (scalarize_elem): Likewise.
1849 (create_total_scalarization_access): Likewise.
1850 (sort_and_splice_var_accesses): Do not track total scalarization
1852 (analyze_access_subtree): New parameter totally, adjust to new meaning
1853 of grp_total_scalarization.
1854 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1855 (can_totally_scalarize_forest_p): New function.
1856 (create_total_scalarization_access): Likewise.
1857 (create_total_access_and_reshape): Likewise.
1858 (total_should_skip_creating_access): Likewise.
1859 (totally_scalarize_subtree): Likewise.
1860 (analyze_all_variable_accesses): Perform total scalarization after
1861 subaccess propagation using the new functions above.
1862 (initialize_constant_pool_replacements): Output initializers by
1863 traversing the access tree.
1865 2020-01-29 Martin Jambor <mjambor@suse.cz>
1867 * tree-sra.c (verify_sra_access_forest): New function.
1868 (verify_all_sra_access_forests): Likewise.
1869 (create_artificial_child_access): Set parent.
1870 (analyze_all_variable_accesses): Call the verifier.
1872 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1874 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1875 if called on indirect edge.
1876 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1877 speculative call if needed.
1879 2020-01-29 Richard Biener <rguenther@suse.de>
1881 PR tree-optimization/93428
1882 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1883 permutation when the load node is created.
1884 (vect_analyze_slp_instance): Re-use it here.
1886 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1888 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1890 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1892 PR rtl-optimization/93272
1893 * ira-lives.c (process_out_of_region_eh_regs): New function.
1894 (process_bb_node_lives): Call it.
1896 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1898 * coverage.c (read_counts_file): Make error message lowercase.
1900 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1902 * profile-count.c (profile_quality_display_names): Fix ordering.
1904 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1907 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1908 hash only when edge is first within the sequence.
1909 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1910 (symbol_table::create_edge): Do not set target_prob.
1911 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1913 (cgraph_edge::make_speculative): Drop target_prob parameter.
1914 (cgraph_edge::speculative_call_info): Remove.
1915 (cgraph_edge::first_speculative_call_target): New member function.
1916 (update_call_stmt_hash_for_removing_direct_edge): New function.
1917 (cgraph_edge::resolve_speculation): Rewrite to new API.
1918 (cgraph_edge::speculative_call_for_target): New member function.
1919 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1920 multiple speculation targets.
1921 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1923 (verify_speculative_call): Verify that targets form an interval.
1924 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1925 (cgraph_edge::first_speculative_call_target): New member function.
1926 (cgraph_edge::next_speculative_call_target): New member function.
1927 (cgraph_edge::speculative_call_target_ref): New member function.
1928 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1929 (cgraph_edge): Remove target_prob.
1930 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1931 Fix handling of speculative calls.
1932 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1933 * ipa-fnsummary.c (analyze_function_body): Likewise.
1934 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1935 * ipa-profile.c (dump_histogram): Fix formating.
1936 (ipa_profile_generate_summary): Watch for overflows.
1937 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1938 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1939 (update_indirect_edges_after_inlining): Update to new API.
1940 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1942 * profile-count.h: (profile_probability::adjusted): New.
1943 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1944 updating of profile.
1945 * value-prof.c (gimple_ic_transform): Rename to ...
1946 (dump_ic_profile): ... this one; update dumping.
1947 (stream_in_histogram_value): Fix formating.
1948 (gimple_value_profile_transformations): Update.
1950 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1953 * config/i386/i386.md (*movoi_internal_avx): Remove
1954 TARGET_SSE_TYPELESS_STORES check.
1955 (*movti_internal): Prefer TARGET_AVX over
1956 TARGET_SSE_TYPELESS_STORES.
1957 (*movtf_internal): Likewise.
1958 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1959 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1960 from TARGET_SSE_TYPELESS_STORES.
1962 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1964 * diagnostic-core.h (warning_at): Rename overload to...
1965 (warning_meta): ...this.
1966 (emit_diagnostic_valist): Delete decl of overload taking
1967 diagnostic_metadata.
1968 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1969 (warning_at): Rename overload taking diagnostic_metadata to...
1970 (warning_meta): ...this.
1972 2020-01-28 Richard Biener <rguenther@suse.de>
1974 PR tree-optimization/93439
1975 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1976 * tree-cfg.c (move_sese_region_to_fn): ... here.
1977 (verify_types_in_gimple_reference): Verify used cliques are
1980 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1983 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1984 argument of a pointer to struct gcc_options and pass it to
1985 parse_mtune_ctrl_str.
1986 (ix86_function_specific_restore): Pass opts to
1987 set_ix86_tune_features.
1988 (ix86_option_override_internal): Likewise.
1989 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1990 gcc_options and use it for x_ix86_tune_ctrl_string.
1992 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1994 PR rtl-optimization/87763
1995 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1996 simplification to handle subregs as well as bare regs.
1997 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1999 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2001 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
2002 for reduction chains that (now) include a call.
2004 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2006 PR tree-optimization/92822
2007 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
2008 out the don't-care elements of a vector whose significant elements
2009 are duplicates, make the don't-care elements duplicates too.
2011 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
2013 PR tree-optimization/93434
2014 * tree-predcom.c (split_data_refs_to_components): Record which
2015 components have had aliasing loads removed. Prevent store-store
2016 commoning for all such components.
2018 2020-01-28 Jakub Jelinek <jakub@redhat.com>
2021 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
2022 -1 or is_vshift is true, use new_vector with number of elts npatterns
2023 rather than new_unary_operation.
2025 PR tree-optimization/93454
2026 * gimple-fold.c (fold_array_ctor_reference): Perform
2027 elt_size.to_uhwi () just once, instead of calling it in every
2028 iteration. Punt if that value is above size of the temporary
2029 buffer. Decrease third native_encode_expr argument when
2030 bufoff + elt_sz is above size of buf.
2032 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2034 * config/mips/mips.c (mips_declare_object_name)
2035 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2037 2020-01-27 Martin Liska <mliska@suse.cz>
2039 PR gcov-profile/93403
2040 * tree-profile.c (gimple_init_gcov_profiler): Generate
2041 both __gcov_indirect_call_profiler_v4 and
2042 __gcov_indirect_call_profiler_v4_atomic.
2044 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2047 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2049 (@aarch64_split_simd_mov<mode>): Use it.
2050 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2051 Leave the vec_extract patterns to handle 2-element vectors.
2052 (aarch64_simd_mov_from_<mode>high): Likewise.
2053 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2054 (vec_extractv2dfv1df): Likewise.
2056 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2058 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2059 jump conditions for *compare_condjump<GPI:mode>.
2061 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2064 * digraph.cc (test_edge::test_edge): Specify template for base
2067 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2069 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2071 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2073 * config/arc/arc-protos.h (gen_mlo): Remove.
2074 (gen_mhi): Likewise.
2075 * config/arc/arc.c (AUX_MULHI): Define.
2076 (arc_must_save_reister): Special handling for r58/59.
2077 (arc_compute_frame_size): Consider mlo/mhi registers.
2078 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2080 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2081 mlo/mhi name selection.
2082 (arc_restore_callee_saves): Don't early restore blink when ISR.
2083 (arc_expand_prologue): Add mlo/mhi saving.
2084 (arc_expand_epilogue): Add mlo/mhi restoring.
2087 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2088 numbering when MUL64 option is used.
2089 (DWARF2_FRAME_REG_OUT): Define.
2090 * config/arc/arc.md (arc600_stall): New pattern.
2091 (VUNSPEC_ARC_ARC600_STALL): Define.
2092 (mulsi64): Use correct mlo/mhi registers.
2093 (mulsi_600): Clean it up.
2094 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2096 (mhi_operand): Likewise.
2098 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2099 Petro Karashchenko <petro.karashchenko@ring.com>
2101 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2102 attributes if needed.
2103 (prepare_move_operands): Generate special unspec instruction for
2105 (arc_isuncached_mem_p): Propagate uncached attribute to each
2107 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2108 (VUNSPEC_ARC_STDI): Likewise.
2109 (ALLI): New mode iterator.
2110 (mALLI): New mode attribute.
2111 (lddi): New instruction pattern.
2113 (stdidi_split): Split instruction for architectures which are not
2114 supporting ll64 option.
2115 (lddidi_split): Likewise.
2117 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2119 PR rtl-optimization/92989
2120 * lra-lives.c (process_bb_lives): Update the live-in set before
2121 processing additional clobbers.
2123 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2125 PR rtl-optimization/93170
2126 * cselib.c (cselib_invalidate_regno_val): New function, split out
2128 (cselib_invalidate_regno): ...here.
2129 (cselib_invalidated_by_call_p): New function.
2130 (cselib_process_insn): Iterate over all the hard-register entries in
2131 REG_VALUES and invalidate any that cross call-clobbered registers.
2133 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2135 * dojump.c (split_comparison): Use HONOR_NANS rather than
2136 HONOR_SNANS when splitting LTGT.
2138 2020-01-27 Martin Liska <mliska@suse.cz>
2141 * opts.c (print_filtered_help): Exclude language-specific
2142 options from --help=common unless enabled in all FEs.
2144 2020-01-27 Martin Liska <mliska@suse.cz>
2146 * opts.c (print_help): Exclude params from
2147 all except --help=param.
2149 2020-01-27 Martin Liska <mliska@suse.cz>
2152 * config/i386/i386-features.c (make_resolver_func):
2153 Align the code with ppc64 target implementation.
2154 Do not generate a unique name for resolver function.
2156 2020-01-27 Richard Biener <rguenther@suse.de>
2158 PR tree-optimization/93397
2159 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2160 converted reduction chain SLP graph adjustment.
2162 2020-01-26 Marek Polacek <polacek@redhat.com>
2165 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2168 2020-01-26 Jason Merrill <jason@redhat.com>
2171 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2174 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2176 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2177 (rx_setmem): Likewise.
2179 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2182 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2183 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2184 drop <di> from constraint of last operand.
2187 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2188 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2189 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2191 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2194 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2197 2020-01-24 Jeff Law <law@redhat.com>
2199 PR tree-optimization/92788
2200 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2203 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2206 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2207 *avx_vperm_broadcast_<mode>,
2208 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2209 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2210 Move before avx2_perm<mode>/avx512f_perm<mode>.
2213 * simplify-rtx.c (simplify_const_unary_operation,
2214 simplify_const_binary_operation): Punt for mode precision above
2215 MAX_BITSIZE_MODE_ANY_INT.
2217 2020-01-24 Andrew Pinski <apinski@marvell.com>
2219 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2222 2020-01-24 Jeff Law <law@redhat.com>
2225 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2226 for REGs. Call output_operand_lossage to get more reasonable
2229 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2231 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2232 gcn_fp_compare_operator.
2233 (vec_cmpu<mode>di): Use gcn_compare_operator.
2234 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2235 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2236 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2237 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2238 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2239 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2240 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2241 gcn_fp_compare_operator.
2242 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2243 gcn_fp_compare_operator.
2244 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2245 gcn_fp_compare_operator.
2246 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2247 gcn_fp_compare_operator.
2249 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2251 * doc/install.texi (Cross-Compiler-Specific Options): Document
2252 `--with-toolexeclibdir' option.
2254 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2256 * target.def (flags_regnum): Also mention effect on delay slot filling.
2257 * doc/tm.texi: Regenerate.
2259 2020-01-23 Jeff Law <law@redhat.com>
2261 PR translation/90162
2262 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2264 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2267 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2270 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2272 PR rtl-optimization/93402
2273 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2276 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2278 * config.in: Regenerated.
2279 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2280 for TARGET_LIBC_GNUSTACK.
2281 * configure: Regenerated.
2282 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2283 found to be 2.31 or greater.
2285 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2287 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2289 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2290 (mips_asm_file_end): New function. Delegate to
2291 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2292 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2294 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2297 * config/i386/i386-modes.def (POImode): New mode.
2298 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2299 * config/i386/i386.md (DPWI): New mode attribute.
2300 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2302 (QPWI): ... this. Use POI instead of OI for TImode.
2303 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2304 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2307 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2310 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2312 (speculation_tracker_rev): New pattern.
2313 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2314 Use speculation_tracker_rev to track the inverse condition.
2316 2020-01-23 Richard Biener <rguenther@suse.de>
2318 PR tree-optimization/93381
2319 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2320 alias-set of the def as argument and record the first one.
2321 (vn_walk_cb_data::first_set): New member.
2322 (vn_reference_lookup_3): Pass the alias-set of the current def
2323 to push_partial_def. Fix alias-set used in the aggregate copy
2325 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2326 * real.c (clear_significand_below): Fix out-of-bound access.
2328 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2331 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2332 New define_insn patterns.
2334 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2336 * doc/sourcebuild.texi (check-function-bodies): Add an
2337 optional target/xfail selector.
2339 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2341 PR rtl-optimization/93124
2342 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2343 bare USE and CLOBBER insns.
2345 2020-01-22 Andrew Pinski <apinski@marvell.com>
2347 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2349 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2352 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2353 diagnostic_manager into "ana" namespace.
2354 * selftest-run-tests.c (selftest::run_tests): Update for move of
2355 selftest::run_analyzer_selftests to
2356 ana::selftest::run_analyzer_selftests.
2358 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2360 * cfgexpand.c (union_stack_vars): Update the size.
2362 2020-01-22 Richard Biener <rguenther@suse.de>
2364 PR tree-optimization/93381
2365 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2366 throughout, handle all conversions the same.
2368 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2371 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2372 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2373 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2374 Call force_reg on high_in2 unconditionally.
2376 2020-01-22 Martin Liska <mliska@suse.cz>
2378 PR tree-optimization/92924
2379 * profile.c (compute_value_histograms): Divide
2382 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2385 * output.h (assemble_name_resolve): Declare.
2386 * varasm.c (assemble_name_resolve): New function.
2387 (assemble_name): Use it.
2388 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2390 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2392 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2393 update_web_docs_git instead of update_web_docs_svn.
2395 2020-01-21 Andrew Pinski <apinski@marvell.com>
2398 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2399 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2400 (*tlsgd_small_<mode>): Likewise.
2401 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2402 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2403 register. Convert that register back to dest using convert_mode.
2405 2020-01-21 Jim Wilson <jimw@sifive.com>
2407 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2410 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2411 Uros Bizjak <ubizjak@gmail.com>
2414 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2416 (legitimize_tls_address): Do GNU2 TLS address computation in
2417 ptr_mode and zero-extend result to Pmode.
2418 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2419 :P with :PTR and Pmode with ptr_mode.
2420 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2421 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2422 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2424 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2427 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2428 the last two operands are CONST_INT_P before using them as such.
2430 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2432 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2433 to get the integer element types.
2435 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2437 * config/aarch64/aarch64-sve-builtins.h
2438 (function_expander::convert_to_pmode): Declare.
2439 * config/aarch64/aarch64-sve-builtins.cc
2440 (function_expander::convert_to_pmode): New function.
2441 (function_expander::get_contiguous_base): Use it.
2442 (function_expander::prepare_gather_address_operands): Likewise.
2443 * config/aarch64/aarch64-sve-builtins-sve2.cc
2444 (svwhilerw_svwhilewr_impl::expand): Likewise.
2446 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2449 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2450 cfun->machine->label_is_assembled.
2451 (aarch64_print_patchable_function_entry): New.
2452 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2453 * config/aarch64/aarch64.h (struct machine_function): New field,
2456 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2459 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2462 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2465 * cgraph.c (cgraph_edge::resolve_speculation,
2466 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2467 call_stmt_site_hash.
2469 2020-01-21 Martin Liska <mliska@suse.cz>
2471 * config/rs6000/rs6000.c (common_mode_defined): Remove
2474 2020-01-21 Richard Biener <rguenther@suse.de>
2476 PR tree-optimization/92328
2477 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2478 type when value-numbering same-sized store by inserting a
2480 (eliminate_dom_walker::eliminate_stmt): When eliminating
2481 a redundant store handle bit-reinterpretation of the same value.
2483 2020-01-21 Andrew Pinski <apinski@marvel.com>
2486 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2488 (prepare_block_for_update): This. Use a worklist instead of
2491 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2493 * gcc/config/arm/arm.c (clear_operation_p):
2494 Initialise last_regno, skip first iteration
2495 based on the first_set value and use ints instead
2496 of the unnecessary HOST_WIDE_INTs.
2498 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2501 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2502 compare_mode other than SFmode or DFmode.
2504 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2507 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2508 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2509 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2511 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2513 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2515 2020-01-20 Andrew Pinski <apinski@marvell.com>
2518 * targhooks.c (default_print_patchable_function_entry): Use
2519 output_asm_insn to emit the nop instruction.
2521 2020-01-20 Fangrui Song <maskray@google.com>
2524 * targhooks.c (default_print_patchable_function_entry): Align to
2527 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2530 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2531 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2532 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2533 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2534 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2535 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2536 Remove the {q} suffix from lea.
2537 (*tls_dynamic_gnu2_call_64): Renamed to ...
2538 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2539 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2540 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2541 Pass Pmode to gen_tls_dynamic_gnu2_64.
2543 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2545 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2547 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2549 * config/aarch64/aarch64-sve-builtins-base.cc
2550 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2552 2020-01-20 Richard Biener <rguenther@suse.de>
2555 * dwarf2out.c (prune_unused_types): Unconditionally mark
2556 called function DIEs.
2558 2020-01-20 Martin Liska <mliska@suse.cz>
2560 PR tree-optimization/93199
2561 * tree-eh.c (struct leh_state): Add
2562 new field outer_non_cleanup.
2563 (cleanup_is_dead_in): Pass leh_state instead
2564 of eh_region. Add a checking that state->outer_non_cleanup
2565 points to outer non-clean up region.
2566 (lower_try_finally): Record outer_non_cleanup
2568 (lower_catch): Likewise.
2569 (lower_eh_filter): Likewise.
2570 (lower_eh_must_not_throw): Likewise.
2571 (lower_cleanup): Likewise.
2573 2020-01-20 Richard Biener <rguenther@suse.de>
2575 PR tree-optimization/93094
2576 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2577 (vect_transform_loop): Likewise.
2578 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2579 loop_vectorized_call to vect_transform_loop.
2580 * tree-vect-loop.c (vect_transform_loop): Pass down
2581 loop_vectorized_call to vect_loop_versioning.
2582 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2583 the earlier discovered loop_vectorized_call.
2585 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2587 * doc/contribute.texi: Update for SVN -> Git transition.
2588 * doc/install.texi: Likewise.
2590 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2593 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2594 speculative targets.
2595 (verify_speculative_call): New function
2596 (cgraph_node::verify_node): Use it.
2597 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2600 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2603 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2604 (cgraph_edge::make_direct): Remove all indirect targets.
2605 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2606 (cgraph_node::verify_node): Verify that only one call_stmt or
2607 lto_stmt_uid is set.
2608 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2610 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2611 (lto_output_ref): Simplify streaming of stmt.
2612 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2614 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2616 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2617 Mark parameter unused.
2619 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2621 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2623 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2625 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2627 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2629 * Makefile.in: Add coroutine-passes.o.
2630 * builtin-types.def (BT_CONST_SIZE): New.
2631 (BT_FN_BOOL_PTR): New.
2632 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2633 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2634 * coroutine-builtins.def: New file.
2635 * coroutine-passes.cc: New file.
2636 * function.h (struct GTY function): Add a bit to indicate that the
2637 function is a coroutine component.
2638 * internal-fn.c (expand_CO_FRAME): New.
2639 (expand_CO_YIELD): New.
2640 (expand_CO_SUSPN): New.
2641 (expand_CO_ACTOR): New.
2642 * internal-fn.def (CO_ACTOR): New.
2646 * passes.def: Add pass_coroutine_lower_builtins,
2647 pass_coroutine_early_expand_ifns.
2648 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2649 (make_pass_coroutine_early_expand_ifns): New.
2650 * doc/invoke.texi: Document the fcoroutines command line
2653 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2655 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2658 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2659 after checking the argument is a REG. Don't use REGNO (reg)
2660 again to set last_regno, reuse regno variable instead.
2662 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2664 * doc/analyzer.texi (Limitations): Add note about NaN.
2666 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2667 Sudakshina Das <sudi.das@arm.com>
2669 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2670 and valid immediate.
2671 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2672 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2673 * config/arm/constraints.md (Pg): New.
2674 * config/arm/predicates.md (long_shift_imm): New.
2675 (arm_reg_or_long_shift_imm): Likewise.
2676 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2677 (thumb2_lsll): Likewise.
2680 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2681 Sudakshina Das <sudi.das@arm.com>
2683 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2684 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2685 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2686 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2687 * config/arm/thumb2.md (thumb2_asrl): New.
2688 (thumb2_lsll): Likewise.
2690 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2692 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2695 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2697 * gdbinit.in (help-gcc-hooks): New command.
2698 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2699 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2702 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2704 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2705 correct target macro.
2707 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2709 * config/aarch64/aarch64-protos.h
2710 (aarch64_sve_ld1ro_operand_p): New.
2711 * config/aarch64/aarch64-sve-builtins-base.cc
2712 (class load_replicate): New.
2713 (class svld1ro_impl): New.
2714 (class svld1rq_impl): Change to inherit from load_replicate.
2715 (svld1ro): New sve intrinsic function base.
2716 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2717 New DEF_SVE_FUNCTION.
2718 * config/aarch64/aarch64-sve-builtins-base.h
2719 (svld1ro): New decl.
2720 * config/aarch64/aarch64-sve-builtins.cc
2721 (function_expander::add_mem_operand): Modify assert to allow
2723 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2725 * config/aarch64/aarch64.c
2726 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2727 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2728 (aarch64_sve_ld1ro_operand_p): New.
2729 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2730 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2731 * config/aarch64/predicates.md
2732 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2734 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2736 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2737 Introduce this ACLE specified predefined macro.
2738 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2739 (fp): Disabling this disables f64mm.
2740 (simd): Disabling this disables f64mm.
2741 (fp16): Disabling this disables f64mm.
2742 (sve): Disabling this disables f64mm.
2743 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2744 (AARCH64_ISA_F64MM): New.
2745 (TARGET_F64MM): New.
2746 * doc/invoke.texi (f64mm): Document new option.
2748 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2750 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2751 (neoversen1_tunings): Likewise.
2753 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2756 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2757 Add assert to ensure prolog has been emitted.
2758 (aarch64_split_atomic_op): Likewise.
2759 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2760 Use epilogue_completed rather than reload_completed.
2761 (aarch64_atomic_exchange<mode>): Likewise.
2762 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2763 (atomic_nand<mode>): Likewise.
2764 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2765 (atomic_fetch_nand<mode>): Likewise.
2766 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2767 (atomic_nand_fetch<mode>): Likewise.
2769 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2772 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2774 (REVERSE_CONDITION): Delete.
2775 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2776 (CCFP_CCFPE): Likewise.
2777 (e): New mode attribute.
2778 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2779 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2780 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2781 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2782 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2783 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2784 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2785 name of generator from gen_ccmpdi to gen_ccmpccdi.
2786 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2787 the previous comparison but aren't able to, use the new ccmp_rev
2790 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2792 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2793 than testing directly for INTEGER_CST.
2794 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2796 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2798 PR tree-optimization/93292
2799 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2800 get_vectype_for_scalar_type returns NULL.
2802 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2804 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2805 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2807 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2809 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2811 * params.opt: (max-predicted-iterations): Set bounds.
2812 * predict.c (real_almost_one, real_br_prob_base,
2813 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2814 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2815 probabilities; do not truncate to reg_br_prob_bases.
2816 (estimate_loops_at_level): Pass max_cyclic_prob.
2817 (estimate_loops): Compute max_cyclic_prob.
2818 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2820 * profile-count.c (profile_probability::to_sreal): New.
2821 * profile-count.h (class sreal): Move up in file.
2822 (profile_probability::to_sreal): Declare.
2824 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2827 (arm_invalid_conversion): New function for target hook.
2828 (arm_invalid_unary_op): New function for target hook.
2829 (arm_invalid_binary_op): New function for target hook.
2831 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2833 * config.gcc: Add arm_bf16.h.
2834 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2835 (arm_simd_builtin_std_type): Add BFmode.
2836 (arm_init_simd_builtin_types): Define element types for vector types.
2837 (arm_init_bf16_types): New function.
2838 (arm_init_builtins): Add arm_init_bf16_types function call.
2839 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2840 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2841 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2842 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2843 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2844 (arm_mangle_type): Add __bf16.
2845 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2846 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2847 arm_bf16_ptr_type_node.
2848 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2849 define_split between ARM registers.
2850 * config/arm/arm_bf16.h: New file.
2851 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2852 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2854 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2855 * config/arm/vfp.md: Add BFmode to movhf patterns.
2857 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2858 Andre Vieira <andre.simoesdiasvieira@arm.com>
2860 * config/arm/arm-cpus.in (mve, mve_float): New features.
2861 (dsp, mve, mve.fp): New options.
2862 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2863 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2864 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2866 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2867 Thomas Preud'homme <thomas.preudhomme@arm.com>
2869 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2871 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2872 error for using -mcmse when targeting Armv8.1-M Mainline.
2874 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2875 Thomas Preud'homme <thomas.preudhomme@arm.com>
2877 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2878 address in r4 when targeting Armv8.1-M Mainline.
2879 (nonsecure_call_value_internal): Likewise.
2880 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2881 a register match_operand again. Emit BLXNS when targeting
2883 (nonsecure_call_value_reg_thumb2): Likewise.
2885 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2886 Thomas Preud'homme <thomas.preudhomme@arm.com>
2888 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2889 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2890 variable as true when floating-point ABI is not hard. Replace
2891 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2892 Generate VLSTM and VLLDM instruction respectively before and
2893 after a function call to cmse_nonsecure_call function.
2894 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2895 (VUNSPEC_VLLDM): Likewise.
2896 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2897 (lazy_load_multiple_insn): Likewise.
2899 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2900 Thomas Preud'homme <thomas.preudhomme@arm.com>
2902 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2903 (arm_emit_vfp_multi_reg_pop): Likewise.
2904 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2905 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2906 restore callee-saved VFP registers.
2908 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2909 Thomas Preud'homme <thomas.preudhomme@arm.com>
2911 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2912 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2913 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2914 callee-saved GPRs as well as clear ip register before doing a nonsecure
2915 call then restore callee-saved GPRs after it when targeting
2917 (arm_reorg): Adapt to function rename.
2919 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2920 Thomas Preud'homme <thomas.preudhomme@arm.com>
2922 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2923 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2924 clear_vfp_multiple pattern based on a new vfp parameter.
2925 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2926 targeting Armv8.1-M Mainline.
2927 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2928 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2929 whether VFP registers are available before looking call_used_regs for a
2931 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2932 of prototype of clear_operation_p.
2933 (clear_vfp_multiple_operation): New predicate.
2934 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2935 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2937 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2938 Thomas Preud'homme <thomas.preudhomme@arm.com>
2940 * config/arm/arm-protos.h (clear_operation_p): Declare.
2941 * config/arm/arm.c (clear_operation_p): New function.
2942 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2943 targeting Armv8.1-M Mainline or successor.
2944 (output_return_instruction): Only output APSR register clearing if
2945 Armv8.1-M Mainline instructions not available.
2946 (thumb_exit): Likewise.
2947 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2948 * config/arm/thumb2.md (clear_apsr): New define_insn.
2949 (clear_multiple): Likewise.
2950 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2952 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2953 Thomas Preud'homme <thomas.preudhomme@arm.com>
2955 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2956 (use_return_insn): Also return false for Armv8.1-M Mainline.
2957 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2958 Mainline instructions are available.
2959 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2960 when targeting Armv8.1-M Mainline Security Extensions.
2961 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2962 Mainline entry function.
2963 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2964 targeting Armv8.1-M Mainline or successor.
2965 (arm_expand_epilogue): Fix indentation of caller-saved register
2966 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2968 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2969 (FP_SYSREGS): Likewise.
2970 (enum vfp_sysregs_encoding): Define enum.
2971 (fp_sysreg_names): Declare.
2972 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2973 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2974 (pop_fpsysreg_insn): Likewise.
2976 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2977 Thomas Preud'homme <thomas.preudhomme@arm.com>
2979 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2980 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2981 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2982 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2983 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2984 (ARMv8_1m_main): New feature group.
2985 (armv8.1-m.main): New architecture.
2986 * config/arm/arm-tables.opt: Regenerate.
2987 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2988 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2989 (arm_options_perform_arch_sanity_checks): Error out when targeting
2990 Armv8.1-M Mainline Security Extensions.
2991 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2993 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2995 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2996 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2997 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2998 aarch64_bfdot_laneq): New.
2999 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
3000 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
3001 vbfdotq_laneq_f32): New.
3002 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
3003 VBFMLA_W, VBF): New.
3004 (isquadop): Add V4BF, V8BF.
3006 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3008 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
3009 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
3010 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
3011 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
3012 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
3013 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
3014 usdot_laneq, sudot_lane,sudot_laneq): New.
3015 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
3016 (aarch64_<sur>dot_lane): New.
3017 * config/aarch64/arm_neon.h (vusdot_s32): New.
3019 (vusdot_lane_s32): New.
3020 (vsudot_lane_s32): New.
3021 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
3022 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
3024 2020-01-16 Martin Liska <mliska@suse.cz>
3026 * value-prof.c (dump_histogram_value): Fix
3027 obvious spacing issue.
3029 2020-01-16 Andrew Pinski <apinski@marvell.com>
3031 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3032 !storage_order_barrier_p.
3034 2020-01-16 Andrew Pinski <apinski@marvell.com>
3036 * sched-int.h (_dep): Add unused bit-field field for the padding.
3037 * sched-deps.c (init_dep_1): Init unused field.
3039 2020-01-16 Andrew Pinski <apinski@marvell.com>
3041 * optabs.h (create_expand_operand): Initialize target field also.
3043 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3045 PR tree-optimization/92429
3046 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3047 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3049 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3052 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3054 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3055 aarch64_sve_int_mode to each mode.
3057 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3059 * doc/analyzer.texi (Overview): Add note about
3060 -fdump-ipa-analyzer.
3062 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3064 PR tree-optimization/93231
3065 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3066 input_type is unsigned. Use tree_to_shwi for shift constant.
3067 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3068 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3069 inputs more efficiently.
3071 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3073 * config/i386/i386.md (*movsf_internal): Do not require
3074 SSE2 ISA for alternatives 14 and 15.
3076 2020-01-15 Richard Biener <rguenther@suse.de>
3079 * tree-eh.c (sink_clobbers): If we already visited the destination
3080 block do not defer insertion.
3081 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3082 the purpose of defered insertion.
3084 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3086 * BASE-VER: Bump to 10.0.1.
3088 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3090 PR tree-optimization/93247
3091 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3092 type of the stmt that we're going to vectorize.
3094 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3096 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3097 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3100 2020-01-15 Martin Liska <mliska@suse.cz>
3102 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3103 2 calls of streamer_read_hwi in a function call.
3105 2020-01-15 Richard Biener <rguenther@suse.de>
3107 * alias.c (record_alias_subset): Avoid redundant work when
3108 subset is already recorded.
3110 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3112 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3113 the analyzer options provide CWE identifiers.
3115 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3117 * tree-diagnostic-path.cc (path_summary::event_range::print):
3118 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3119 using get_pure_location.
3121 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3123 PR tree-optimization/93262
3124 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3125 perform head trimming only if the last argument is constant,
3126 either all ones, or larger or equal to head trim, in the latter
3127 case decrease the last argument by head_trim.
3129 PR tree-optimization/93249
3130 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3131 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3132 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3133 perform head trim unless we can prove there are no '\0' chars
3134 from the source among the first head_trim chars.
3136 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3138 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3140 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3143 * config/i386/sse.md
3144 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3145 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3146 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3147 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3148 just a single alternative instead of two, make operands 1 and 2
3151 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3154 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3157 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3159 * Makefile.in (lang_opt_files): Add analyzer.opt.
3160 (ANALYZER_OBJS): New.
3161 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3162 tristate.o and ANALYZER_OBJS.
3163 (TEXI_GCCINT_FILES): Add analyzer.texi.
3164 * common.opt (-fanalyzer): New driver option.
3165 * config.in: Regenerate.
3166 * configure: Regenerate.
3167 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3168 (gccdepdir): Also create depdir for "analyzer" subdir.
3169 * digraph.cc: New file.
3170 * digraph.h: New file.
3171 * doc/analyzer.texi: New file.
3172 * doc/gccint.texi ("Static Analyzer") New menu item.
3173 (analyzer.texi): Include it.
3174 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3175 ("Warning Options"): Add static analysis warnings to the list.
3176 (-Wno-analyzer-double-fclose): New option.
3177 (-Wno-analyzer-double-free): New option.
3178 (-Wno-analyzer-exposure-through-output-file): New option.
3179 (-Wno-analyzer-file-leak): New option.
3180 (-Wno-analyzer-free-of-non-heap): New option.
3181 (-Wno-analyzer-malloc-leak): New option.
3182 (-Wno-analyzer-possible-null-argument): New option.
3183 (-Wno-analyzer-possible-null-dereference): New option.
3184 (-Wno-analyzer-null-argument): New option.
3185 (-Wno-analyzer-null-dereference): New option.
3186 (-Wno-analyzer-stale-setjmp-buffer): New option.
3187 (-Wno-analyzer-tainted-array-index): New option.
3188 (-Wno-analyzer-use-after-free): New option.
3189 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3190 (-Wno-analyzer-use-of-uninitialized-value): New option.
3191 (-Wanalyzer-too-complex): New option.
3192 (-fanalyzer-call-summaries): New warning.
3193 (-fanalyzer-checker=): New warning.
3194 (-fanalyzer-fine-grained): New warning.
3195 (-fno-analyzer-state-merge): New warning.
3196 (-fno-analyzer-state-purge): New warning.
3197 (-fanalyzer-transitivity): New warning.
3198 (-fanalyzer-verbose-edges): New warning.
3199 (-fanalyzer-verbose-state-changes): New warning.
3200 (-fanalyzer-verbosity=): New warning.
3201 (-fdump-analyzer): New warning.
3202 (-fdump-analyzer-callgraph): New warning.
3203 (-fdump-analyzer-exploded-graph): New warning.
3204 (-fdump-analyzer-exploded-nodes): New warning.
3205 (-fdump-analyzer-exploded-nodes-2): New warning.
3206 (-fdump-analyzer-exploded-nodes-3): New warning.
3207 (-fdump-analyzer-supergraph): New warning.
3208 * doc/sourcebuild.texi (dg-require-dot): New.
3209 (dg-check-dot): New.
3210 * gdbinit.in (break-on-saved-diagnostic): New command.
3211 * graphviz.cc: New file.
3212 * graphviz.h: New file.
3213 * ordered-hash-map-tests.cc: New file.
3214 * ordered-hash-map.h: New file.
3215 * passes.def (pass_analyzer): Add before
3216 pass_ipa_whole_program_visibility.
3217 * selftest-run-tests.c (selftest::run_tests): Call
3218 selftest::ordered_hash_map_tests_cc_tests.
3219 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3221 * shortest-paths.h: New file.
3222 * timevar.def (TV_ANALYZER): New timevar.
3223 (TV_ANALYZER_SUPERGRAPH): Likewise.
3224 (TV_ANALYZER_STATE_PURGE): Likewise.
3225 (TV_ANALYZER_PLAN): Likewise.
3226 (TV_ANALYZER_SCC): Likewise.
3227 (TV_ANALYZER_WORKLIST): Likewise.
3228 (TV_ANALYZER_DUMP): Likewise.
3229 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3230 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3231 * tree-pass.h (make_pass_analyzer): New decl.
3232 * tristate.cc: New file.
3233 * tristate.h: New file.
3235 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3238 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3239 alternatives 9 and 10.
3241 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3243 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3244 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3245 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3246 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3247 (selftest::hash_map_tests_c_tests): Call it.
3248 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3249 New static constant, using the value of = H::empty_zero_p.
3250 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3251 from default_hash_traits <Value>.
3252 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3254 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3255 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3256 calls to mark_empty with !Descriptor::empty_zero_p.
3257 (hash_table::empty_slow): Conditionalize the memset call with a
3258 check that Descriptor::empty_zero_p; otherwise, loop through the
3259 entries calling mark_empty on them.
3260 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3261 (pointer_hash::empty_zero_p): Likewise.
3262 (pair_hash::empty_zero_p): Likewise.
3263 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3265 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3266 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3267 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3268 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3269 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3270 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3271 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3272 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3274 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3277 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3279 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3280 fix typo on return value.
3282 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3285 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3287 (cgraph_edge::make_speculative): Add param for setting speculative_id
3289 (cgraph_edge::speculative_call_info): Update comments and find reference
3290 by speculative_id for multiple indirect targets.
3291 (cgraph_edge::resolve_speculation): Decrease the speculations
3292 for indirect edge, drop it's speculative if not direct target
3293 left. Update comments.
3294 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3295 (cgraph_node::dump): Print num_speculative_call_targets.
3296 (cgraph_node::verify_node): Don't report error if speculative
3297 edge not include statement.
3298 (cgraph_edge::num_speculative_call_targets_p): New function.
3299 * cgraph.h (int common_target_id): Remove.
3300 (int common_target_probability): Remove.
3301 (num_speculative_call_targets): New variable.
3302 (make_speculative): Add param for setting speculative_id.
3303 (cgraph_edge::num_speculative_call_targets_p): New declare.
3304 (target_prob): New variable.
3305 (speculative_id): New variable.
3306 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3307 call summaries for multiple speculative call targets.
3308 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3309 * ipa-profile.c (struct speculative_call_target): New struct.
3310 (class speculative_call_summary): New class.
3311 (class speculative_call_summaries): New class.
3312 (call_sums): New variable.
3313 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3314 (ipa_profile_write_edge_summary): New function.
3315 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3316 (ipa_profile_dump_all_summaries): New function.
3317 (ipa_profile_read_edge_summary): New function.
3318 (ipa_profile_read_summary_section): New function.
3319 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3320 (ipa_profile): Generate num_speculative_call_targets from
3322 * ipa-ref.h (speculative_id): New variable.
3323 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3324 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3325 common_target_probability. Stream out speculative_id and
3326 num_speculative_call_targets.
3327 (input_edge): Likewise.
3328 * predict.c (dump_prediction): Remove edges count assert to be
3330 * symtab.c (symtab_node::create_reference): Init speculative_id.
3331 (symtab_node::clone_references): Clone speculative_id.
3332 (symtab_node::clone_referring): Clone speculative_id.
3333 (symtab_node::clone_reference): Clone speculative_id.
3334 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3335 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3336 if indirect call contains multiple speculative targets.
3337 * value-prof.h (check_ic_target): Remove.
3338 * value-prof.c (gimple_value_profile_transformations):
3339 Use void function gimple_ic_transform.
3340 * value-prof.c (gimple_ic_transform): Handle topn case.
3341 Fix comment typos. Change it to a void function.
3343 2020-01-13 Andrew Pinski <apinski@marvell.com>
3345 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3346 (octeontx2t98): New define.
3347 (octeontx2t96): New define.
3348 (octeontx2t93): New define.
3349 (octeontx2f95): New define.
3350 (octeontx2f95n): New define.
3351 (octeontx2f95mm): New define.
3352 * config/aarch64/aarch64-tune.md: Regenerate.
3353 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3355 2020-01-13 Jason Merrill <jason@redhat.com>
3357 PR c++/33799 - destroy return value if local cleanup throws.
3358 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3360 2020-01-13 Martin Liska <mliska@suse.cz>
3362 * ipa-cp.c (get_max_overall_size): Use newly
3363 renamed param param_ipa_cp_unit_growth.
3364 * params.opt: Remove legacy param name.
3366 2020-01-13 Martin Sebor <msebor@redhat.com>
3368 PR tree-optimization/93213
3369 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3370 stores to be eliminated.
3372 2020-01-13 Martin Liska <mliska@suse.cz>
3374 * opts.c (print_help): Do not print CL_PARAM
3375 and CL_WARNING for CL_OPTIMIZATION.
3377 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3380 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3381 depending on optimization settings.
3383 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3385 PR tree-optimization/90838
3386 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3387 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3388 argument rather than to initialize temporary for targets that
3389 don't use the mode argument at all. Initialize ctzval to avoid
3392 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3394 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3395 * tree-core.h: Document it.
3396 * gimplify.c (gimplify_omp_workshare): Set it.
3397 * omp-low.c (lower_omp_target): Use it.
3398 * tree-pretty-print.c (dump_omp_clause): Print it.
3400 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3401 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3403 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3405 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3406 * common.opt (fdiagnostics-path-format=): New option.
3407 (diagnostic_path_format): New enum.
3408 (fdiagnostics-show-path-depths): New option.
3409 * coretypes.h (diagnostic_event_id_t): New forward decl.
3410 * diagnostic-color.c (color_dict): Add "path".
3411 * diagnostic-event-id.h: New file.
3412 * diagnostic-format-json.cc (json_from_expanded_location): Make
3414 (json_end_diagnostic): Call context->make_json_for_path if it
3415 exists and the diagnostic has a path.
3416 (diagnostic_output_format_init): Clear context->print_path.
3417 * diagnostic-path.h: New file.
3418 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3419 when printing a run of events in a diagnostic_path so that they
3420 all get the same color.
3421 (layout::m_diagnostic_path_p): New field.
3422 (layout::layout): Initialize it.
3423 (layout::print_any_labels): Don't colorize the label text for an
3424 event in a diagnostic_path.
3425 (gcc_rich_location::add_location_if_nearby): Add
3426 "restrict_to_current_line_spans" and "label" params. Pass the
3427 former to layout.maybe_add_location_range; pass the latter
3428 when calling add_range.
3429 * diagnostic.c: Include "diagnostic-path.h".
3430 (diagnostic_initialize): Initialize context->path_format and
3431 context->show_path_depths.
3432 (diagnostic_show_any_path): New function.
3433 (diagnostic_path::interprocedural_p): New function.
3434 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3435 (simple_diagnostic_path::num_events): New function.
3436 (simple_diagnostic_path::get_event): New function.
3437 (simple_diagnostic_path::add_event): New function.
3438 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3439 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3440 (debug): New overload taking a diagnostic_path *.
3441 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3442 * diagnostic.h (enum diagnostic_path_format): New enum.
3443 (json::value): New forward decl.
3444 (diagnostic_context::path_format): New field.
3445 (diagnostic_context::show_path_depths): New field.
3446 (diagnostic_context::print_path): New callback field.
3447 (diagnostic_context::make_json_for_path): New callback field.
3448 (diagnostic_show_any_path): New decl.
3449 (json_from_expanded_location): New decl.
3450 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3451 (-fdiagnostics-show-path-depths): New option.
3452 (-fdiagnostics-color): Add "path" to description of default
3453 GCC_COLORS; describe it.
3454 (-fdiagnostics-format=json): Document how diagnostic paths are
3455 represented in the JSON output format.
3456 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3457 Add optional params "restrict_to_current_line_spans" and "label".
3458 * opts.c (common_handle_option): Handle
3459 OPT_fdiagnostics_path_format_ and
3460 OPT_fdiagnostics_show_path_depths.
3461 * pretty-print.c: Include "diagnostic-event-id.h".
3462 (pp_format): Implement "%@" format code for printing
3463 diagnostic_event_id_t *.
3464 (selftest::test_pp_format): Add tests for "%@".
3465 * selftest-run-tests.c (selftest::run_tests): Call
3466 selftest::tree_diagnostic_path_cc_tests.
3467 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3468 * toplev.c (general_init): Initialize global_dc->path_format and
3469 global_dc->show_path_depths.
3470 * tree-diagnostic-path.cc: New file.
3471 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3472 non-static. Drop "diagnostic" param in favor of storing the
3473 original value of "where" and re-using it.
3474 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3475 maybe_unwind_expanded_macro_loc.
3476 (tree_diagnostics_defaults): Initialize context->print_path and
3477 context->make_json_for_path.
3478 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3480 (default_tree_make_json_for_path): New decl.
3481 (maybe_unwind_expanded_macro_loc): New decl.
3483 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3485 PR tree-optimization/93210
3486 * fold-const.h (native_encode_initializer,
3487 can_native_interpret_type_p): Declare.
3488 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3490 (native_encode_initializer): New function, moved from dwarf2out.c.
3491 Adjust to native_encode_expr compatible arguments, including dry-run
3492 and partial extraction modes. Don't handle STRING_CST.
3493 (can_native_interpret_type_p): No longer static.
3494 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3495 offset / BITS_PER_UNIT fits into int and don't call it if
3496 can_native_interpret_type_p fails. If suboff is NULL and for
3497 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3498 native_encode_initializer.
3499 (fold_const_aggregate_ref_1): Formatting fix.
3500 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3501 (tree_add_const_value_attribute): Adjust caller.
3503 PR tree-optimization/90838
3504 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3505 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3506 CTZ_DEFINED_VALUE_AT_ZERO.
3508 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3511 * lra-constraints.c (match_reload): Permit input operands have the
3512 same mode as output while other input operands have a different
3515 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3517 PR tree-optimization/90838
3518 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3519 (check_ctz_string): Likewise.
3520 (optimize_count_trailing_zeroes): Likewise.
3521 (simplify_count_trailing_zeroes): Likewise.
3522 (pass_forwprop::execute): Try ctz simplification.
3523 * match.pd: Add matching for ctz idioms.
3525 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3527 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3529 (aarch64_invalid_unary_op): New function for target hook.
3530 (aarch64_invalid_binary_op): New function for target hook.
3532 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3534 * config.gcc: Add arm_bf16.h.
3535 * config/aarch64/aarch64-builtins.c
3536 (aarch64_simd_builtin_std_type): Add BFmode.
3537 (aarch64_init_simd_builtin_types): Define element types for vector
3539 (aarch64_init_bf16_types): New function.
3540 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3541 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3543 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3544 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3546 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3547 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3548 * config/aarch64/aarch64.c
3549 (aarch64_classify_vector_mode): Add support for BF types.
3550 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3551 (aarch64_vq_mode): Add support for BF types.
3552 (aarch64_simd_container_mode): Add support for BF types.
3553 (aarch64_mangle_type): Add support for BF scalar type.
3554 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3555 * config/aarch64/arm_bf16.h: New file.
3556 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3557 * config/aarch64/iterators.md: Add BF types to mode attributes.
3558 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3560 2020-01-10 Jason Merrill <jason@redhat.com>
3562 PR c++/93173 - incorrect tree sharing.
3563 * gimplify.c (copy_if_shared): No longer static.
3564 * gimplify.h: Declare it.
3566 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3568 * doc/invoke.texi (-msve-vector-bits=): Document that
3569 -msve-vector-bits=128 now generates VL-specific code for
3570 little-endian targets.
3571 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3572 build_vector_type_for_mode to construct the data vector types.
3573 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3574 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3575 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3576 for 128-bit vectors.
3578 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3580 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3583 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3585 * config/aarch64/aarch64-builtins.c
3586 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3587 rather than checking the number of elements and the element mode.
3589 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3591 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3592 get_related_vectype_for_scalar_type rather than build_vector_type
3593 to create the index type for a conditional reduction.
3595 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3597 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3598 for any type of gather or scatter, including strided accesses.
3600 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3602 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3605 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3607 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3609 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3610 parameter and its use to reset DR_OFFSET's.
3611 (vect_transform_loop): Remove orig_drs_init argument.
3612 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3613 member of dr_vec_info rather than the offset of the associated
3614 data_reference's innermost_loop_behavior.
3615 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3616 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3617 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3618 get_dr_vinfo_offset.
3619 (vectorizable_store): Likewise.
3620 (vectorizable_load): Likewise.
3622 2020-01-10 Richard Biener <rguenther@suse.de>
3624 * gimple-ssa-store-merging
3625 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3627 2020-01-10 Martin Liska <mliska@suse.cz>
3630 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3631 encapsulation that was there before r280040.
3633 2020-01-10 Richard Biener <rguenther@suse.de>
3636 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3637 sequences to avoid walking them again for secondary opportunities.
3638 (pass_lower_eh_dispatch::execute): Instead actually insert
3641 2020-01-10 Richard Biener <rguenther@suse.de>
3644 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3645 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3646 avoid quadraticness.
3648 2020-01-10 Martin Jambor <mjambor@suse.cz>
3650 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3651 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3652 to get param_ipa_sra_max_replacements.
3653 (param_splitting_across_edge): Pass the caller to
3654 pull_accesses_from_callee.
3656 2020-01-10 Martin Jambor <mjambor@suse.cz>
3658 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3659 * ipa-cp.c (max_new_size): Removed.
3660 (orig_overall_size): New variable.
3661 (get_max_overall_size): New function.
3662 (estimate_local_effects): Use it. Adjust dump.
3663 (decide_about_value): Likewise.
3664 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3665 orig_overall_size. Adjust dump.
3666 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3668 2020-01-10 Martin Jambor <mjambor@suse.cz>
3670 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3671 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3672 instead of param_ipa_max_agg_items.
3673 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3674 optimization info for the callee.
3676 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3678 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3679 markers if debug_inline_points is false.
3681 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3683 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3685 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3686 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3687 aarch64-sve-builtins-sve2.h.
3688 (aarch64-sve-builtins-sve2.o): New rule.
3689 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3690 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3691 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3692 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3693 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3694 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3696 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3697 instructions that are handled here.
3698 (@cond_asrd<mode>): Generalize to...
3699 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3700 (*cond_asrd<mode>_2): Generalize to...
3701 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3702 (*cond_asrd<mode>_z): Generalize to...
3703 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3704 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3705 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3706 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3707 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3709 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3710 (@aarch64_scatter_stnt<mode>): Likewise.
3711 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3712 (@aarch64_mul_lane_<mode>): Likewise.
3713 (@aarch64_sve_suqadd<mode>_const): Likewise.
3714 (*<sur>h<addsub><mode>): Generalize to...
3715 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3717 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3718 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3719 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3720 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3721 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3722 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3723 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3724 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3725 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3726 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3727 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3728 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3729 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3730 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3731 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3732 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3733 (@aarch64_sve2_xar<mode>): Likewise.
3734 (@aarch64_sve2_bcax<mode>): Likewise.
3735 (*aarch64_sve2_eor3<mode>): Rename to...
3736 (@aarch64_sve2_eor3<mode>): ...this.
3737 (@aarch64_sve2_bsl<mode>): New expander.
3738 (@aarch64_sve2_nbsl<mode>): Likewise.
3739 (@aarch64_sve2_bsl1n<mode>): Likewise.
3740 (@aarch64_sve2_bsl2n<mode>): Likewise.
3741 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3742 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3743 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3744 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3745 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3746 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3747 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3748 (<su>mull<bt><Vwide>): Generalize to...
3749 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3751 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3752 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3753 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3754 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3755 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3756 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3757 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3758 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3759 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3760 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3761 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3762 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3763 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3764 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3765 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3766 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3767 (<SHRNB:r>shrnb<mode>): Generalize to...
3768 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3770 (<SHRNT:r>shrnt<mode>): Generalize to...
3771 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3773 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3774 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3775 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3776 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3777 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3778 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3779 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3780 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3781 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3782 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3783 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3784 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3785 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3786 (@aarch64_sve2_cvtnt<mode>): Likewise.
3787 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3788 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3789 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3790 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3791 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3792 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3793 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3794 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3795 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3796 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3797 (@aarch64_sve2_pmul<mode>): Likewise.
3798 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3799 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3800 (@aarch64_sve2_tbl2<mode>): Likewise.
3801 (@aarch64_sve2_tbx<mode>): Likewise.
3802 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3803 (@aarch64_sve2_histcnt<mode>): Likewise.
3804 (@aarch64_sve2_histseg<mode>): Likewise.
3805 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3806 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3807 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3808 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3809 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3810 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3811 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3812 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3813 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3814 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3815 (SVE2_PMULL_PAIR_I): New mode iterators.
3816 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3817 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3818 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3819 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3820 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3821 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3822 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3823 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3824 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3825 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3826 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3827 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3828 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3829 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3830 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3831 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3832 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3833 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3834 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3835 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3836 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3837 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3838 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3839 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3840 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3841 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3842 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3843 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3844 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3845 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3846 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3848 (VNARROW, Ventype): New mode attributes.
3849 (Vewtype): Handle VNx2DI. Fix typo in comment.
3850 (VDOUBLE): New mode attribute.
3851 (sve_lane_con): Handle VNx8HI.
3852 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3853 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3854 (sve_int_op, sve_int_op_rev): Handle the above codes.
3855 (sve_pred_int_rhs2_operand): Likewise.
3856 (MULLBT, SHRNB, SHRNT): Delete.
3857 (SVE_INT_SHIFT_IMM): New int iterator.
3858 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3859 and UNSPEC_WHILEHS for TARGET_SVE2.
3860 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3861 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3862 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3863 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3864 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3865 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3866 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3867 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3868 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3869 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3870 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3871 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3872 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3873 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3874 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3875 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3876 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3877 (optab): Handle the new unspecs.
3878 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3880 (lr): Handle the new unspecs.
3882 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3883 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3884 (sve_int_qsub_op): New int attributes.
3885 (sve_fp_op, rot): Handle the new unspecs.
3886 * config/aarch64/aarch64-sve-builtins.h
3887 (function_resolver::require_matching_pointer_type): Declare.
3888 (function_resolver::resolve_unary): Add an optional boolean argument.
3889 (function_resolver::finish_opt_n_resolution): Add an optional
3890 type_suffix_index argument.
3891 (gimple_folder::redirect_call): Declare.
3892 (gimple_expander::prepare_gather_address_operands): Add an optional
3894 * config/aarch64/aarch64-sve-builtins.cc: Include
3895 aarch64-sve-builtins-sve2.h.
3896 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3897 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3898 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3899 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3900 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3901 (TYPES_s_unsigned): Likewise.
3902 (TYPES_s_integer): Use TYPES_s_unsigned.
3903 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3904 (TYPES_sd_integer): Use them.
3905 (TYPES_d_unsigned): New macro.
3906 (TYPES_d_integer): Use it.
3907 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3908 (TYPES_cvt_narrow): Likewise.
3909 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3910 (preds_mx): New variable.
3911 (function_builder::add_overloaded_function): Allow the new feature
3912 set to be more restrictive than the original one.
3913 (function_resolver::infer_pointer_type): Remove qualifiers from
3914 the pointer type before printing it.
3915 (function_resolver::require_matching_pointer_type): New function.
3916 (function_resolver::resolve_sv_displacement): Handle functions
3917 that don't support 32-bit vector indices or svint32_t vector offsets.
3918 (function_resolver::finish_opt_n_resolution): Take the inferred type
3919 as a separate argument.
3920 (function_resolver::resolve_unary): Optionally treat all forms in
3921 the same way as normal merging functions.
3922 (gimple_folder::redirect_call): New function.
3923 (function_expander::prepare_gather_address_operands): Add an argument
3924 that says whether scaled forms are available. If they aren't,
3925 handle scaling of vector indices and don't add the extension and
3927 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3928 fall back to using cond_* instead.
3929 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3930 Split out the member variables into...
3931 (rtx_code_function_base): ...this new base class.
3932 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3933 (unspec_based_function): Split out the member variables into...
3934 (unspec_based_function_base): ...this new base class.
3935 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3936 (unspec_based_function_exact_insn): New class.
3937 (unspec_based_add_function, unspec_based_add_lane_function)
3938 (unspec_based_lane_function, unspec_based_pred_function)
3939 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3940 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3941 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3943 (unspec_based_fused_function): New class.
3944 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3945 (unspec_based_fused_lane_function): New class.
3946 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3948 (CODE_FOR_MODE1): New macro.
3949 (fixed_insn_function): New class.
3950 (while_comparison): Likewise.
3951 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3952 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3953 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3954 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3955 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3956 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3957 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3958 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3959 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3960 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3961 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3962 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3963 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3964 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3965 Also add an initial argument for unary_convert_narrowt, regardless
3966 of the predication type.
3967 (build_32_64): Allow loads and stores to specify MODE_none.
3968 (build_sv_index64, build_sv_uint_offset): New functions.
3969 (long_type_suffix): New function.
3970 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3971 (binary_imm_long_base, load_gather_sv_base): Likewise.
3972 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3973 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3974 (unary_narrowb_base, unary_narrowt_base): Likewise.
3975 (binary_long_lane_def, binary_long_lane): New shape.
3976 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3977 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3978 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3979 (binary_to_uint_def, binary_to_uint): Likewise.
3980 (binary_wide_def, binary_wide): Likewise.
3981 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3982 (compare_def, compare): Likewise.
3983 (compare_ptr_def, compare_ptr): Likewise.
3984 (load_ext_gather_index_restricted_def,
3985 load_ext_gather_index_restricted): Likewise.
3986 (load_ext_gather_offset_restricted_def,
3987 load_ext_gather_offset_restricted): Likewise.
3988 (load_gather_sv_def): Inherit from load_gather_sv_base.
3989 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3990 (shift_left_imm_def, shift_left_imm): Likewise.
3991 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3992 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3993 (store_scatter_index_restricted_def,
3994 store_scatter_index_restricted): Likewise.
3995 (store_scatter_offset_restricted_def,
3996 store_scatter_offset_restricted): Likewise.
3997 (tbl_tuple_def, tbl_tuple): Likewise.
3998 (ternary_long_lane_def, ternary_long_lane): Likewise.
3999 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
4000 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
4001 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
4002 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
4003 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
4004 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
4005 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
4006 (ternary_uint_def, ternary_uint): Likewise.
4007 (unary_convert): Fix typo in comment.
4008 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
4009 (unary_long_def, unary_long): Likewise.
4010 (unary_narrowb_def, unary_narrowb): Likewise.
4011 (unary_narrowt_def, unary_narrowt): Likewise.
4012 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
4013 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
4014 (unary_to_int_def, unary_to_int): Likewise.
4015 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
4016 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
4017 (svasrd_impl): Delete.
4018 (svcadd_impl::expand): Handle integer operations too.
4019 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
4020 new functions to derive the unspec numbers.
4021 (svmla_svmls_lane_impl): Replace with...
4022 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
4023 integer operations too.
4024 (svwhile_impl): Rename to...
4025 (svwhilelx_impl): ...this and inherit from while_comparison.
4026 (svasrd): Use unspec_based_function.
4027 (svmla_lane): Use svmla_lane_impl.
4028 (svmls_lane): Use svmls_lane_impl.
4029 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4030 (svwhilele, svwhilelt): Use svwhilelx_impl.
4031 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4032 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4033 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4034 * config/aarch64/aarch64-sve-builtins.def: Include
4035 aarch64-sve-builtins-sve2.def.
4037 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4039 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4040 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4041 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4042 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4043 immediates as well as vector ones.
4044 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4045 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4046 (aarch64_sve_qsub_immediate): Update calls accordingly.
4048 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4050 * config/aarch64/aarch64-sve2.md: Add banner comments.
4051 (<su>mulh<r>s<mode>3): Move further up file.
4052 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4053 (*aarch64_sve2_sra<mode>): Move further down file.
4054 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4056 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4058 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4060 (while_optab_cmp): Handle them.
4061 * config/aarch64/aarch64-sve.md
4062 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4063 and add a "@" marker.
4064 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4065 instead of gen_aarch64_sve2_while_ptest.
4066 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4068 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4070 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4071 (UNSPEC_WHILELE): ...this.
4072 (UNSPEC_WHILE_LO): Rename to...
4073 (UNSPEC_WHILELO): ...this.
4074 (UNSPEC_WHILE_LS): Rename to...
4075 (UNSPEC_WHILELS): ...this.
4076 (UNSPEC_WHILE_LT): Rename to...
4077 (UNSPEC_WHILELT): ...this.
4078 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4079 (cmp_op, while_optab_cmp): Likewise.
4080 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4081 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4082 (svwhilelt): Likewise.
4084 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4086 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4087 (unary_to_uint): Define.
4088 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4089 (unary_count): Rename to...
4090 (unary_to_uint_def, unary_to_uint): ...this.
4091 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4093 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4095 * config/aarch64/aarch64-sve-builtins-functions.h
4096 (code_for_mode_function): New class.
4097 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4098 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4099 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4100 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4101 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4103 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4105 * config/aarch64/iterators.md (addsub): New code attribute.
4106 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4108 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4109 in the asm string and attributes. Fix indentation.
4110 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4112 (@aarch64_sve_<optab><mode>): ...this.
4113 * config/aarch64/aarch64-sve-builtins.h
4114 (function_expander::expand_signed_unpred_op): Delete.
4115 * config/aarch64/aarch64-sve-builtins.cc
4116 (function_expander::expand_signed_unpred_op): Likewise.
4117 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4118 try using code_for_aarch64_sve instead.
4119 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4120 (svqsub_impl): Likewise.
4121 (svqadd, svqsub): Use rtx_code_function instead.
4123 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4125 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4126 (HADDSUB, sur, addsub): Remove them.
4128 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4130 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4131 functions the same way, rather than singling out those that
4132 aren't mapped directly to optabs.
4134 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4136 * target.def (compatible_vector_types_p): New target hook.
4137 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4138 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4139 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4140 * doc/tm.texi: Regenerate.
4141 * gimple-expr.c: Include target.h.
4142 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4143 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4145 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4146 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4147 Use the original predicate if it already has a suitable type.
4149 2020-01-09 Martin Jambor <mjambor@suse.cz>
4151 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4152 resolve_speculation and redirect_call_stmt_to_callee static. Change
4153 return type of set_call_stmt to cgraph_edge *.
4154 * auto-profile.c (afdo_indirect_call): Adjust call to
4155 redirect_call_stmt_to_callee.
4156 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4157 make the this pointer explicit, adjust self-recursive calls and the
4158 call top make_direct. Return the resulting edge.
4159 (cgraph_edge::remove): Make this pointer explicit.
4160 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4161 (cgraph_edge::make_direct): Likewise, adjust call to
4162 resolve_speculation.
4163 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4164 call to set_call_stmt.
4165 (cgraph_update_edges_for_call_stmt_node): Update call to
4166 set_call_stmt and remove.
4167 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4168 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4169 (cgraph_node::create_edge_including_clones): Moved "first" definition
4170 of edge to the block where it was used. Adjusted calls to
4172 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4173 cgraph_edge::remove.
4174 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4175 make_direct and redirect_call_stmt_to_callee.
4176 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4177 resolve_speculation and make_direct.
4178 * ipa-inline-transform.c (inline_transform): Adjust call to
4179 redirect_call_stmt_to_callee.
4180 (check_speculations_1):: Adjust call to resolve_speculation.
4181 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4182 resolve-speculation.
4183 (inline_small_functions): Adjust call to resolve_speculation.
4184 (ipa_inline): Likewise.
4185 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4187 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4188 safe with regards to edge removal, adjust calls to
4189 redirect_call_stmt_to_callee.
4190 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4191 and redirect_call_stmt_to_callee.
4192 * multiple_target.c (create_dispatcher_calls): Adjust call to
4193 redirect_call_stmt_to_callee
4194 (redirect_to_specific_clone): Likewise.
4195 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4196 Adjust calls to cgraph_edge::remove.
4197 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4198 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4199 (expand_call_inline): Adjust call to cgraph_edge::remove.
4201 2020-01-09 Martin Liska <mliska@suse.cz>
4203 * params.opt: Set Optimization for
4204 param_max_speculative_devirt_maydefs.
4206 2020-01-09 Martin Sebor <msebor@redhat.com>
4210 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4212 2020-01-09 Martin Liska <mliska@suse.cz>
4214 * auto-profile.c (auto_profile): Use opt_for_fn
4216 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4217 (propagate_vals_across_arith_jfunc): Likewise.
4218 (hint_time_bonus): Likewise.
4219 (incorporate_penalties): Likewise.
4220 (good_cloning_opportunity_p): Likewise.
4221 (perform_estimation_of_a_value): Likewise.
4222 (estimate_local_effects): Likewise.
4223 (ipcp_propagate_stage): Likewise.
4224 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4225 (set_switch_stmt_execution_predicate): Likewise.
4226 (analyze_function_body): Likewise.
4227 * ipa-inline-analysis.c (offline_size): Likewise.
4228 * ipa-inline.c (early_inliner): Likewise.
4229 * ipa-prop.c (ipa_analyze_node): Likewise.
4230 (ipcp_transform_function): Likewise.
4231 * ipa-sra.c (process_scan_results): Likewise.
4232 (ipa_sra_summarize_function): Likewise.
4233 * params.opt: Rename ipcp-unit-growth to
4234 ipa-cp-unit-growth. Add Optimization for various
4235 IPA-related parameters.
4237 2020-01-09 Richard Biener <rguenther@suse.de>
4240 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4242 2020-01-09 Richard Biener <rguenther@suse.de>
4244 PR tree-optimization/93040
4245 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4247 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4249 * common/config/avr/avr-common.c (avr_option_optimization_table)
4250 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4252 2020-01-09 Martin Liska <mliska@suse.cz>
4254 * cgraphclones.c (symbol_table::materialize_all_clones):
4255 Use cgraph_node::dump_name.
4257 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4260 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4261 output_operand_lossage instead of gcc_unreachable.
4262 * doc/md.texi (riscv f constraint): Fix typo.
4265 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4266 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4267 CONST_SCALAR_INT_P instead of CONST_INT_P.
4268 (*subv<mode>4_1): Rename to ...
4269 (subv<mode>4_1): ... this.
4270 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4271 define_insn_and_split patterns.
4272 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4275 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4277 * vec.c (class selftest::count_dtor): New class.
4278 (selftest::test_auto_delete_vec): New test.
4279 (selftest::vec_c_tests): Call it.
4280 * vec.h (class auto_delete_vec): New class template.
4281 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4283 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4285 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4287 2020-01-08 Jim Wilson <jimw@sifive.com>
4289 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4290 use of TLS_MODEL_LOCAL_EXEC when not pic.
4292 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4294 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4297 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4300 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4301 *stack_protect_set_3 peephole2): Also check that the second
4302 insns source is general_operand.
4305 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4306 predicate for output operand instead of register_operand.
4307 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4308 memory destination and non-memory operands[2].
4310 2020-01-08 Martin Liska <mliska@suse.cz>
4312 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4313 ::dump_asm_name instead of (::name or ::asm_name).
4314 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4315 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4316 (analyze_functions): Likewise.
4317 (expand_all_functions): Likewise.
4318 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4319 (propagate_bits_across_jump_function): Likewise.
4320 (dump_profile_updates): Likewise.
4321 (ipcp_store_bits_results): Likewise.
4322 (ipcp_store_vr_results): Likewise.
4323 * ipa-devirt.c (dump_targets): Likewise.
4324 * ipa-fnsummary.c (analyze_function_body): Likewise.
4325 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4326 (process_hsa_functions): Likewise.
4327 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4328 (set_alias_uids): Likewise.
4329 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4330 * ipa-inline.c (recursive_inlining): Likewise.
4331 (inline_to_all_callers_1): Likewise.
4332 (ipa_inline): Likewise.
4333 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4334 (ipa_propagate_frequency): Likewise.
4335 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4336 (remove_described_reference): Likewise.
4337 * ipa-pure-const.c (worse_state): Likewise.
4338 (check_retval_uses): Likewise.
4339 (analyze_function): Likewise.
4340 (propagate_pure_const): Likewise.
4341 (propagate_nothrow): Likewise.
4342 (dump_malloc_lattice): Likewise.
4343 (propagate_malloc): Likewise.
4344 (pass_local_pure_const::execute): Likewise.
4345 * ipa-visibility.c (optimize_weakref): Likewise.
4346 (function_and_variable_visibility): Likewise.
4347 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4348 (ipa_discover_variable_flags): Likewise.
4349 * lto-streamer-out.c (output_function): Likewise.
4350 (output_constructor): Likewise.
4351 * tree-inline.c (copy_bb): Likewise.
4352 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4353 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4355 2020-01-08 Richard Biener <rguenther@suse.de>
4358 * tree-eh.c (sink_clobbers): Update virtual operands for
4359 the first and last stmt only. Add a dry-run capability.
4360 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4361 after CFG manipulations and in RPO order to catch all
4362 secondary opportunities reliably.
4364 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4367 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4369 2019-01-08 Richard Biener <rguenther@suse.de>
4372 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4373 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4374 virtual operand, also updating SSA use.
4375 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4376 Update stmt after resetting virtual operand.
4377 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4378 * gimple-iterator.c (gsi_remove): When not removing the stmt
4379 permanently do not delink immediate uses or mark the stmt modified.
4381 2020-01-08 Martin Liska <mliska@suse.cz>
4383 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4384 (ipa_call_context::estimate_size_and_time): Likewise.
4385 (inline_analyze_function): Likewise.
4387 2020-01-08 Martin Liska <mliska@suse.cz>
4389 * cgraph.c (cgraph_node::dump): Use systematically
4392 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4394 Add -nodevicespecs option for avr.
4397 * config/avr/avr.opt (-nodevicespecs): New driver option.
4398 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4399 "-specs=device-specs/..." if that option is not set.
4400 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4402 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4404 Implement 64-bit double functions for avr.
4407 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4408 --with-double-comparison.
4409 * doc/install.texi: Document them.
4410 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4411 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4412 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4413 * doc/invoke.texi (AVR Built-in Macros): Document them.
4414 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4415 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4416 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4418 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4421 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4422 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4423 when only building rm-profile multilibs.
4425 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4428 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4429 lattice for a value to check.
4430 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4431 finite propagation in self-recursive scc.
4433 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4435 * ipa-inline.c (caller_growth_limits): Restore the AND.
4437 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4439 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4440 (VEC_ALLREG_ALT): New iterator.
4441 (VEC_ALLREG_INT_MODE): New iterator.
4442 (VCMP_MODE): New iterator.
4443 (VCMP_MODE_INT): New iterator.
4444 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4445 (vec_cmp<u>v64qidi): New define_expand.
4446 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4447 (vec_cmpu<mode>di_exec): New define_expand.
4448 (vec_cmp<u>v64qidi_exec): New define_expand.
4449 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4450 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4451 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4452 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4453 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4454 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4455 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4456 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4457 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4458 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4460 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4461 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4463 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4465 * config/gcn/constraints.md (DA): Update description and match.
4467 (Db): New constraint.
4468 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4470 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4471 Implement 'Db' mixed immediate type.
4472 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4473 (addcv64si3_dup<exec_vcc>): Delete.
4474 (subcv64si3<exec_vcc>): Rework constraints.
4475 (addv64di3): Rework constraints.
4476 (addv64di3_exec): Rework constraints.
4477 (subv64di3): Rework constraints.
4478 (addv64di3_dup): Delete.
4479 (addv64di3_dup_exec): Delete.
4480 (addv64di3_zext): Rework constraints.
4481 (addv64di3_zext_exec): Rework constraints.
4482 (addv64di3_zext_dup): Rework constraints.
4483 (addv64di3_zext_dup_exec): Rework constraints.
4484 (addv64di3_zext_dup2): Rework constraints.
4485 (addv64di3_zext_dup2_exec): Rework constraints.
4486 (addv64di3_sext_dup2): Rework constraints.
4487 (addv64di3_sext_dup2_exec): Rework constraints.
4489 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4491 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4492 existing target checks.
4494 2020-01-07 Richard Biener <rguenther@suse.de>
4496 * doc/install.texi: Bump minimal supported MPC version.
4498 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4500 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4501 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4502 * langhooks.c: Include stor-layout.h.
4503 (lhd_simulate_enum_decl): New function.
4504 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4505 handle_arm_sve_h for the LTO frontend.
4506 (register_vector_type): Cope with null returns from pushdecl.
4508 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4510 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4511 (aarch64_sve::nvectors_if_data_type): Replace with...
4512 (aarch64_sve::builtin_type_p): ...this.
4513 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4514 (find_vector_type): Delete.
4515 (add_sve_type_attribute): New function.
4516 (lookup_sve_type_attribute): Likewise.
4517 (register_builtin_types): Add an "SVE type" attribute to each type.
4518 (register_tuple_type): Likewise.
4519 (svbool_type_p, nvectors_if_data_type): Delete.
4520 (mangle_builtin_type): Use lookup_sve_type_attribute.
4521 (builtin_type_p): Likewise. Add an overload that returns the
4522 number of constituent vector and predicate registers.
4523 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4524 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4525 instead of aarch64_sve_argument_p.
4526 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4527 (aarch64_pass_by_reference): Likewise.
4528 (aarch64_function_value_1): Likewise.
4529 (aarch64_return_in_memory): Likewise.
4530 (aarch64_layout_arg): Likewise.
4532 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4534 PR tree-optimization/93156
4535 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4536 least significant bit is always clear.
4538 PR tree-optimization/93118
4539 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4540 simplifier with two intermediate conversions.
4542 2020-01-07 Martin Liska <mliska@suse.cz>
4544 * params.opt: Add Optimization for various parameters.
4546 2020-01-07 Martin Liska <mliska@suse.cz>
4549 * doc/extend.texi: Explain cloning for target_clone
4552 2020-01-07 Martin Liska <mliska@suse.cz>
4554 PR tree-optimization/92860
4555 * common.opt: Make in Optimization option
4556 as it is affected by -O0, which is an Optimization
4558 * tree-inline.c (tree_inlinable_function_p):
4559 Use opt_for_fn for warn_inline.
4560 (expand_call_inline): Likewise.
4562 2020-01-07 Martin Liska <mliska@suse.cz>
4564 PR tree-optimization/92860
4565 * common.opt: Make flag_ree as optimization
4568 2020-01-07 Martin Liska <mliska@suse.cz>
4570 PR optimization/92860
4571 * params.opt: Mark param_min_crossjump_insns with Optimization
4574 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4576 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4577 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4579 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4581 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4582 helper function to return the valid addressing formats for a given
4583 hard register and mode.
4584 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4586 * config/rs6000/constraints.md (Q constraint): Update
4588 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4591 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4592 Use 'Q' for doing vector extract from memory.
4593 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4595 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4596 doing vector extract from memory.
4597 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4598 extract from memory.
4600 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4601 for the offset being 34-bits when -mcpu=future is used.
4603 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4605 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4606 instead of cmpib_comparison_operator in cmpib patterns.
4607 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4608 of cmpib_comparison_operator. Revise comment.
4610 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4612 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4613 in an IFN_DIV_POW2 node to be equal.
4615 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4617 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4618 (vect_check_scalar_mask): ...this.
4619 (vectorizable_store, vectorizable_load): Update call accordingly.
4620 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4621 argument in calls to conditional internal functions.
4623 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4625 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4626 '0' matching inputs.
4627 (subv64di3_exec): Likewise.
4629 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4631 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4632 * doc/md.texi (movstr): Likewise.
4634 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4636 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4639 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4641 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4643 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4644 to a temporary file and use move-if-change to update the real
4645 file where necessary.
4647 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4649 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4650 rather than Upa for CPY /M.
4652 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4654 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4657 2020-01-06 Martin Liska <mliska@suse.cz>
4659 PR tree-optimization/92860
4660 * params.opt: Mark param_max_combine_insns with Optimization
4663 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4666 * config/i386/i386.md (SWIDWI): New mode iterator.
4667 (DWI, dwi): Add TImode variants.
4668 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4669 <general_hilo_operand> instead of <general_operand>. Use
4670 CONST_SCALAR_INT_P instead of CONST_INT_P.
4671 (*addv<mode>4_1): Rename to ...
4672 (addv<mode>4_1): ... this.
4673 (QWI): New mode attribute.
4674 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4675 define_insn_and_split patterns.
4676 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4678 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4679 <general_hilo_operand> instead of <general_operand>.
4680 (*addcarry<mode>_1): New define_insn.
4681 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4683 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4685 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4686 Use "call" instead of "set".
4688 2020-01-03 Martin Jambor <mjambor@suse.cz>
4691 * ipa-cp.c (print_all_lattices): Skip functions without info.
4693 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4696 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4697 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4698 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4699 for 'e' simd clones.
4702 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4704 (mprefer-vector-width=): Add Save.
4705 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4706 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4707 (ix86_debug_options, ix86_function_specific_print): Adjust
4708 ix86_target_string callers.
4709 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4710 (ix86_valid_target_attribute_tree): Likewise.
4711 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4712 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4713 ix86_target_string caller.
4716 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4717 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4718 instead of gen_int_shift_amount + convert_modes.
4720 PR rtl-optimization/93088
4721 * loop-iv.c (find_single_def_src): Punt after looking through
4722 128 reg copies for regs with single definitions. Move definitions
4725 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4727 * config/arm/arm-c.c (arm_cpu_builtins): Define
4728 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4729 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4730 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4731 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4732 * config/arm/arm-tables.opt: Regenerated.
4733 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4734 arm_arch_i8mm and arm_arch_bf16 when enabled.
4735 * config/arm/arm.h (TARGET_I8MM): New macro.
4736 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4737 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4738 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4739 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4740 (v8_6_a_simd_variants): New.
4741 (v8_*_a_simd_variants): Add i8mm and bf16.
4742 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4744 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4747 * predict.c (compute_function_frequency): Don't call
4748 warn_function_cold on functions that already have cold attribute.
4750 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4753 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4754 COMDAT group function labels in .data.rel.ro.local section.
4755 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4758 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4759 comparison_operator in B and S integer comparisons. Likewise, use
4760 ordered_comparison_operator instead of cmpib_comparison_operator in
4762 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4764 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4766 Update copyright years.
4768 * gcc.c (process_command): Update copyright notice dates.
4769 * gcov-dump.c (print_version): Ditto.
4770 * gcov.c (print_version): Ditto.
4771 * gcov-tool.c (print_version): Ditto.
4772 * gengtype.c (create_file): Ditto.
4773 * doc/cpp.texi: Bump @copying's copyright year.
4774 * doc/cppinternals.texi: Ditto.
4775 * doc/gcc.texi: Ditto.
4776 * doc/gccint.texi: Ditto.
4777 * doc/gcov.texi: Ditto.
4778 * doc/install.texi: Ditto.
4779 * doc/invoke.texi: Ditto.
4781 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4783 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4786 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4788 PR tree-optimization/93098
4789 * match.pd (popcount): For shift amounts, use integer_onep
4790 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4791 tests. Make sure that precision is power of two larger than or equal
4792 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4793 instead of ULL suffixed constants. Formatting fixes.
4795 Copyright (C) 2020 Free Software Foundation, Inc.
4797 Copying and distribution of this file, with or without modification,
4798 are permitted in any medium without royalty provided the copyright
4799 notice and this notice are preserved.