1 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
3 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
4 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
5 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
6 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
7 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
8 * config/arm/iterators.md (VSF2BF): New attribute.
9 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
10 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
11 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
13 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
15 * config/arm/arm.md (required_for_purecode): New attribute.
16 (enabled): Handle required_for_purecode.
17 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
18 work with -mpure-code.
20 2020-02-25 Jakub Jelinek <jakub@redhat.com>
22 PR rtl-optimization/93908
23 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
26 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
28 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
30 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
32 * doc/install.texi (--enable-checking): Adjust wording.
34 2020-02-25 Richard Biener <rguenther@suse.de>
36 PR tree-optimization/93868
37 * tree-vect-slp.c (slp_copy_subtree): New function.
38 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
39 re-arranging stmts in it.
41 2020-02-25 Jakub Jelinek <jakub@redhat.com>
44 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
45 dummy function and remove it at the end.
48 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
49 paramter -> parameter.
50 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
51 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
53 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
55 * doc/install.texi (--enable-checking): Properly document current
57 (--enable-stage1-checking): Minor clarification about bootstrap.
59 2020-02-24 David Malcolm <dmalcolm@redhat.com>
62 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
63 -fanalyzer-checker=taint is also required.
64 (-fanalyzer-checker=): Note that providing this option enables the
65 given checker, and doing so may be required for checkers that are
68 2020-02-24 David Malcolm <dmalcolm@redhat.com>
70 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
71 significant control flow events; add a "3" which shows all
72 control flow events; the old "3" becomes "4".
74 2020-02-24 Jakub Jelinek <jakub@redhat.com>
76 PR tree-optimization/93582
77 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
78 pd.offset and pd.size to be counted in bits rather than bytes, add
79 support for maxsizei that is not a multiple of BITS_PER_UNIT and
80 handle bitfield stores and loads.
81 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
82 uncomparable quantities - bytes vs. bits. Allow push_partial_def
83 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
84 pd.offset/pd.size to be counted in bits rather than bytes.
85 Formatting fix. Rename shadowed len variable to buflen.
87 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
88 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
91 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
92 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
93 * opts-common.c (parse_options_from_collect_gcc_options): New function.
94 (prepend_xassembler_to_collect_as_options): Likewise.
95 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
96 (prepend_xassembler_to_collect_as_options): Likewise.
97 * lto-opts.c (lto_write_options): Stream assembler options
98 in COLLECT_AS_OPTIONS.
99 * lto-wrapper.c (xassembler_options_error): New static variable.
100 (get_options_from_collect_gcc_options): Move parsing options code to
101 parse_options_from_collect_gcc_options and call it.
102 (merge_and_complain): Validate -Xassembler options.
103 (append_compiler_options): Handle OPT_Xassembler.
104 (run_gcc): Append command line -Xassembler options to
106 * doc/invoke.texi: Add documentation about using Xassembler
109 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
111 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
113 (riscv_rtx_costs): Update cost model for LTGT.
115 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
117 PR rtl-optimization/93564
118 * ira-color.c (struct update_cost_queue_elem): New member start.
119 (queue_update_cost, get_next_update_cost): Add new arg start.
120 (allocnos_conflict_p): New function.
121 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
122 Add checking conflicts with allocnos_conflict_p.
123 (update_costs_from_prefs, restore_costs_from_copies): Adjust
124 update_costs_from_allocno calls.
125 (update_conflict_hard_regno_costs): Add checking conflicts with
126 allocnos_conflict_p. Adjust calls of queue_update_cost and
127 get_next_update_cost.
128 (assign_hard_reg): Adjust calls of queue_update_cost. Add
130 (bucket_allocno_compare_func): Restore previous version.
132 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
134 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
135 double-word size when handling aggregate return values.
136 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
137 that homogeneous SFmode and DFmode aggregates are passed and returned
138 in general registers.
140 2020-02-21 Jakub Jelinek <jakub@redhat.com>
143 * opts.c (print_filtered_help): Translate help before appending
144 messages to it rather than after that.
146 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
148 PR rtl-optimization/PR92989
149 * lra-lives.c (process_bb_lives): Restore the original order
150 of the bb liveness update. Call make_hard_regno_dead for each
151 register clobbered at the start of an EH receiver.
153 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
156 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
157 self-recursively generated.
159 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
162 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
165 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
167 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
168 Document new target supports option.
170 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
172 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
173 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
174 * config/arm/iterators.md (MATMUL): New iterator.
175 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
176 (mmla_sfx): New attribute.
177 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
178 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
179 (UNSPEC_MATMUL_US): New.
181 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
183 * config/arm/arm.md: Prevent scalar shifts from being used when big
186 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
187 Richard Biener <rguenther@suse.de>
189 PR tree-optimization/93586
190 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
191 after mismatched array refs; do not sure type size information to
192 recover from unmatched referneces with !flag_strict_aliasing_p.
194 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
196 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
197 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
198 (scatter_store<mode>): Rename to ...
199 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
200 (scatter<mode>_exec): Delete. Move contents ...
201 (mask_scatter_store<mode>): ... here, and rename that to ...
202 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
203 Remove mode conversion.
204 (mask_gather_load<mode>): Rename to ...
205 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
206 Remove mode conversion.
207 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
209 2020-02-21 Martin Jambor <mjambor@suse.cz>
211 PR tree-optimization/93845
212 * tree-sra.c (verify_sra_access_forest): Only test access size of
215 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
217 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
218 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
219 (addv64di3_exec): Likewise.
220 (subv64di3): Likewise.
221 (subv64di3_exec): Likewise.
222 (addv64di3_zext): Likewise.
223 (addv64di3_zext_exec): Likewise.
224 (addv64di3_zext_dup): Likewise.
225 (addv64di3_zext_dup_exec): Likewise.
226 (addv64di3_zext_dup2): Likewise.
227 (addv64di3_zext_dup2_exec): Likewise.
228 (addv64di3_sext_dup2): Likewise.
229 (addv64di3_sext_dup2_exec): Likewise.
230 (<expander>v64di3): Likewise.
231 (<expander>v64di3_exec): Likewise.
232 (*<reduc_op>_dpp_shr_v64di): Likewise.
233 (*plus_carry_dpp_shr_v64di): Likewise.
234 * config/gcn/gcn.md (adddi3): Likewise.
235 (addptrdi3): Likewise.
236 (<expander>di3): Likewise.
238 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
240 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
242 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
244 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
245 support. Use aarch64_emit_mult instead of emitting multiplication
246 instructions directly.
247 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
248 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
250 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
252 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
253 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
254 instead of emitting multiplication instructions directly.
255 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
256 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
257 (@aarch64_frecps<mode>): New expanders.
259 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
261 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
262 on and produce uint64_ts rather than ints.
263 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
264 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
266 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
268 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
269 an unused xmsk register when handling approximate rsqrt.
271 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
273 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
274 flag_finite_math_only condition.
276 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
279 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
280 to destination operand for shufps alternative.
281 (*vec_extractv2si_1): Ditto.
283 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
286 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
289 2020-02-20 Martin Liska <mliska@suse.cz>
292 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
294 2020-02-20 Martin Liska <mliska@suse.cz>
297 * common/config/avr/avr-common.c: Remote trailing "|".
299 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
301 * collect2.c (maybe_run_lto_and_relink): Fix typo in
304 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
306 PR tree-optimization/93767
307 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
308 access-size bias from the offset calculations for negative strides.
310 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
312 * collect2.c (c_file, o_file): Make const again.
313 (ldout,lderrout, dump_ld_file): Remove.
314 (tool_cleanup): Avoid calling not signal-safe functions.
315 (maybe_run_lto_and_relink): Avoid possible signal handler
316 access to unintialzed memory (lto_o_files).
317 (main): Avoid leaking temp files in $TMPDIR.
318 Initialize c_file/o_file with concat, which avoids exposing
319 uninitialized memory to signal handler, which calls unlink(!).
320 Avoid calling maybe_unlink when the main function returns,
321 since the atexit handler is already doing this.
322 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
324 2020-02-19 Martin Jambor <mjambor@suse.cz>
326 PR tree-optimization/93776
327 * tree-sra.c (create_access): Do not create zero size accesses.
328 (get_access_for_expr): Do not search for zero sized accesses.
330 2020-02-19 Martin Jambor <mjambor@suse.cz>
332 PR tree-optimization/93667
333 * tree-sra.c (scalarizable_type_p): Return false if record fields
334 do not follow wach other.
336 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
338 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
339 rather than fmv.x.s/fmv.s.x.
341 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
343 * config/aarch64/aarch64-simd-builtins.def
344 (intrinsic_vec_smult_lo_): New.
345 (intrinsic_vec_umult_lo_): Likewise.
346 (vec_widen_smult_hi_): Likewise.
347 (vec_widen_umult_hi_): Likewise.
348 * config/aarch64/aarch64-simd.md
349 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
350 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
351 (vmull_high_s16): Likewise.
352 (vmull_high_s32): Likewise.
353 (vmull_high_u8): Likewise.
354 (vmull_high_u16): Likewise.
355 (vmull_high_u32): Likewise.
356 (vmull_s8): Likewise.
357 (vmull_s16): Likewise.
358 (vmull_s32): Likewise.
359 (vmull_u8): Likewise.
360 (vmull_u16): Likewise.
361 (vmull_u32): Likewise.
363 2020-02-18 Martin Liska <mliska@suse.cz>
365 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
366 bootstrap by missing removal of invalid sanity check.
368 2020-02-18 Martin Liska <mliska@suse.cz>
371 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
372 Always compare LHS of gimple_assign.
374 2020-02-18 Martin Liska <mliska@suse.cz>
377 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
378 and return type of functions.
379 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
380 Drop MALLOC attribute for void functions.
381 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
382 malloc_state for a new VOID clone.
384 2020-02-18 Martin Liska <mliska@suse.cz>
387 * common.opt: Add -fprofile-reproducibility.
388 * doc/invoke.texi: Document it.
389 * value-prof.c (dump_histogram_value):
390 Document and support behavior for counters[0]
391 being a negative value.
392 (get_nth_most_common_value): Handle negative
393 counters[0] in respect to flag_profile_reproducible.
395 2020-02-18 Jakub Jelinek <jakub@redhat.com>
398 * cgraph.c (verify_speculative_call): Use speculative_id instead of
399 speculative_uid in messages. Remove trailing whitespace from error
400 message. Use num_speculative_call_targets instead of
401 num_speculative_targets in a message.
402 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
403 edge messages and stmt instead of cal_stmt in reference message.
405 PR tree-optimization/93780
406 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
407 before calling build_vector_type.
408 (execute_update_addresses_taken): Likewise.
411 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
412 typo, functoin -> function.
413 * tree.c (free_lang_data_in_decl): Fix comment typo,
414 functoin -> function.
415 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
417 2020-02-17 David Malcolm <dmalcolm@redhat.com>
419 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
421 (print_option_information): Don't call get_option_url if URLs
424 2020-02-17 Alexandre Oliva <oliva@adacore.com>
426 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
427 handling of register_common-less targets.
429 2020-02-17 Martin Liska <mliska@suse.cz>
432 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
434 2020-02-17 Martin Liska <mliska@suse.cz>
437 * config/rs6000/rs6000.c (rs6000_option_override_internal):
440 2020-02-17 Martin Liska <mliska@suse.cz>
443 * config/rx/elf.opt: Fix typo.
445 2020-02-17 Richard Biener <rguenther@suse.de>
448 * opts-global.c (print_ignored_options): Use inform and
451 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
454 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
456 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
459 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
460 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
462 2020-02-15 Jason Merrill <jason@redhat.com>
464 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
466 2020-02-15 Jakub Jelinek <jakub@redhat.com>
468 PR tree-optimization/93744
469 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
470 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
471 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
472 sure @2 in the first and @1 in the other patterns has no side-effects.
474 2020-02-15 David Malcolm <dmalcolm@redhat.com>
475 Bernd Edlinger <bernd.edlinger@hotmail.de>
479 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
480 * configure.ac (--with-diagnostics-urls): New configuration
481 option, based on --with-diagnostics-color.
482 (DIAGNOSTICS_URLS_DEFAULT): New define.
483 * config.h: Regenerate.
484 * configure: Regenerate.
485 * diagnostic.c (diagnostic_urls_init): Handle -1 for
486 DIAGNOSTICS_URLS_DEFAULT from configure-time
487 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
488 and TERM_URLS environment variable.
489 * diagnostic-url.h (diagnostic_url_format): New enum type.
490 (diagnostic_urls_enabled_p): rename to...
491 (determine_url_format): ... this, and change return type.
492 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
493 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
494 the linux console, and mingw.
495 (diagnostic_urls_enabled_p): rename to...
496 (determine_url_format): ... this, and adjust.
497 * pretty-print.h (pretty_printer::show_urls): rename to...
498 (pretty_printer::url_format): ... this, and change to enum.
499 * pretty-print.c (pretty_printer::pretty_printer,
500 pp_begin_url, pp_end_url, test_urls): Adjust.
501 * doc/install.texi (--with-diagnostics-urls): Document the new
502 configuration option.
503 (--with-diagnostics-color): Document the existing interaction
504 with GCC_COLORS better.
505 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
506 vindex reference. Update description of defaults based on the above.
507 (-fdiagnostics-color): Update description of how -fdiagnostics-color
508 interacts with GCC_COLORS.
510 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
513 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
514 conjunction with TARGET_GNU_TLS in early return.
516 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
518 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
519 the mode is not wider than UNITS_PER_WORD.
521 2020-02-14 Martin Jambor <mjambor@suse.cz>
523 PR tree-optimization/93516
524 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
525 access of the same type as the parent.
526 (propagate_subaccesses_from_lhs): Likewise.
528 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
531 * config/i386/avx512vbmi2intrin.h
532 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
533 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
534 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
535 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
536 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
537 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
538 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
539 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
540 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
541 of lacking a closing parenthesis.
542 * config/i386/avx512vbmi2vlintrin.h
543 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
544 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
545 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
546 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
547 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
548 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
549 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
550 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
551 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
552 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
553 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
554 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
555 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
556 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
557 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
558 _mm_shldi_epi32, _mm_mask_shldi_epi32,
559 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
560 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
562 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
565 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
566 the target function entry.
568 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
570 * common/config/arc/arc-common.c (arc_option_optimization_table):
571 Disable if-conversion step when optimized for size.
573 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
575 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
576 R12-R15 are always in ARCOMPACT16_REGS register class.
577 * config/arc/arc.opt (mq-class): Deprecate.
578 * config/arc/constraint.md ("q"): Remove dependency on mq-class
580 * doc/invoke.texi (mq-class): Update text.
581 * common/config/arc/arc-common.c (arc_option_optimization_table):
584 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
586 * config/arc/arc.c (arc_insn_cost): New function.
587 (TARGET_INSN_COST): Define.
588 * config/arc/arc.md (cost): New attribute.
589 (add_n): Use arc_nonmemory_operand.
590 (ashlsi3_insn): Likewise, also update constraints.
591 (ashrsi3_insn): Likewise.
593 (add_shift): Likewise.
594 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
596 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
598 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
600 (umulsidi_600): Likewise.
602 2020-02-13 Jakub Jelinek <jakub@redhat.com>
605 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
606 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
607 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
608 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
609 pass __A to the builtin followed by __W instead of __A followed by
611 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
612 _mm512_mask_popcnt_epi64): Likewise.
613 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
614 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
615 _mm256_mask_popcnt_epi64): Likewise.
617 PR tree-optimization/93582
618 * fold-const.h (shift_bytes_in_array_left,
619 shift_bytes_in_array_right): Declare.
620 * fold-const.c (shift_bytes_in_array_left,
621 shift_bytes_in_array_right): New function, moved from
622 gimple-ssa-store-merging.c, no longer static.
623 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
624 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
625 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
626 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
627 shift_bytes_in_array.
628 (verify_shift_bytes_in_array): Rename to ...
629 (verify_shift_bytes_in_array_left): ... this. Use
630 shift_bytes_in_array_left instead of shift_bytes_in_array.
631 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
632 instead of verify_shift_bytes_in_array.
633 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
634 / native_interpret_expr where the store covers all needed bits,
635 punt on PDP-endian, otherwise allow all involved offsets and sizes
636 not to be byte-aligned.
639 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
640 use const_0_to_255_operand predicate instead of immediate_operand.
641 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
642 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
643 vgf2p8affineinvqb_<mode><mask_name>,
644 vgf2p8affineqb_<mode><mask_name>): Drop mode from
645 const_0_to_255_operand predicated operands.
647 2020-02-12 Jeff Law <law@redhat.com>
649 * config/h8300/h8300.md (comparison shortening peepholes): Use
650 a mode iterator to merge the HImode and SImode peepholes.
652 2020-02-12 Jakub Jelinek <jakub@redhat.com>
655 * real.c (is_even): Make static. Function comment fix.
656 (is_halfway_below): Make static, don't assert R is not inf/nan,
657 instead return false for those. Small formatting fixes.
659 2020-02-12 Martin Sebor <msebor@redhat.com>
662 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
663 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
664 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
665 (strlen_check_and_optimize_call): Adjust callee name.
667 2020-02-12 Jeff Law <law@redhat.com>
669 * config/h8300/h8300.md (comparison shortening peepholes): Drop
670 (and (xor)) variant. Combine other two into single peephole.
672 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
674 PR rtl-optimization/93565
675 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
677 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
679 * config/aarch64/aarch64-simd.md
680 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
681 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
682 generating separate ADDV and zero_extend patterns.
683 * config/aarch64/iterators.md (VDQV_E): New iterator.
685 2020-02-12 Jeff Law <law@redhat.com>
687 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
688 expanders, splits, etc.
689 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
690 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
691 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
692 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
693 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
694 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
696 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
698 2020-02-12 Jakub Jelinek <jakub@redhat.com>
701 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
702 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
703 TARGET_AVX512DQ from condition.
704 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
705 instead of <mask_mode512bit_condition> in condition. If
706 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
708 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
711 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
714 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
716 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
718 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
719 where strlen is more legible.
720 (rs6000_builtin_vectorized_libmass): Ditto.
721 (rs6000_print_options_internal): Ditto.
723 2020-02-11 Martin Sebor <msebor@redhat.com>
725 PR tree-optimization/93683
726 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
728 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
730 * config/rs6000/predicates.md (cint34_operand): Rename the
731 -mprefixed-addr option to be -mprefixed.
732 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
733 the -mprefixed-addr option to be -mprefixed.
734 (OTHER_FUTURE_MASKS): Likewise.
735 (POWERPC_MASKS): Likewise.
736 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
737 the -mprefixed-addr option to be -mprefixed. Change error
738 messages to refer to -mprefixed.
739 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
741 (rs6000_legitimate_offset_address_p): Likewise.
742 (rs6000_mode_dependent_address): Likewise.
743 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
744 "-mprefixed" for target attributes and pragmas.
745 (address_to_insn_form): Rename the -mprefixed-addr option to be
747 (rs6000_adjust_insn_length): Likewise.
748 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
749 -mprefixed-addr option to be -mprefixed.
750 (ASM_OUTPUT_OPCODE): Likewise.
751 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
752 -mprefixed-addr option to be -mprefixed.
753 * config/rs6000/rs6000.opt (-mprefixed): Rename the
754 -mprefixed-addr option to be prefixed. Change the option from
755 being undocumented to being documented.
756 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
757 -mprefixed option. Update the -mpcrel documentation to mention
760 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
762 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
763 including FIRST_PSEUDO_REGISTER - 1.
764 * ira-color.c (print_hard_reg_set): Ditto.
766 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
768 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
769 (USTERNOP_QUALIFIERS): New define.
770 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
771 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
772 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
773 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
774 * config/arm/arm_neon.h (vusdot_s32): New.
775 (vusdot_lane_s32): New.
776 (vusdotq_lane_s32): New.
777 (vsudot_lane_s32): New.
778 (vsudotq_lane_s32): New.
779 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
780 * config/arm/iterators.md (DOTPROD_I8MM): New.
781 (sup, opsuffix): Add <us/su>.
782 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
783 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
785 2020-02-11 Richard Biener <rguenther@suse.de>
787 PR tree-optimization/93661
788 PR tree-optimization/93662
789 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
791 * tree-sra.c (get_access_for_expr): Likewise.
793 2020-02-10 Jakub Jelinek <jakub@redhat.com>
796 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
797 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
798 Change condition from TARGET_AVX2 to TARGET_AVX.
800 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
803 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
806 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
808 Try to generate zero-based comparisons.
809 * config/cris/cris.c (cris_reduce_compare): New function.
810 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
811 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
812 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
814 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
817 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
818 in Thumb state and also as a destination in Arm state. Add T16
821 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
823 * md.texi (Define Subst): Match closing paren in example.
825 2020-02-10 Jakub Jelinek <jakub@redhat.com>
829 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
830 arguments of strncmp.
832 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
835 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
836 but different source value.
837 (adjust_callers_for_value_intersection): New function.
838 (gather_edges_for_value): Adjust order of callers to let a
839 non-self-recursive caller be the first element.
840 (self_recursive_pass_through_p): Add a new parameter "simple", and
841 check generalized self-recursive pass-through jump function.
842 (self_recursive_agg_pass_through_p): Likewise.
843 (find_more_scalar_values_for_callers_subset): Compute value from
844 pass-through jump function for self-recursive.
845 (intersect_with_plats): Cleanup previous implementation code for value
846 itersection with self-recursive call edge.
847 (intersect_with_agg_replacements): Likewise.
848 (intersect_aggregates_with_edge): Deduce value from pass-through jump
849 function for self-recursive call edge. Cleanup previous implementation
850 code for value intersection with self-recursive call edge.
851 (decide_whether_version_node): Remove dead callers and adjust order
852 to let a non-self-recursive caller be the first element.
854 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
856 * recog.c: Move pass_split_before_sched2 code in front of
857 pass_split_before_regstack.
858 (pass_data_split_before_sched2): Rename pass to split3 from split4.
859 (pass_data_split_before_regstack): Rename pass to split4 from split3.
860 (rest_of_handle_split_before_sched2): Remove.
861 (pass_split_before_sched2::execute): Unconditionally call
863 (enable_split_before_sched2): New function.
864 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
865 (pass_split_before_regstack::gate): Ditto.
866 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
867 Update name check for renamed split4 pass.
868 * config/sh/sh.c (register_sh_passes): Update pass insertion
869 point for renamed split4 pass.
871 2020-02-09 Jakub Jelinek <jakub@redhat.com>
873 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
874 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
875 copying them around between host and target.
877 2020-02-08 Andrew Pinski <apinski@marvell.com>
880 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
881 STRICT_ALIGNMENT also.
883 2020-02-08 Jim Wilson <jimw@sifive.com>
886 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
888 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
889 Jakub Jelinek <jakub@redhat.com>
892 * config/i386/i386.h (CALL_USED_REGISTERS): Make
893 xmm16-xmm31 call-used even in 64-bit ms-abi.
895 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
897 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
898 (simd_ummla, simd_usmmla): Likewise.
899 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
900 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
903 2020-02-07 Richard Biener <rguenther@suse.de>
906 * tree-inline.c (fold_marked_statements): Do a PRE walk,
907 skipping unreachable regions.
908 (optimize_inline_calls): Skip folding stmts when we didn't
911 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
914 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
915 Don't return aggregates with only SFmode and DFmode in SSE
917 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
919 2020-02-07 Jakub Jelinek <jakub@redhat.com>
922 * config/rs6000/rs6000-logue.c
923 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
924 if it fails, move rs into end_addr and retry. Add
925 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
926 the insn pattern doesn't describe well what exactly happens to
930 * config/i386/predicates.md (avx_identity_operand): Remove.
931 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
932 (avx_<castmode><avxsizesuffix>_<castmode>,
933 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
934 a VEC_CONCAT of the operand and UNSPEC_CAST.
935 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
936 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
940 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
941 recog_data.insn if distance_non_agu_define changed it.
943 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
946 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
947 we only had X-FORM (reg+reg) addressing for vectors. Also before
948 ISA 3.0, we only had X-FORM addressing for scalars in the
949 traditional Altivec registers.
951 2020-02-06 <zhongyunde@huawei.com>
952 Vladimir Makarov <vmakarov@redhat.com>
954 PR rtl-optimization/93561
955 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
958 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
960 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
963 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
965 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
966 where the low and the high 32 bits are equal to each other specially,
967 with an rldimi instruction.
969 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
971 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
973 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
975 * config/arm/arm-tables.opt: Regenerate.
977 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
980 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
981 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
982 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
984 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
986 PR rtl-optimization/87763
987 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
989 2020-02-06 Delia Burduv <delia.burduv@arm.com>
991 * config/aarch64/aarch64-simd-builtins.def
992 (bfmlaq): New built-in function.
993 (bfmlalb): New built-in function.
994 (bfmlalt): New built-in function.
995 (bfmlalb_lane): New built-in function.
996 (bfmlalt_lane): New built-in function.
997 * config/aarch64/aarch64-simd.md
998 (aarch64_bfmmlaqv4sf): New pattern.
999 (aarch64_bfmlal<bt>v4sf): New pattern.
1000 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1001 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1002 (vbfmlalbq_f32): New intrinsic.
1003 (vbfmlaltq_f32): New intrinsic.
1004 (vbfmlalbq_lane_f32): New intrinsic.
1005 (vbfmlaltq_lane_f32): New intrinsic.
1006 (vbfmlalbq_laneq_f32): New intrinsic.
1007 (vbfmlaltq_laneq_f32): New intrinsic.
1008 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1009 (bt): New int attribute.
1011 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1013 * config/i386/i386.md (*pushtf): Emit "#" instead of
1014 calling gcc_unreachable in insn output.
1017 (*pushsf_rex64): Ditto for alternatives other than 1.
1018 (*pushsf): Ditto for alternatives other than 1.
1020 2020-02-06 Martin Liska <mliska@suse.cz>
1022 PR gcov-profile/91971
1023 PR gcov-profile/93466
1024 * coverage.c (coverage_init): Revert mangling of
1025 path into filename. It can lead to huge filename length.
1026 Creation of subfolders seem more natural.
1028 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1031 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1032 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1033 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1035 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1038 * config/i386/predicates.md (avx_identity_operand): New predicate.
1039 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1040 define_insn_and_split.
1043 * omp-low.c (use_pointer_for_field): For nested constructs, also
1044 look for map clauses on target construct.
1045 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1046 taskreg_nesting_level.
1049 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1050 shared clause, call omp_notice_variable on outer context if any.
1052 2020-02-05 Jason Merrill <jason@redhat.com>
1055 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1056 non-zero address even if weak and not yet defined.
1058 2020-02-05 Martin Sebor <msebor@redhat.com>
1060 PR tree-optimization/92765
1061 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1062 * tree-ssa-strlen.c (compute_string_length): Remove.
1063 (determine_min_objsize): Remove.
1064 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1065 Avoid using type size as the upper bound on string length.
1066 (handle_builtin_string_cmp): Add an argument. Adjust.
1067 (strlen_check_and_optimize_call): Pass additional argument to
1068 handle_builtin_string_cmp.
1070 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1072 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1073 (*pushdi2_rex64 peephole2): Unconditionally split after
1075 (*ashl<mode>3_doubleword): Ditto.
1076 (*<shift_insn><mode>3_doubleword): Ditto.
1078 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1081 * config/rs6000/rs6000.c (get_vector_offset): Fix
1083 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1085 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1087 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1090 (Special Functions for Debugging the Analyzer): Update description
1091 of __analyzer_dump_exploded_nodes.
1093 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1096 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1097 include sets and not clobbers in the vzeroupper pattern.
1098 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1099 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1100 (*avx_vzeroupper_1): New define_insn_and_split.
1103 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1104 don't run when !optimize.
1105 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1108 2020-02-05 Richard Biener <rguenther@suse.de>
1111 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1112 checks before matching calls.
1114 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1116 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1117 function comment typo.
1120 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1121 simd_clone_create failed when i == 0, adjust clone->nargs by
1124 2020-02-05 Martin Liska <mliska@suse.cz>
1127 * doc/invoke.texi: Document that one should
1128 not combine ASLR and -fpch.
1130 2020-02-04 Richard Biener <rguenther@suse.de>
1132 PR tree-optimization/93538
1133 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1135 2020-02-04 Richard Biener <rguenther@suse.de>
1137 PR tree-optimization/91123
1138 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1139 (vn_walk_cb_data::last_vuse): New member.
1140 (vn_walk_cb_data::saved_operands): Likewsie.
1141 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1142 (vn_walk_cb_data::push_partial_def): Use finish.
1143 (vn_reference_lookup_2): Update last_vuse and use finish if
1144 we've saved operands.
1145 (vn_reference_lookup_3): Use finish and update calls to
1146 push_partial_defs everywhere. When translating through
1147 memcpy or aggregate copies save off operands and alias-set.
1148 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1149 operation for redundant store removal.
1151 2020-02-04 Richard Biener <rguenther@suse.de>
1153 PR tree-optimization/92819
1154 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1155 generating more stmts than before.
1157 2020-02-04 Martin Liska <mliska@suse.cz>
1159 * config/arm/arm.c (arm_gen_far_branch): Move the function
1160 outside of selftests.
1162 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1164 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1165 function to adjust PC-relative vector addresses.
1166 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1167 handle vectors with PC-relative addresses.
1169 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1171 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1173 (hard_reg_and_mode_to_addr_mask): Delete.
1174 (rs6000_adjust_vec_address): If the original vector address
1175 was REG+REG or REG+OFFSET and the element is not zero, do the add
1176 of the elements in the original address before adding the offset
1177 for the vector element. Use address_to_insn_form to validate the
1178 address using the register being loaded, rather than guessing
1179 whether the address is a DS-FORM or DQ-FORM address.
1181 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1183 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1184 to calculate the offset in memory from the start of a vector of a
1185 particular element. Add code to keep the element number in
1186 bounds if the element number is variable.
1187 (rs6000_adjust_vec_address): Move calculation of offset of the
1188 vector element to get_vector_offset.
1189 (rs6000_split_vec_extract_var): Do not do the initial AND of
1190 element here, move the code to get_vector_offset.
1192 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1194 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1197 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1199 * config/rs6000/constraints.md: Improve documentation.
1201 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1204 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1205 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1207 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1209 * config.gcc: Remove "carrizo" support.
1210 * config/gcn/gcn-opts.h (processor_type): Likewise.
1211 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1212 * config/gcn/gcn.opt (gpu_type): Likewise.
1213 * config/gcn/t-omp-device: Likewise.
1215 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1218 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1219 * config/arm/arm.c (arm_gen_far_branch): New function
1221 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1223 2020-02-03 Julian Brown <julian@codesourcery.com>
1224 Tobias Burnus <tobias@codesourcery.com>
1226 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1228 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1231 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1232 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1235 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1237 PR rtl-optimization/91333
1238 * ira-color.c (struct allocno_color_data): Add member
1240 (init_allocno_threads): Set the member up.
1241 (bucket_allocno_compare_func): Add compare hard reg
1244 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1246 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1248 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1249 * config.in: Regenerated.
1250 * configure: Regenerated.
1251 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1252 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1253 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1255 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1257 * configure: Regenerate.
1259 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1261 PR rtl-optimization/91333
1262 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1263 reg preferences comparison up.
1265 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1267 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1268 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1269 aarch64-sve-builtins-base.h.
1270 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1271 aarch64-sve-builtins-base.cc.
1272 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1273 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1275 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1276 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1277 (svcvtnt): New functions.
1278 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1279 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1280 (svcvtnt): New functions.
1281 (svcvt): Add a form that converts f32 to bf16.
1282 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1283 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1285 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1286 Treat B as bfloat16_t.
1287 (ternary_bfloat_lane_base): New class.
1288 (ternary_bfloat_def): Likewise.
1289 (ternary_bfloat): New shape.
1290 (ternary_bfloat_lane_def): New class.
1291 (ternary_bfloat_lane): New shape.
1292 (ternary_bfloat_lanex2_def): New class.
1293 (ternary_bfloat_lanex2): New shape.
1294 (ternary_bfloat_opt_n_def): New class.
1295 (ternary_bfloat_opt_n): New shape.
1296 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1297 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1298 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1299 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1300 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1301 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1302 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1303 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1304 the pattern off the narrow mode instead of the wider one.
1305 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1306 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1307 (sve_fp_op): Handle them.
1308 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1309 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1311 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1313 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1314 * config/aarch64/aarch64-modes.def (BF): Move definition before
1315 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1316 (SVE_MODES): Handle BF modes.
1317 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1319 (aarch64_full_sve_mode): Likewise.
1320 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1322 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1323 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1324 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1325 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1327 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1329 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1330 (TYPES_all_data): Add bf16.
1331 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1332 (register_tuple_type): Increase buffer size.
1333 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1334 (bf16): New type suffix.
1335 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1336 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1337 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1338 Change type from all_data to all_arith.
1339 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1342 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1343 Matthew Malcomson <matthew.malcomson@arm.com>
1344 Richard Sandiford <richard.sandiford@arm.com>
1346 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1347 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1348 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1349 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1350 __ARM_FEATURE_MATMUL_FP64.
1351 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1352 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1353 be disabled at the same time.
1354 (f32mm): New extension.
1355 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1356 (AARCH64_FL_F64MM): Bump to the next bit up.
1357 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1358 (TARGET_SVE_F64MM): New macros.
1359 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1360 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1361 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1362 (UNSPEC_ZIP2Q): New unspeccs.
1363 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1364 (optab, sur, perm_insn): Handle the new unspecs.
1365 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1366 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1367 TARGET_SVE_F64MM instead of separate tests.
1368 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1369 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1370 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1371 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1372 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1373 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1374 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1375 (TYPES_s_signed): New macro.
1376 (TYPES_s_integer): Use it.
1377 (TYPES_d_float): New macro.
1378 (TYPES_d_data): Use it.
1379 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1380 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1381 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1382 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1383 (svmmla): New shape.
1384 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1385 template parameters.
1386 (ternary_resize2_lane_base): Likewise.
1387 (ternary_resize2_base): New class.
1388 (ternary_qq_lane_base): Likewise.
1389 (ternary_intq_uintq_lane_def): Likewise.
1390 (ternary_intq_uintq_lane): New shape.
1391 (ternary_intq_uintq_opt_n_def): New class
1392 (ternary_intq_uintq_opt_n): New shape.
1393 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1394 (ternary_uintq_intq_def): New class.
1395 (ternary_uintq_intq): New shape.
1396 (ternary_uintq_intq_lane_def): New class.
1397 (ternary_uintq_intq_lane): New shape.
1398 (ternary_uintq_intq_opt_n_def): New class.
1399 (ternary_uintq_intq_opt_n): New shape.
1400 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1401 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1402 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1403 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1405 (svdotprod_lane_impl): ...this new class.
1406 (svmmla_impl, svusdot_impl): New classes.
1407 (svdot_lane): Update to use svdotprod_lane_impl.
1408 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1409 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1411 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1412 function, with no types defined.
1413 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1414 AARCH64_FL_I8MM functions.
1415 (svmmla): New AARCH64_FL_F32MM function.
1416 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1417 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1418 AARCH64_FL_F64MM function.
1419 (REQUIRED_EXTENSIONS):
1421 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1423 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1426 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1428 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1429 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1430 (*movti_internal): Do not check for
1431 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1432 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1433 just after check for TARGET_AVX.
1434 (*movdf_internal): Ditto.
1435 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1436 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1437 * config/i386/sse.md (mov<mode>_internal): Only check
1438 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1439 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1440 (<sse>_andnot<mode>3<mask_name>): Move check for
1441 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1442 (<code><mode>3<mask_name>): Ditto.
1443 (*andnot<mode>3): Ditto.
1444 (*andnottf3): Ditto.
1445 (*<code><mode>3): Ditto.
1446 (*<code>tf3): Ditto.
1447 (*andnot<VI:mode>3): Remove
1448 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1449 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1450 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1451 (sse4_1_blendv<ssemodesuffix>): Ditto.
1452 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1453 Explain that tune applies to 128bit instructions only.
1455 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1457 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1458 to definition of hsa_kernel_description. Parse assembly to find SGPR
1459 and VGPR count of kernel and store in hsa_kernel_description.
1461 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1463 PR rtl-optimization/91838
1464 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1465 to truncate if allowed or reject combination.
1467 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1469 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1470 (find_inv_vars_cb): Likewise.
1472 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1474 * calls.c (special_function_p): Split out the check for DECL_NAME
1475 being non-NULL and fndecl being extern at file scope into a
1476 new maybe_special_function_p and call it. Drop check for fndecl
1477 being non-NULL that was after a usage of DECL_NAME (fndecl).
1478 * tree.h (maybe_special_function_p): New inline function.
1480 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1482 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1483 (mask_gather_load<mode>): ... here, and zero-initialize the
1485 (maskload<mode>di): Zero-initialize the destination.
1488 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1491 * doc/analyzer.texi (Limitations): Note that constraints on
1492 floating-point values are currently ignored.
1494 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1497 * symtab.c (symtab_node::noninterposable_alias): If localalias
1498 already exists, but is not usable, append numbers after it until
1499 a unique name is found. Formatting fix.
1502 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1505 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1507 * config/gcn/gcn.c (print_operand): Handle LTGT.
1508 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1510 2020-01-30 Richard Biener <rguenther@suse.de>
1512 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1513 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1515 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1517 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1518 without a DECL in .data.rel.ro.local.
1520 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1523 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1527 * config/i386/sse.md
1528 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1529 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1530 any_extend code iterator instead of always zero_extend.
1531 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1532 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1533 Use any_extend code iterator instead of always zero_extend.
1534 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1535 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1536 Use any_extend code iterator instead of always zero_extend.
1537 (*sse2_pmovmskb_ext): New define_insn.
1538 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1541 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1542 (*popcountsi2_zext_falsedep): New define_insn.
1544 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1546 * config.in: Regenerated.
1547 * configure: Regenerated.
1549 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1552 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1553 LLVM's assembler changed the default in version 9.
1555 2020-01-24 Jeff Law <law@redhat.com>
1557 PR tree-optimization/89689
1558 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1560 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1564 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1566 PR rtl-optimization/87763
1567 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1568 simplification to handle subregs as well as bare regs.
1569 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1571 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1574 * ira.c (ira): Revert use of simplified LRA algorithm.
1576 2020-01-29 Martin Jambor <mjambor@suse.cz>
1578 PR tree-optimization/92706
1579 * tree-sra.c (struct access): Fields first_link, last_link,
1580 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1581 next_rhs_queued and grp_rhs_queued respectively, new fields
1582 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1583 (struct assign_link): Field next renamed to next_rhs, new field
1584 next_lhs. Updated comment.
1585 (work_queue_head): Renamed to rhs_work_queue_head.
1586 (lhs_work_queue_head): New variable.
1587 (add_link_to_lhs): New function.
1588 (relink_to_new_repr): Also relink LHS lists.
1589 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1590 (add_access_to_lhs_work_queue): New function.
1591 (pop_access_from_work_queue): Renamed to
1592 pop_access_from_rhs_work_queue.
1593 (pop_access_from_lhs_work_queue): New function.
1594 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1596 (child_would_conflict_in_lacc): Renamed to
1597 child_would_conflict_in_acc. Adjusted parameter names.
1598 (create_artificial_child_access): New parameter set_grp_read, use it.
1599 (subtree_mark_written_and_enqueue): Renamed to
1600 subtree_mark_written_and_rhs_enqueue.
1601 (propagate_subaccesses_across_link): Renamed to
1602 propagate_subaccesses_from_rhs.
1603 (propagate_subaccesses_from_lhs): New function.
1604 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1607 2020-01-29 Martin Jambor <mjambor@suse.cz>
1609 PR tree-optimization/92706
1610 * tree-sra.c (struct access): Adjust comment of
1611 grp_total_scalarization.
1612 (find_access_in_subtree): Look for single children spanning an entire
1614 (scalarizable_type_p): Allow register accesses, adjust callers.
1615 (completely_scalarize): Remove function.
1616 (scalarize_elem): Likewise.
1617 (create_total_scalarization_access): Likewise.
1618 (sort_and_splice_var_accesses): Do not track total scalarization
1620 (analyze_access_subtree): New parameter totally, adjust to new meaning
1621 of grp_total_scalarization.
1622 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1623 (can_totally_scalarize_forest_p): New function.
1624 (create_total_scalarization_access): Likewise.
1625 (create_total_access_and_reshape): Likewise.
1626 (total_should_skip_creating_access): Likewise.
1627 (totally_scalarize_subtree): Likewise.
1628 (analyze_all_variable_accesses): Perform total scalarization after
1629 subaccess propagation using the new functions above.
1630 (initialize_constant_pool_replacements): Output initializers by
1631 traversing the access tree.
1633 2020-01-29 Martin Jambor <mjambor@suse.cz>
1635 * tree-sra.c (verify_sra_access_forest): New function.
1636 (verify_all_sra_access_forests): Likewise.
1637 (create_artificial_child_access): Set parent.
1638 (analyze_all_variable_accesses): Call the verifier.
1640 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1642 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1643 if called on indirect edge.
1644 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1645 speculative call if needed.
1647 2020-01-29 Richard Biener <rguenther@suse.de>
1649 PR tree-optimization/93428
1650 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1651 permutation when the load node is created.
1652 (vect_analyze_slp_instance): Re-use it here.
1654 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1656 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1658 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1660 PR rtl-optimization/93272
1661 * ira-lives.c (process_out_of_region_eh_regs): New function.
1662 (process_bb_node_lives): Call it.
1664 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1666 * coverage.c (read_counts_file): Make error message lowercase.
1668 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1670 * profile-count.c (profile_quality_display_names): Fix ordering.
1672 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1675 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1676 hash only when edge is first within the sequence.
1677 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1678 (symbol_table::create_edge): Do not set target_prob.
1679 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1681 (cgraph_edge::make_speculative): Drop target_prob parameter.
1682 (cgraph_edge::speculative_call_info): Remove.
1683 (cgraph_edge::first_speculative_call_target): New member function.
1684 (update_call_stmt_hash_for_removing_direct_edge): New function.
1685 (cgraph_edge::resolve_speculation): Rewrite to new API.
1686 (cgraph_edge::speculative_call_for_target): New member function.
1687 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1688 multiple speculation targets.
1689 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1691 (verify_speculative_call): Verify that targets form an interval.
1692 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1693 (cgraph_edge::first_speculative_call_target): New member function.
1694 (cgraph_edge::next_speculative_call_target): New member function.
1695 (cgraph_edge::speculative_call_target_ref): New member function.
1696 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1697 (cgraph_edge): Remove target_prob.
1698 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1699 Fix handling of speculative calls.
1700 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1701 * ipa-fnsummary.c (analyze_function_body): Likewise.
1702 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1703 * ipa-profile.c (dump_histogram): Fix formating.
1704 (ipa_profile_generate_summary): Watch for overflows.
1705 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1706 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1707 (update_indirect_edges_after_inlining): Update to new API.
1708 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1710 * profile-count.h: (profile_probability::adjusted): New.
1711 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1712 updating of profile.
1713 * value-prof.c (gimple_ic_transform): Rename to ...
1714 (dump_ic_profile): ... this one; update dumping.
1715 (stream_in_histogram_value): Fix formating.
1716 (gimple_value_profile_transformations): Update.
1718 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1721 * config/i386/i386.md (*movoi_internal_avx): Remove
1722 TARGET_SSE_TYPELESS_STORES check.
1723 (*movti_internal): Prefer TARGET_AVX over
1724 TARGET_SSE_TYPELESS_STORES.
1725 (*movtf_internal): Likewise.
1726 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1727 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1728 from TARGET_SSE_TYPELESS_STORES.
1730 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1732 * diagnostic-core.h (warning_at): Rename overload to...
1733 (warning_meta): ...this.
1734 (emit_diagnostic_valist): Delete decl of overload taking
1735 diagnostic_metadata.
1736 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1737 (warning_at): Rename overload taking diagnostic_metadata to...
1738 (warning_meta): ...this.
1740 2020-01-28 Richard Biener <rguenther@suse.de>
1742 PR tree-optimization/93439
1743 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1744 * tree-cfg.c (move_sese_region_to_fn): ... here.
1745 (verify_types_in_gimple_reference): Verify used cliques are
1748 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1751 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1752 argument of a pointer to struct gcc_options and pass it to
1753 parse_mtune_ctrl_str.
1754 (ix86_function_specific_restore): Pass opts to
1755 set_ix86_tune_features.
1756 (ix86_option_override_internal): Likewise.
1757 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1758 gcc_options and use it for x_ix86_tune_ctrl_string.
1760 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1762 PR rtl-optimization/87763
1763 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1764 simplification to handle subregs as well as bare regs.
1765 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1767 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1769 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1770 for reduction chains that (now) include a call.
1772 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1774 PR tree-optimization/92822
1775 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1776 out the don't-care elements of a vector whose significant elements
1777 are duplicates, make the don't-care elements duplicates too.
1779 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1781 PR tree-optimization/93434
1782 * tree-predcom.c (split_data_refs_to_components): Record which
1783 components have had aliasing loads removed. Prevent store-store
1784 commoning for all such components.
1786 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1789 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1790 -1 or is_vshift is true, use new_vector with number of elts npatterns
1791 rather than new_unary_operation.
1793 PR tree-optimization/93454
1794 * gimple-fold.c (fold_array_ctor_reference): Perform
1795 elt_size.to_uhwi () just once, instead of calling it in every
1796 iteration. Punt if that value is above size of the temporary
1797 buffer. Decrease third native_encode_expr argument when
1798 bufoff + elt_sz is above size of buf.
1800 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1802 * config/mips/mips.c (mips_declare_object_name)
1803 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1805 2020-01-27 Martin Liska <mliska@suse.cz>
1807 PR gcov-profile/93403
1808 * tree-profile.c (gimple_init_gcov_profiler): Generate
1809 both __gcov_indirect_call_profiler_v4 and
1810 __gcov_indirect_call_profiler_v4_atomic.
1812 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1815 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1817 (@aarch64_split_simd_mov<mode>): Use it.
1818 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1819 Leave the vec_extract patterns to handle 2-element vectors.
1820 (aarch64_simd_mov_from_<mode>high): Likewise.
1821 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1822 (vec_extractv2dfv1df): Likewise.
1824 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1826 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1827 jump conditions for *compare_condjump<GPI:mode>.
1829 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1832 * digraph.cc (test_edge::test_edge): Specify template for base
1835 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1837 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1839 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1841 * config/arc/arc-protos.h (gen_mlo): Remove.
1842 (gen_mhi): Likewise.
1843 * config/arc/arc.c (AUX_MULHI): Define.
1844 (arc_must_save_reister): Special handling for r58/59.
1845 (arc_compute_frame_size): Consider mlo/mhi registers.
1846 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1848 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1849 mlo/mhi name selection.
1850 (arc_restore_callee_saves): Don't early restore blink when ISR.
1851 (arc_expand_prologue): Add mlo/mhi saving.
1852 (arc_expand_epilogue): Add mlo/mhi restoring.
1855 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1856 numbering when MUL64 option is used.
1857 (DWARF2_FRAME_REG_OUT): Define.
1858 * config/arc/arc.md (arc600_stall): New pattern.
1859 (VUNSPEC_ARC_ARC600_STALL): Define.
1860 (mulsi64): Use correct mlo/mhi registers.
1861 (mulsi_600): Clean it up.
1862 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1864 (mhi_operand): Likewise.
1866 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1867 Petro Karashchenko <petro.karashchenko@ring.com>
1869 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1870 attributes if needed.
1871 (prepare_move_operands): Generate special unspec instruction for
1873 (arc_isuncached_mem_p): Propagate uncached attribute to each
1875 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1876 (VUNSPEC_ARC_STDI): Likewise.
1877 (ALLI): New mode iterator.
1878 (mALLI): New mode attribute.
1879 (lddi): New instruction pattern.
1881 (stdidi_split): Split instruction for architectures which are not
1882 supporting ll64 option.
1883 (lddidi_split): Likewise.
1885 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1887 PR rtl-optimization/92989
1888 * lra-lives.c (process_bb_lives): Update the live-in set before
1889 processing additional clobbers.
1891 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1893 PR rtl-optimization/93170
1894 * cselib.c (cselib_invalidate_regno_val): New function, split out
1896 (cselib_invalidate_regno): ...here.
1897 (cselib_invalidated_by_call_p): New function.
1898 (cselib_process_insn): Iterate over all the hard-register entries in
1899 REG_VALUES and invalidate any that cross call-clobbered registers.
1901 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1903 * dojump.c (split_comparison): Use HONOR_NANS rather than
1904 HONOR_SNANS when splitting LTGT.
1906 2020-01-27 Martin Liska <mliska@suse.cz>
1909 * opts.c (print_filtered_help): Exclude language-specific
1910 options from --help=common unless enabled in all FEs.
1912 2020-01-27 Martin Liska <mliska@suse.cz>
1914 * opts.c (print_help): Exclude params from
1915 all except --help=param.
1917 2020-01-27 Martin Liska <mliska@suse.cz>
1920 * config/i386/i386-features.c (make_resolver_func):
1921 Align the code with ppc64 target implementation.
1922 Do not generate a unique name for resolver function.
1924 2020-01-27 Richard Biener <rguenther@suse.de>
1926 PR tree-optimization/93397
1927 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1928 converted reduction chain SLP graph adjustment.
1930 2020-01-26 Marek Polacek <polacek@redhat.com>
1933 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1936 2020-01-26 Jason Merrill <jason@redhat.com>
1939 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1942 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1944 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1945 (rx_setmem): Likewise.
1947 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1950 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1951 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1952 drop <di> from constraint of last operand.
1955 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1956 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1957 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1959 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1962 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1965 2020-01-24 Jeff Law <law@redhat.com>
1967 PR tree-optimization/92788
1968 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1971 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1974 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1975 *avx_vperm_broadcast_<mode>,
1976 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1977 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1978 Move before avx2_perm<mode>/avx512f_perm<mode>.
1981 * simplify-rtx.c (simplify_const_unary_operation,
1982 simplify_const_binary_operation): Punt for mode precision above
1983 MAX_BITSIZE_MODE_ANY_INT.
1985 2020-01-24 Andrew Pinski <apinski@marvell.com>
1987 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1990 2020-01-24 Jeff Law <law@redhat.com>
1993 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1994 for REGs. Call output_operand_lossage to get more reasonable
1997 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1999 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2000 gcn_fp_compare_operator.
2001 (vec_cmpu<mode>di): Use gcn_compare_operator.
2002 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2003 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2004 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2005 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2006 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2007 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2008 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2009 gcn_fp_compare_operator.
2010 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2011 gcn_fp_compare_operator.
2012 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2013 gcn_fp_compare_operator.
2014 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2015 gcn_fp_compare_operator.
2017 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2019 * doc/install.texi (Cross-Compiler-Specific Options): Document
2020 `--with-toolexeclibdir' option.
2022 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2024 * target.def (flags_regnum): Also mention effect on delay slot filling.
2025 * doc/tm.texi: Regenerate.
2027 2020-01-23 Jeff Law <law@redhat.com>
2029 PR translation/90162
2030 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2032 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2035 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2038 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2040 PR rtl-optimization/93402
2041 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2044 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2046 * config.in: Regenerated.
2047 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2048 for TARGET_LIBC_GNUSTACK.
2049 * configure: Regenerated.
2050 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2051 found to be 2.31 or greater.
2053 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2055 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2057 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2058 (mips_asm_file_end): New function. Delegate to
2059 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2060 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2062 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2065 * config/i386/i386-modes.def (POImode): New mode.
2066 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2067 * config/i386/i386.md (DPWI): New mode attribute.
2068 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2070 (QPWI): ... this. Use POI instead of OI for TImode.
2071 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2072 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2075 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2078 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2080 (speculation_tracker_rev): New pattern.
2081 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2082 Use speculation_tracker_rev to track the inverse condition.
2084 2020-01-23 Richard Biener <rguenther@suse.de>
2086 PR tree-optimization/93381
2087 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2088 alias-set of the def as argument and record the first one.
2089 (vn_walk_cb_data::first_set): New member.
2090 (vn_reference_lookup_3): Pass the alias-set of the current def
2091 to push_partial_def. Fix alias-set used in the aggregate copy
2093 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2094 * real.c (clear_significand_below): Fix out-of-bound access.
2096 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2099 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2100 New define_insn patterns.
2102 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2104 * doc/sourcebuild.texi (check-function-bodies): Add an
2105 optional target/xfail selector.
2107 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2109 PR rtl-optimization/93124
2110 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2111 bare USE and CLOBBER insns.
2113 2020-01-22 Andrew Pinski <apinski@marvell.com>
2115 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2117 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2120 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2121 diagnostic_manager into "ana" namespace.
2122 * selftest-run-tests.c (selftest::run_tests): Update for move of
2123 selftest::run_analyzer_selftests to
2124 ana::selftest::run_analyzer_selftests.
2126 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2128 * cfgexpand.c (union_stack_vars): Update the size.
2130 2020-01-22 Richard Biener <rguenther@suse.de>
2132 PR tree-optimization/93381
2133 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2134 throughout, handle all conversions the same.
2136 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2139 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2140 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2141 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2142 Call force_reg on high_in2 unconditionally.
2144 2020-01-22 Martin Liska <mliska@suse.cz>
2146 PR tree-optimization/92924
2147 * profile.c (compute_value_histograms): Divide
2150 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2153 * output.h (assemble_name_resolve): Declare.
2154 * varasm.c (assemble_name_resolve): New function.
2155 (assemble_name): Use it.
2156 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2158 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2160 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2161 update_web_docs_git instead of update_web_docs_svn.
2163 2020-01-21 Andrew Pinski <apinski@marvell.com>
2166 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2167 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2168 (*tlsgd_small_<mode>): Likewise.
2169 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2170 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2171 register. Convert that register back to dest using convert_mode.
2173 2020-01-21 Jim Wilson <jimw@sifive.com>
2175 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2178 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2179 Uros Bizjak <ubizjak@gmail.com>
2182 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2184 (legitimize_tls_address): Do GNU2 TLS address computation in
2185 ptr_mode and zero-extend result to Pmode.
2186 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2187 :P with :PTR and Pmode with ptr_mode.
2188 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2189 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2190 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2192 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2195 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2196 the last two operands are CONST_INT_P before using them as such.
2198 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2200 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2201 to get the integer element types.
2203 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2205 * config/aarch64/aarch64-sve-builtins.h
2206 (function_expander::convert_to_pmode): Declare.
2207 * config/aarch64/aarch64-sve-builtins.cc
2208 (function_expander::convert_to_pmode): New function.
2209 (function_expander::get_contiguous_base): Use it.
2210 (function_expander::prepare_gather_address_operands): Likewise.
2211 * config/aarch64/aarch64-sve-builtins-sve2.cc
2212 (svwhilerw_svwhilewr_impl::expand): Likewise.
2214 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2217 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2218 cfun->machine->label_is_assembled.
2219 (aarch64_print_patchable_function_entry): New.
2220 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2221 * config/aarch64/aarch64.h (struct machine_function): New field,
2224 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2227 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2230 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2233 * cgraph.c (cgraph_edge::resolve_speculation,
2234 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2235 call_stmt_site_hash.
2237 2020-01-21 Martin Liska <mliska@suse.cz>
2239 * config/rs6000/rs6000.c (common_mode_defined): Remove
2242 2020-01-21 Richard Biener <rguenther@suse.de>
2244 PR tree-optimization/92328
2245 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2246 type when value-numbering same-sized store by inserting a
2248 (eliminate_dom_walker::eliminate_stmt): When eliminating
2249 a redundant store handle bit-reinterpretation of the same value.
2251 2020-01-21 Andrew Pinski <apinski@marvel.com>
2254 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2256 (prepare_block_for_update): This. Use a worklist instead of
2259 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2261 * gcc/config/arm/arm.c (clear_operation_p):
2262 Initialise last_regno, skip first iteration
2263 based on the first_set value and use ints instead
2264 of the unnecessary HOST_WIDE_INTs.
2266 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2269 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2270 compare_mode other than SFmode or DFmode.
2272 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2275 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2276 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2277 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2279 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2281 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2283 2020-01-20 Andrew Pinski <apinski@marvell.com>
2286 * targhooks.c (default_print_patchable_function_entry): Use
2287 output_asm_insn to emit the nop instruction.
2289 2020-01-20 Fangrui Song <maskray@google.com>
2292 * targhooks.c (default_print_patchable_function_entry): Align to
2295 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2298 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2299 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2300 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2301 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2302 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2303 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2304 Remove the {q} suffix from lea.
2305 (*tls_dynamic_gnu2_call_64): Renamed to ...
2306 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2307 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2308 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2309 Pass Pmode to gen_tls_dynamic_gnu2_64.
2311 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2313 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2315 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2317 * config/aarch64/aarch64-sve-builtins-base.cc
2318 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2320 2020-01-20 Richard Biener <rguenther@suse.de>
2323 * dwarf2out.c (prune_unused_types): Unconditionally mark
2324 called function DIEs.
2326 2020-01-20 Martin Liska <mliska@suse.cz>
2328 PR tree-optimization/93199
2329 * tree-eh.c (struct leh_state): Add
2330 new field outer_non_cleanup.
2331 (cleanup_is_dead_in): Pass leh_state instead
2332 of eh_region. Add a checking that state->outer_non_cleanup
2333 points to outer non-clean up region.
2334 (lower_try_finally): Record outer_non_cleanup
2336 (lower_catch): Likewise.
2337 (lower_eh_filter): Likewise.
2338 (lower_eh_must_not_throw): Likewise.
2339 (lower_cleanup): Likewise.
2341 2020-01-20 Richard Biener <rguenther@suse.de>
2343 PR tree-optimization/93094
2344 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2345 (vect_transform_loop): Likewise.
2346 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2347 loop_vectorized_call to vect_transform_loop.
2348 * tree-vect-loop.c (vect_transform_loop): Pass down
2349 loop_vectorized_call to vect_loop_versioning.
2350 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2351 the earlier discovered loop_vectorized_call.
2353 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2355 * doc/contribute.texi: Update for SVN -> Git transition.
2356 * doc/install.texi: Likewise.
2358 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2361 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2362 speculative targets.
2363 (verify_speculative_call): New function
2364 (cgraph_node::verify_node): Use it.
2365 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2368 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2371 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2372 (cgraph_edge::make_direct): Remove all indirect targets.
2373 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2374 (cgraph_node::verify_node): Verify that only one call_stmt or
2375 lto_stmt_uid is set.
2376 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2378 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2379 (lto_output_ref): Simplify streaming of stmt.
2380 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2382 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2384 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2385 Mark parameter unused.
2387 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2389 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2391 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2393 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2395 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2397 * Makefile.in: Add coroutine-passes.o.
2398 * builtin-types.def (BT_CONST_SIZE): New.
2399 (BT_FN_BOOL_PTR): New.
2400 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2401 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2402 * coroutine-builtins.def: New file.
2403 * coroutine-passes.cc: New file.
2404 * function.h (struct GTY function): Add a bit to indicate that the
2405 function is a coroutine component.
2406 * internal-fn.c (expand_CO_FRAME): New.
2407 (expand_CO_YIELD): New.
2408 (expand_CO_SUSPN): New.
2409 (expand_CO_ACTOR): New.
2410 * internal-fn.def (CO_ACTOR): New.
2414 * passes.def: Add pass_coroutine_lower_builtins,
2415 pass_coroutine_early_expand_ifns.
2416 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2417 (make_pass_coroutine_early_expand_ifns): New.
2418 * doc/invoke.texi: Document the fcoroutines command line
2421 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2423 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2426 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2427 after checking the argument is a REG. Don't use REGNO (reg)
2428 again to set last_regno, reuse regno variable instead.
2430 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2432 * doc/analyzer.texi (Limitations): Add note about NaN.
2434 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2435 Sudakshina Das <sudi.das@arm.com>
2437 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2438 and valid immediate.
2439 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2440 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2441 * config/arm/constraints.md (Pg): New.
2442 * config/arm/predicates.md (long_shift_imm): New.
2443 (arm_reg_or_long_shift_imm): Likewise.
2444 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2445 (thumb2_lsll): Likewise.
2448 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2449 Sudakshina Das <sudi.das@arm.com>
2451 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2452 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2453 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2454 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2455 * config/arm/thumb2.md (thumb2_asrl): New.
2456 (thumb2_lsll): Likewise.
2458 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2460 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2463 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2465 * gdbinit.in (help-gcc-hooks): New command.
2466 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2467 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2470 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2472 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2473 correct target macro.
2475 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2477 * config/aarch64/aarch64-protos.h
2478 (aarch64_sve_ld1ro_operand_p): New.
2479 * config/aarch64/aarch64-sve-builtins-base.cc
2480 (class load_replicate): New.
2481 (class svld1ro_impl): New.
2482 (class svld1rq_impl): Change to inherit from load_replicate.
2483 (svld1ro): New sve intrinsic function base.
2484 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2485 New DEF_SVE_FUNCTION.
2486 * config/aarch64/aarch64-sve-builtins-base.h
2487 (svld1ro): New decl.
2488 * config/aarch64/aarch64-sve-builtins.cc
2489 (function_expander::add_mem_operand): Modify assert to allow
2491 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2493 * config/aarch64/aarch64.c
2494 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2495 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2496 (aarch64_sve_ld1ro_operand_p): New.
2497 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2498 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2499 * config/aarch64/predicates.md
2500 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2502 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2504 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2505 Introduce this ACLE specified predefined macro.
2506 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2507 (fp): Disabling this disables f64mm.
2508 (simd): Disabling this disables f64mm.
2509 (fp16): Disabling this disables f64mm.
2510 (sve): Disabling this disables f64mm.
2511 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2512 (AARCH64_ISA_F64MM): New.
2513 (TARGET_F64MM): New.
2514 * doc/invoke.texi (f64mm): Document new option.
2516 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2518 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2519 (neoversen1_tunings): Likewise.
2521 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2524 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2525 Add assert to ensure prolog has been emitted.
2526 (aarch64_split_atomic_op): Likewise.
2527 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2528 Use epilogue_completed rather than reload_completed.
2529 (aarch64_atomic_exchange<mode>): Likewise.
2530 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2531 (atomic_nand<mode>): Likewise.
2532 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2533 (atomic_fetch_nand<mode>): Likewise.
2534 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2535 (atomic_nand_fetch<mode>): Likewise.
2537 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2540 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2542 (REVERSE_CONDITION): Delete.
2543 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2544 (CCFP_CCFPE): Likewise.
2545 (e): New mode attribute.
2546 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2547 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2548 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2549 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2550 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2551 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2552 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2553 name of generator from gen_ccmpdi to gen_ccmpccdi.
2554 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2555 the previous comparison but aren't able to, use the new ccmp_rev
2558 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2560 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2561 than testing directly for INTEGER_CST.
2562 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2564 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2566 PR tree-optimization/93292
2567 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2568 get_vectype_for_scalar_type returns NULL.
2570 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2572 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2573 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2575 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2577 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2579 * params.opt: (max-predicted-iterations): Set bounds.
2580 * predict.c (real_almost_one, real_br_prob_base,
2581 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2582 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2583 probabilities; do not truncate to reg_br_prob_bases.
2584 (estimate_loops_at_level): Pass max_cyclic_prob.
2585 (estimate_loops): Compute max_cyclic_prob.
2586 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2588 * profile-count.c (profile_probability::to_sreal): New.
2589 * profile-count.h (class sreal): Move up in file.
2590 (profile_probability::to_sreal): Declare.
2592 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2595 (arm_invalid_conversion): New function for target hook.
2596 (arm_invalid_unary_op): New function for target hook.
2597 (arm_invalid_binary_op): New function for target hook.
2599 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2601 * config.gcc: Add arm_bf16.h.
2602 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2603 (arm_simd_builtin_std_type): Add BFmode.
2604 (arm_init_simd_builtin_types): Define element types for vector types.
2605 (arm_init_bf16_types): New function.
2606 (arm_init_builtins): Add arm_init_bf16_types function call.
2607 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2608 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2609 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2610 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2611 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2612 (arm_mangle_type): Add __bf16.
2613 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2614 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2615 arm_bf16_ptr_type_node.
2616 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2617 define_split between ARM registers.
2618 * config/arm/arm_bf16.h: New file.
2619 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2620 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2622 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2623 * config/arm/vfp.md: Add BFmode to movhf patterns.
2625 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2626 Andre Vieira <andre.simoesdiasvieira@arm.com>
2628 * config/arm/arm-cpus.in (mve, mve_float): New features.
2629 (dsp, mve, mve.fp): New options.
2630 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2631 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2632 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2634 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2635 Thomas Preud'homme <thomas.preudhomme@arm.com>
2637 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2639 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2640 error for using -mcmse when targeting Armv8.1-M Mainline.
2642 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2643 Thomas Preud'homme <thomas.preudhomme@arm.com>
2645 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2646 address in r4 when targeting Armv8.1-M Mainline.
2647 (nonsecure_call_value_internal): Likewise.
2648 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2649 a register match_operand again. Emit BLXNS when targeting
2651 (nonsecure_call_value_reg_thumb2): Likewise.
2653 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2654 Thomas Preud'homme <thomas.preudhomme@arm.com>
2656 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2657 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2658 variable as true when floating-point ABI is not hard. Replace
2659 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2660 Generate VLSTM and VLLDM instruction respectively before and
2661 after a function call to cmse_nonsecure_call function.
2662 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2663 (VUNSPEC_VLLDM): Likewise.
2664 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2665 (lazy_load_multiple_insn): Likewise.
2667 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2668 Thomas Preud'homme <thomas.preudhomme@arm.com>
2670 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2671 (arm_emit_vfp_multi_reg_pop): Likewise.
2672 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2673 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2674 restore callee-saved VFP registers.
2676 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2677 Thomas Preud'homme <thomas.preudhomme@arm.com>
2679 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2680 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2681 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2682 callee-saved GPRs as well as clear ip register before doing a nonsecure
2683 call then restore callee-saved GPRs after it when targeting
2685 (arm_reorg): Adapt to function rename.
2687 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2688 Thomas Preud'homme <thomas.preudhomme@arm.com>
2690 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2691 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2692 clear_vfp_multiple pattern based on a new vfp parameter.
2693 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2694 targeting Armv8.1-M Mainline.
2695 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2696 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2697 whether VFP registers are available before looking call_used_regs for a
2699 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2700 of prototype of clear_operation_p.
2701 (clear_vfp_multiple_operation): New predicate.
2702 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2703 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2705 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2706 Thomas Preud'homme <thomas.preudhomme@arm.com>
2708 * config/arm/arm-protos.h (clear_operation_p): Declare.
2709 * config/arm/arm.c (clear_operation_p): New function.
2710 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2711 targeting Armv8.1-M Mainline or successor.
2712 (output_return_instruction): Only output APSR register clearing if
2713 Armv8.1-M Mainline instructions not available.
2714 (thumb_exit): Likewise.
2715 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2716 * config/arm/thumb2.md (clear_apsr): New define_insn.
2717 (clear_multiple): Likewise.
2718 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2720 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2721 Thomas Preud'homme <thomas.preudhomme@arm.com>
2723 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2724 (use_return_insn): Also return false for Armv8.1-M Mainline.
2725 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2726 Mainline instructions are available.
2727 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2728 when targeting Armv8.1-M Mainline Security Extensions.
2729 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2730 Mainline entry function.
2731 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2732 targeting Armv8.1-M Mainline or successor.
2733 (arm_expand_epilogue): Fix indentation of caller-saved register
2734 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2736 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2737 (FP_SYSREGS): Likewise.
2738 (enum vfp_sysregs_encoding): Define enum.
2739 (fp_sysreg_names): Declare.
2740 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2741 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2742 (pop_fpsysreg_insn): Likewise.
2744 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2745 Thomas Preud'homme <thomas.preudhomme@arm.com>
2747 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2748 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2749 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2750 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2751 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2752 (ARMv8_1m_main): New feature group.
2753 (armv8.1-m.main): New architecture.
2754 * config/arm/arm-tables.opt: Regenerate.
2755 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2756 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2757 (arm_options_perform_arch_sanity_checks): Error out when targeting
2758 Armv8.1-M Mainline Security Extensions.
2759 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2761 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2763 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2764 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2765 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2766 aarch64_bfdot_laneq): New.
2767 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2768 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2769 vbfdotq_laneq_f32): New.
2770 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2771 VBFMLA_W, VBF): New.
2772 (isquadop): Add V4BF, V8BF.
2774 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2776 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2777 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2778 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2779 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2780 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2781 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2782 usdot_laneq, sudot_lane,sudot_laneq): New.
2783 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2784 (aarch64_<sur>dot_lane): New.
2785 * config/aarch64/arm_neon.h (vusdot_s32): New.
2787 (vusdot_lane_s32): New.
2788 (vsudot_lane_s32): New.
2789 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2790 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2792 2020-01-16 Martin Liska <mliska@suse.cz>
2794 * value-prof.c (dump_histogram_value): Fix
2795 obvious spacing issue.
2797 2020-01-16 Andrew Pinski <apinski@marvell.com>
2799 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2800 !storage_order_barrier_p.
2802 2020-01-16 Andrew Pinski <apinski@marvell.com>
2804 * sched-int.h (_dep): Add unused bit-field field for the padding.
2805 * sched-deps.c (init_dep_1): Init unused field.
2807 2020-01-16 Andrew Pinski <apinski@marvell.com>
2809 * optabs.h (create_expand_operand): Initialize target field also.
2811 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2813 PR tree-optimization/92429
2814 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2815 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2817 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2820 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2822 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2823 aarch64_sve_int_mode to each mode.
2825 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2827 * doc/analyzer.texi (Overview): Add note about
2828 -fdump-ipa-analyzer.
2830 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2832 PR tree-optimization/93231
2833 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2834 input_type is unsigned. Use tree_to_shwi for shift constant.
2835 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2836 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2837 inputs more efficiently.
2839 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2841 * config/i386/i386.md (*movsf_internal): Do not require
2842 SSE2 ISA for alternatives 14 and 15.
2844 2020-01-15 Richard Biener <rguenther@suse.de>
2847 * tree-eh.c (sink_clobbers): If we already visited the destination
2848 block do not defer insertion.
2849 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2850 the purpose of defered insertion.
2852 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2854 * BASE-VER: Bump to 10.0.1.
2856 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2858 PR tree-optimization/93247
2859 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2860 type of the stmt that we're going to vectorize.
2862 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2864 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2865 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2868 2020-01-15 Martin Liska <mliska@suse.cz>
2870 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2871 2 calls of streamer_read_hwi in a function call.
2873 2020-01-15 Richard Biener <rguenther@suse.de>
2875 * alias.c (record_alias_subset): Avoid redundant work when
2876 subset is already recorded.
2878 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2880 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2881 the analyzer options provide CWE identifiers.
2883 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2885 * tree-diagnostic-path.cc (path_summary::event_range::print):
2886 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2887 using get_pure_location.
2889 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2891 PR tree-optimization/93262
2892 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2893 perform head trimming only if the last argument is constant,
2894 either all ones, or larger or equal to head trim, in the latter
2895 case decrease the last argument by head_trim.
2897 PR tree-optimization/93249
2898 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2899 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2900 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2901 perform head trim unless we can prove there are no '\0' chars
2902 from the source among the first head_trim chars.
2904 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2906 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2908 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2911 * config/i386/sse.md
2912 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2913 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2914 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2915 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2916 just a single alternative instead of two, make operands 1 and 2
2919 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2922 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2925 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2927 * Makefile.in (lang_opt_files): Add analyzer.opt.
2928 (ANALYZER_OBJS): New.
2929 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2930 tristate.o and ANALYZER_OBJS.
2931 (TEXI_GCCINT_FILES): Add analyzer.texi.
2932 * common.opt (-fanalyzer): New driver option.
2933 * config.in: Regenerate.
2934 * configure: Regenerate.
2935 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2936 (gccdepdir): Also create depdir for "analyzer" subdir.
2937 * digraph.cc: New file.
2938 * digraph.h: New file.
2939 * doc/analyzer.texi: New file.
2940 * doc/gccint.texi ("Static Analyzer") New menu item.
2941 (analyzer.texi): Include it.
2942 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2943 ("Warning Options"): Add static analysis warnings to the list.
2944 (-Wno-analyzer-double-fclose): New option.
2945 (-Wno-analyzer-double-free): New option.
2946 (-Wno-analyzer-exposure-through-output-file): New option.
2947 (-Wno-analyzer-file-leak): New option.
2948 (-Wno-analyzer-free-of-non-heap): New option.
2949 (-Wno-analyzer-malloc-leak): New option.
2950 (-Wno-analyzer-possible-null-argument): New option.
2951 (-Wno-analyzer-possible-null-dereference): New option.
2952 (-Wno-analyzer-null-argument): New option.
2953 (-Wno-analyzer-null-dereference): New option.
2954 (-Wno-analyzer-stale-setjmp-buffer): New option.
2955 (-Wno-analyzer-tainted-array-index): New option.
2956 (-Wno-analyzer-use-after-free): New option.
2957 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2958 (-Wno-analyzer-use-of-uninitialized-value): New option.
2959 (-Wanalyzer-too-complex): New option.
2960 (-fanalyzer-call-summaries): New warning.
2961 (-fanalyzer-checker=): New warning.
2962 (-fanalyzer-fine-grained): New warning.
2963 (-fno-analyzer-state-merge): New warning.
2964 (-fno-analyzer-state-purge): New warning.
2965 (-fanalyzer-transitivity): New warning.
2966 (-fanalyzer-verbose-edges): New warning.
2967 (-fanalyzer-verbose-state-changes): New warning.
2968 (-fanalyzer-verbosity=): New warning.
2969 (-fdump-analyzer): New warning.
2970 (-fdump-analyzer-callgraph): New warning.
2971 (-fdump-analyzer-exploded-graph): New warning.
2972 (-fdump-analyzer-exploded-nodes): New warning.
2973 (-fdump-analyzer-exploded-nodes-2): New warning.
2974 (-fdump-analyzer-exploded-nodes-3): New warning.
2975 (-fdump-analyzer-supergraph): New warning.
2976 * doc/sourcebuild.texi (dg-require-dot): New.
2977 (dg-check-dot): New.
2978 * gdbinit.in (break-on-saved-diagnostic): New command.
2979 * graphviz.cc: New file.
2980 * graphviz.h: New file.
2981 * ordered-hash-map-tests.cc: New file.
2982 * ordered-hash-map.h: New file.
2983 * passes.def (pass_analyzer): Add before
2984 pass_ipa_whole_program_visibility.
2985 * selftest-run-tests.c (selftest::run_tests): Call
2986 selftest::ordered_hash_map_tests_cc_tests.
2987 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2989 * shortest-paths.h: New file.
2990 * timevar.def (TV_ANALYZER): New timevar.
2991 (TV_ANALYZER_SUPERGRAPH): Likewise.
2992 (TV_ANALYZER_STATE_PURGE): Likewise.
2993 (TV_ANALYZER_PLAN): Likewise.
2994 (TV_ANALYZER_SCC): Likewise.
2995 (TV_ANALYZER_WORKLIST): Likewise.
2996 (TV_ANALYZER_DUMP): Likewise.
2997 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2998 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2999 * tree-pass.h (make_pass_analyzer): New decl.
3000 * tristate.cc: New file.
3001 * tristate.h: New file.
3003 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3006 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3007 alternatives 9 and 10.
3009 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3011 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3012 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3013 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3014 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3015 (selftest::hash_map_tests_c_tests): Call it.
3016 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3017 New static constant, using the value of = H::empty_zero_p.
3018 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3019 from default_hash_traits <Value>.
3020 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3022 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3023 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3024 calls to mark_empty with !Descriptor::empty_zero_p.
3025 (hash_table::empty_slow): Conditionalize the memset call with a
3026 check that Descriptor::empty_zero_p; otherwise, loop through the
3027 entries calling mark_empty on them.
3028 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3029 (pointer_hash::empty_zero_p): Likewise.
3030 (pair_hash::empty_zero_p): Likewise.
3031 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3033 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3034 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3035 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3036 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3037 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3038 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3039 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3040 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3042 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3045 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3047 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3048 fix typo on return value.
3050 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3053 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3055 (cgraph_edge::make_speculative): Add param for setting speculative_id
3057 (cgraph_edge::speculative_call_info): Update comments and find reference
3058 by speculative_id for multiple indirect targets.
3059 (cgraph_edge::resolve_speculation): Decrease the speculations
3060 for indirect edge, drop it's speculative if not direct target
3061 left. Update comments.
3062 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3063 (cgraph_node::dump): Print num_speculative_call_targets.
3064 (cgraph_node::verify_node): Don't report error if speculative
3065 edge not include statement.
3066 (cgraph_edge::num_speculative_call_targets_p): New function.
3067 * cgraph.h (int common_target_id): Remove.
3068 (int common_target_probability): Remove.
3069 (num_speculative_call_targets): New variable.
3070 (make_speculative): Add param for setting speculative_id.
3071 (cgraph_edge::num_speculative_call_targets_p): New declare.
3072 (target_prob): New variable.
3073 (speculative_id): New variable.
3074 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3075 call summaries for multiple speculative call targets.
3076 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3077 * ipa-profile.c (struct speculative_call_target): New struct.
3078 (class speculative_call_summary): New class.
3079 (class speculative_call_summaries): New class.
3080 (call_sums): New variable.
3081 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3082 (ipa_profile_write_edge_summary): New function.
3083 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3084 (ipa_profile_dump_all_summaries): New function.
3085 (ipa_profile_read_edge_summary): New function.
3086 (ipa_profile_read_summary_section): New function.
3087 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3088 (ipa_profile): Generate num_speculative_call_targets from
3090 * ipa-ref.h (speculative_id): New variable.
3091 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3092 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3093 common_target_probability. Stream out speculative_id and
3094 num_speculative_call_targets.
3095 (input_edge): Likewise.
3096 * predict.c (dump_prediction): Remove edges count assert to be
3098 * symtab.c (symtab_node::create_reference): Init speculative_id.
3099 (symtab_node::clone_references): Clone speculative_id.
3100 (symtab_node::clone_referring): Clone speculative_id.
3101 (symtab_node::clone_reference): Clone speculative_id.
3102 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3103 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3104 if indirect call contains multiple speculative targets.
3105 * value-prof.h (check_ic_target): Remove.
3106 * value-prof.c (gimple_value_profile_transformations):
3107 Use void function gimple_ic_transform.
3108 * value-prof.c (gimple_ic_transform): Handle topn case.
3109 Fix comment typos. Change it to a void function.
3111 2020-01-13 Andrew Pinski <apinski@marvell.com>
3113 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3114 (octeontx2t98): New define.
3115 (octeontx2t96): New define.
3116 (octeontx2t93): New define.
3117 (octeontx2f95): New define.
3118 (octeontx2f95n): New define.
3119 (octeontx2f95mm): New define.
3120 * config/aarch64/aarch64-tune.md: Regenerate.
3121 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3123 2020-01-13 Jason Merrill <jason@redhat.com>
3125 PR c++/33799 - destroy return value if local cleanup throws.
3126 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3128 2020-01-13 Martin Liska <mliska@suse.cz>
3130 * ipa-cp.c (get_max_overall_size): Use newly
3131 renamed param param_ipa_cp_unit_growth.
3132 * params.opt: Remove legacy param name.
3134 2020-01-13 Martin Sebor <msebor@redhat.com>
3136 PR tree-optimization/93213
3137 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3138 stores to be eliminated.
3140 2020-01-13 Martin Liska <mliska@suse.cz>
3142 * opts.c (print_help): Do not print CL_PARAM
3143 and CL_WARNING for CL_OPTIMIZATION.
3145 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3148 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3149 depending on optimization settings.
3151 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3153 PR tree-optimization/90838
3154 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3155 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3156 argument rather than to initialize temporary for targets that
3157 don't use the mode argument at all. Initialize ctzval to avoid
3160 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3162 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3163 * tree-core.h: Document it.
3164 * gimplify.c (gimplify_omp_workshare): Set it.
3165 * omp-low.c (lower_omp_target): Use it.
3166 * tree-pretty-print.c (dump_omp_clause): Print it.
3168 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3169 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3171 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3173 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3174 * common.opt (fdiagnostics-path-format=): New option.
3175 (diagnostic_path_format): New enum.
3176 (fdiagnostics-show-path-depths): New option.
3177 * coretypes.h (diagnostic_event_id_t): New forward decl.
3178 * diagnostic-color.c (color_dict): Add "path".
3179 * diagnostic-event-id.h: New file.
3180 * diagnostic-format-json.cc (json_from_expanded_location): Make
3182 (json_end_diagnostic): Call context->make_json_for_path if it
3183 exists and the diagnostic has a path.
3184 (diagnostic_output_format_init): Clear context->print_path.
3185 * diagnostic-path.h: New file.
3186 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3187 when printing a run of events in a diagnostic_path so that they
3188 all get the same color.
3189 (layout::m_diagnostic_path_p): New field.
3190 (layout::layout): Initialize it.
3191 (layout::print_any_labels): Don't colorize the label text for an
3192 event in a diagnostic_path.
3193 (gcc_rich_location::add_location_if_nearby): Add
3194 "restrict_to_current_line_spans" and "label" params. Pass the
3195 former to layout.maybe_add_location_range; pass the latter
3196 when calling add_range.
3197 * diagnostic.c: Include "diagnostic-path.h".
3198 (diagnostic_initialize): Initialize context->path_format and
3199 context->show_path_depths.
3200 (diagnostic_show_any_path): New function.
3201 (diagnostic_path::interprocedural_p): New function.
3202 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3203 (simple_diagnostic_path::num_events): New function.
3204 (simple_diagnostic_path::get_event): New function.
3205 (simple_diagnostic_path::add_event): New function.
3206 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3207 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3208 (debug): New overload taking a diagnostic_path *.
3209 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3210 * diagnostic.h (enum diagnostic_path_format): New enum.
3211 (json::value): New forward decl.
3212 (diagnostic_context::path_format): New field.
3213 (diagnostic_context::show_path_depths): New field.
3214 (diagnostic_context::print_path): New callback field.
3215 (diagnostic_context::make_json_for_path): New callback field.
3216 (diagnostic_show_any_path): New decl.
3217 (json_from_expanded_location): New decl.
3218 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3219 (-fdiagnostics-show-path-depths): New option.
3220 (-fdiagnostics-color): Add "path" to description of default
3221 GCC_COLORS; describe it.
3222 (-fdiagnostics-format=json): Document how diagnostic paths are
3223 represented in the JSON output format.
3224 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3225 Add optional params "restrict_to_current_line_spans" and "label".
3226 * opts.c (common_handle_option): Handle
3227 OPT_fdiagnostics_path_format_ and
3228 OPT_fdiagnostics_show_path_depths.
3229 * pretty-print.c: Include "diagnostic-event-id.h".
3230 (pp_format): Implement "%@" format code for printing
3231 diagnostic_event_id_t *.
3232 (selftest::test_pp_format): Add tests for "%@".
3233 * selftest-run-tests.c (selftest::run_tests): Call
3234 selftest::tree_diagnostic_path_cc_tests.
3235 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3236 * toplev.c (general_init): Initialize global_dc->path_format and
3237 global_dc->show_path_depths.
3238 * tree-diagnostic-path.cc: New file.
3239 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3240 non-static. Drop "diagnostic" param in favor of storing the
3241 original value of "where" and re-using it.
3242 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3243 maybe_unwind_expanded_macro_loc.
3244 (tree_diagnostics_defaults): Initialize context->print_path and
3245 context->make_json_for_path.
3246 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3248 (default_tree_make_json_for_path): New decl.
3249 (maybe_unwind_expanded_macro_loc): New decl.
3251 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3253 PR tree-optimization/93210
3254 * fold-const.h (native_encode_initializer,
3255 can_native_interpret_type_p): Declare.
3256 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3258 (native_encode_initializer): New function, moved from dwarf2out.c.
3259 Adjust to native_encode_expr compatible arguments, including dry-run
3260 and partial extraction modes. Don't handle STRING_CST.
3261 (can_native_interpret_type_p): No longer static.
3262 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3263 offset / BITS_PER_UNIT fits into int and don't call it if
3264 can_native_interpret_type_p fails. If suboff is NULL and for
3265 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3266 native_encode_initializer.
3267 (fold_const_aggregate_ref_1): Formatting fix.
3268 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3269 (tree_add_const_value_attribute): Adjust caller.
3271 PR tree-optimization/90838
3272 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3273 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3274 CTZ_DEFINED_VALUE_AT_ZERO.
3276 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3279 * lra-constraints.c (match_reload): Permit input operands have the
3280 same mode as output while other input operands have a different
3283 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3285 PR tree-optimization/90838
3286 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3287 (check_ctz_string): Likewise.
3288 (optimize_count_trailing_zeroes): Likewise.
3289 (simplify_count_trailing_zeroes): Likewise.
3290 (pass_forwprop::execute): Try ctz simplification.
3291 * match.pd: Add matching for ctz idioms.
3293 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3295 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3297 (aarch64_invalid_unary_op): New function for target hook.
3298 (aarch64_invalid_binary_op): New function for target hook.
3300 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3302 * config.gcc: Add arm_bf16.h.
3303 * config/aarch64/aarch64-builtins.c
3304 (aarch64_simd_builtin_std_type): Add BFmode.
3305 (aarch64_init_simd_builtin_types): Define element types for vector
3307 (aarch64_init_bf16_types): New function.
3308 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3309 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3311 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3312 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3314 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3315 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3316 * config/aarch64/aarch64.c
3317 (aarch64_classify_vector_mode): Add support for BF types.
3318 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3319 (aarch64_vq_mode): Add support for BF types.
3320 (aarch64_simd_container_mode): Add support for BF types.
3321 (aarch64_mangle_type): Add support for BF scalar type.
3322 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3323 * config/aarch64/arm_bf16.h: New file.
3324 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3325 * config/aarch64/iterators.md: Add BF types to mode attributes.
3326 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3328 2020-01-10 Jason Merrill <jason@redhat.com>
3330 PR c++/93173 - incorrect tree sharing.
3331 * gimplify.c (copy_if_shared): No longer static.
3332 * gimplify.h: Declare it.
3334 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3336 * doc/invoke.texi (-msve-vector-bits=): Document that
3337 -msve-vector-bits=128 now generates VL-specific code for
3338 little-endian targets.
3339 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3340 build_vector_type_for_mode to construct the data vector types.
3341 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3342 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3343 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3344 for 128-bit vectors.
3346 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3348 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3351 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3353 * config/aarch64/aarch64-builtins.c
3354 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3355 rather than checking the number of elements and the element mode.
3357 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3359 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3360 get_related_vectype_for_scalar_type rather than build_vector_type
3361 to create the index type for a conditional reduction.
3363 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3365 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3366 for any type of gather or scatter, including strided accesses.
3368 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3370 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3373 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3375 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3377 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3378 parameter and its use to reset DR_OFFSET's.
3379 (vect_transform_loop): Remove orig_drs_init argument.
3380 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3381 member of dr_vec_info rather than the offset of the associated
3382 data_reference's innermost_loop_behavior.
3383 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3384 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3385 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3386 get_dr_vinfo_offset.
3387 (vectorizable_store): Likewise.
3388 (vectorizable_load): Likewise.
3390 2020-01-10 Richard Biener <rguenther@suse.de>
3392 * gimple-ssa-store-merging
3393 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3395 2020-01-10 Martin Liska <mliska@suse.cz>
3398 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3399 encapsulation that was there before r280040.
3401 2020-01-10 Richard Biener <rguenther@suse.de>
3404 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3405 sequences to avoid walking them again for secondary opportunities.
3406 (pass_lower_eh_dispatch::execute): Instead actually insert
3409 2020-01-10 Richard Biener <rguenther@suse.de>
3412 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3413 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3414 avoid quadraticness.
3416 2020-01-10 Martin Jambor <mjambor@suse.cz>
3418 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3419 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3420 to get param_ipa_sra_max_replacements.
3421 (param_splitting_across_edge): Pass the caller to
3422 pull_accesses_from_callee.
3424 2020-01-10 Martin Jambor <mjambor@suse.cz>
3426 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3427 * ipa-cp.c (max_new_size): Removed.
3428 (orig_overall_size): New variable.
3429 (get_max_overall_size): New function.
3430 (estimate_local_effects): Use it. Adjust dump.
3431 (decide_about_value): Likewise.
3432 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3433 orig_overall_size. Adjust dump.
3434 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3436 2020-01-10 Martin Jambor <mjambor@suse.cz>
3438 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3439 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3440 instead of param_ipa_max_agg_items.
3441 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3442 optimization info for the callee.
3444 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3446 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3447 markers if debug_inline_points is false.
3449 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3451 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3453 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3454 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3455 aarch64-sve-builtins-sve2.h.
3456 (aarch64-sve-builtins-sve2.o): New rule.
3457 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3458 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3459 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3460 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3461 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3462 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3464 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3465 instructions that are handled here.
3466 (@cond_asrd<mode>): Generalize to...
3467 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3468 (*cond_asrd<mode>_2): Generalize to...
3469 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3470 (*cond_asrd<mode>_z): Generalize to...
3471 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3472 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3473 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3474 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3475 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3477 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3478 (@aarch64_scatter_stnt<mode>): Likewise.
3479 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3480 (@aarch64_mul_lane_<mode>): Likewise.
3481 (@aarch64_sve_suqadd<mode>_const): Likewise.
3482 (*<sur>h<addsub><mode>): Generalize to...
3483 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3485 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3486 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3487 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3488 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3489 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3490 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3491 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3492 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3493 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3494 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3495 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3496 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3497 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3498 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3499 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3500 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3501 (@aarch64_sve2_xar<mode>): Likewise.
3502 (@aarch64_sve2_bcax<mode>): Likewise.
3503 (*aarch64_sve2_eor3<mode>): Rename to...
3504 (@aarch64_sve2_eor3<mode>): ...this.
3505 (@aarch64_sve2_bsl<mode>): New expander.
3506 (@aarch64_sve2_nbsl<mode>): Likewise.
3507 (@aarch64_sve2_bsl1n<mode>): Likewise.
3508 (@aarch64_sve2_bsl2n<mode>): Likewise.
3509 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3510 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3511 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3512 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3513 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3514 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3515 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3516 (<su>mull<bt><Vwide>): Generalize to...
3517 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3519 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3520 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3521 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3522 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3523 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3524 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3525 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3526 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3527 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3528 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3529 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3530 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3531 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3532 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3533 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3534 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3535 (<SHRNB:r>shrnb<mode>): Generalize to...
3536 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3538 (<SHRNT:r>shrnt<mode>): Generalize to...
3539 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3541 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3542 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3543 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3544 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3545 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3546 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3547 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3548 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3549 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3550 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3551 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3552 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3553 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3554 (@aarch64_sve2_cvtnt<mode>): Likewise.
3555 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3556 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3557 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3558 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3559 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3560 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3561 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3562 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3563 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3564 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3565 (@aarch64_sve2_pmul<mode>): Likewise.
3566 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3567 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3568 (@aarch64_sve2_tbl2<mode>): Likewise.
3569 (@aarch64_sve2_tbx<mode>): Likewise.
3570 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3571 (@aarch64_sve2_histcnt<mode>): Likewise.
3572 (@aarch64_sve2_histseg<mode>): Likewise.
3573 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3574 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3575 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3576 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3577 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3578 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3579 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3580 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3581 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3582 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3583 (SVE2_PMULL_PAIR_I): New mode iterators.
3584 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3585 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3586 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3587 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3588 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3589 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3590 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3591 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3592 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3593 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3594 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3595 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3596 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3597 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3598 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3599 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3600 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3601 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3602 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3603 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3604 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3605 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3606 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3607 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3608 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3609 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3610 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3611 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3612 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3613 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3614 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3616 (VNARROW, Ventype): New mode attributes.
3617 (Vewtype): Handle VNx2DI. Fix typo in comment.
3618 (VDOUBLE): New mode attribute.
3619 (sve_lane_con): Handle VNx8HI.
3620 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3621 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3622 (sve_int_op, sve_int_op_rev): Handle the above codes.
3623 (sve_pred_int_rhs2_operand): Likewise.
3624 (MULLBT, SHRNB, SHRNT): Delete.
3625 (SVE_INT_SHIFT_IMM): New int iterator.
3626 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3627 and UNSPEC_WHILEHS for TARGET_SVE2.
3628 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3629 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3630 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3631 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3632 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3633 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3634 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3635 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3636 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3637 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3638 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3639 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3640 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3641 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3642 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3643 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3644 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3645 (optab): Handle the new unspecs.
3646 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3648 (lr): Handle the new unspecs.
3650 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3651 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3652 (sve_int_qsub_op): New int attributes.
3653 (sve_fp_op, rot): Handle the new unspecs.
3654 * config/aarch64/aarch64-sve-builtins.h
3655 (function_resolver::require_matching_pointer_type): Declare.
3656 (function_resolver::resolve_unary): Add an optional boolean argument.
3657 (function_resolver::finish_opt_n_resolution): Add an optional
3658 type_suffix_index argument.
3659 (gimple_folder::redirect_call): Declare.
3660 (gimple_expander::prepare_gather_address_operands): Add an optional
3662 * config/aarch64/aarch64-sve-builtins.cc: Include
3663 aarch64-sve-builtins-sve2.h.
3664 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3665 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3666 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3667 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3668 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3669 (TYPES_s_unsigned): Likewise.
3670 (TYPES_s_integer): Use TYPES_s_unsigned.
3671 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3672 (TYPES_sd_integer): Use them.
3673 (TYPES_d_unsigned): New macro.
3674 (TYPES_d_integer): Use it.
3675 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3676 (TYPES_cvt_narrow): Likewise.
3677 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3678 (preds_mx): New variable.
3679 (function_builder::add_overloaded_function): Allow the new feature
3680 set to be more restrictive than the original one.
3681 (function_resolver::infer_pointer_type): Remove qualifiers from
3682 the pointer type before printing it.
3683 (function_resolver::require_matching_pointer_type): New function.
3684 (function_resolver::resolve_sv_displacement): Handle functions
3685 that don't support 32-bit vector indices or svint32_t vector offsets.
3686 (function_resolver::finish_opt_n_resolution): Take the inferred type
3687 as a separate argument.
3688 (function_resolver::resolve_unary): Optionally treat all forms in
3689 the same way as normal merging functions.
3690 (gimple_folder::redirect_call): New function.
3691 (function_expander::prepare_gather_address_operands): Add an argument
3692 that says whether scaled forms are available. If they aren't,
3693 handle scaling of vector indices and don't add the extension and
3695 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3696 fall back to using cond_* instead.
3697 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3698 Split out the member variables into...
3699 (rtx_code_function_base): ...this new base class.
3700 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3701 (unspec_based_function): Split out the member variables into...
3702 (unspec_based_function_base): ...this new base class.
3703 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3704 (unspec_based_function_exact_insn): New class.
3705 (unspec_based_add_function, unspec_based_add_lane_function)
3706 (unspec_based_lane_function, unspec_based_pred_function)
3707 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3708 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3709 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3711 (unspec_based_fused_function): New class.
3712 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3713 (unspec_based_fused_lane_function): New class.
3714 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3716 (CODE_FOR_MODE1): New macro.
3717 (fixed_insn_function): New class.
3718 (while_comparison): Likewise.
3719 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3720 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3721 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3722 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3723 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3724 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3725 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3726 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3727 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3728 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3729 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3730 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3731 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3732 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3733 Also add an initial argument for unary_convert_narrowt, regardless
3734 of the predication type.
3735 (build_32_64): Allow loads and stores to specify MODE_none.
3736 (build_sv_index64, build_sv_uint_offset): New functions.
3737 (long_type_suffix): New function.
3738 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3739 (binary_imm_long_base, load_gather_sv_base): Likewise.
3740 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3741 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3742 (unary_narrowb_base, unary_narrowt_base): Likewise.
3743 (binary_long_lane_def, binary_long_lane): New shape.
3744 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3745 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3746 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3747 (binary_to_uint_def, binary_to_uint): Likewise.
3748 (binary_wide_def, binary_wide): Likewise.
3749 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3750 (compare_def, compare): Likewise.
3751 (compare_ptr_def, compare_ptr): Likewise.
3752 (load_ext_gather_index_restricted_def,
3753 load_ext_gather_index_restricted): Likewise.
3754 (load_ext_gather_offset_restricted_def,
3755 load_ext_gather_offset_restricted): Likewise.
3756 (load_gather_sv_def): Inherit from load_gather_sv_base.
3757 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3758 (shift_left_imm_def, shift_left_imm): Likewise.
3759 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3760 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3761 (store_scatter_index_restricted_def,
3762 store_scatter_index_restricted): Likewise.
3763 (store_scatter_offset_restricted_def,
3764 store_scatter_offset_restricted): Likewise.
3765 (tbl_tuple_def, tbl_tuple): Likewise.
3766 (ternary_long_lane_def, ternary_long_lane): Likewise.
3767 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3768 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3769 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3770 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3771 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3772 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3773 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3774 (ternary_uint_def, ternary_uint): Likewise.
3775 (unary_convert): Fix typo in comment.
3776 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3777 (unary_long_def, unary_long): Likewise.
3778 (unary_narrowb_def, unary_narrowb): Likewise.
3779 (unary_narrowt_def, unary_narrowt): Likewise.
3780 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3781 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3782 (unary_to_int_def, unary_to_int): Likewise.
3783 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3784 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3785 (svasrd_impl): Delete.
3786 (svcadd_impl::expand): Handle integer operations too.
3787 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3788 new functions to derive the unspec numbers.
3789 (svmla_svmls_lane_impl): Replace with...
3790 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3791 integer operations too.
3792 (svwhile_impl): Rename to...
3793 (svwhilelx_impl): ...this and inherit from while_comparison.
3794 (svasrd): Use unspec_based_function.
3795 (svmla_lane): Use svmla_lane_impl.
3796 (svmls_lane): Use svmls_lane_impl.
3797 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3798 (svwhilele, svwhilelt): Use svwhilelx_impl.
3799 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3800 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3801 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3802 * config/aarch64/aarch64-sve-builtins.def: Include
3803 aarch64-sve-builtins-sve2.def.
3805 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3807 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3808 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3809 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3810 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3811 immediates as well as vector ones.
3812 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3813 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3814 (aarch64_sve_qsub_immediate): Update calls accordingly.
3816 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3818 * config/aarch64/aarch64-sve2.md: Add banner comments.
3819 (<su>mulh<r>s<mode>3): Move further up file.
3820 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3821 (*aarch64_sve2_sra<mode>): Move further down file.
3822 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3824 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3826 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3828 (while_optab_cmp): Handle them.
3829 * config/aarch64/aarch64-sve.md
3830 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3831 and add a "@" marker.
3832 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3833 instead of gen_aarch64_sve2_while_ptest.
3834 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3836 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3838 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3839 (UNSPEC_WHILELE): ...this.
3840 (UNSPEC_WHILE_LO): Rename to...
3841 (UNSPEC_WHILELO): ...this.
3842 (UNSPEC_WHILE_LS): Rename to...
3843 (UNSPEC_WHILELS): ...this.
3844 (UNSPEC_WHILE_LT): Rename to...
3845 (UNSPEC_WHILELT): ...this.
3846 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3847 (cmp_op, while_optab_cmp): Likewise.
3848 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3849 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3850 (svwhilelt): Likewise.
3852 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3854 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3855 (unary_to_uint): Define.
3856 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3857 (unary_count): Rename to...
3858 (unary_to_uint_def, unary_to_uint): ...this.
3859 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3861 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3863 * config/aarch64/aarch64-sve-builtins-functions.h
3864 (code_for_mode_function): New class.
3865 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3866 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3867 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3868 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3869 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3871 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3873 * config/aarch64/iterators.md (addsub): New code attribute.
3874 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3876 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3877 in the asm string and attributes. Fix indentation.
3878 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3880 (@aarch64_sve_<optab><mode>): ...this.
3881 * config/aarch64/aarch64-sve-builtins.h
3882 (function_expander::expand_signed_unpred_op): Delete.
3883 * config/aarch64/aarch64-sve-builtins.cc
3884 (function_expander::expand_signed_unpred_op): Likewise.
3885 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3886 try using code_for_aarch64_sve instead.
3887 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3888 (svqsub_impl): Likewise.
3889 (svqadd, svqsub): Use rtx_code_function instead.
3891 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3893 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3894 (HADDSUB, sur, addsub): Remove them.
3896 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3898 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3899 functions the same way, rather than singling out those that
3900 aren't mapped directly to optabs.
3902 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3904 * target.def (compatible_vector_types_p): New target hook.
3905 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3906 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3907 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3908 * doc/tm.texi: Regenerate.
3909 * gimple-expr.c: Include target.h.
3910 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3911 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3913 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3914 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3915 Use the original predicate if it already has a suitable type.
3917 2020-01-09 Martin Jambor <mjambor@suse.cz>
3919 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3920 resolve_speculation and redirect_call_stmt_to_callee static. Change
3921 return type of set_call_stmt to cgraph_edge *.
3922 * auto-profile.c (afdo_indirect_call): Adjust call to
3923 redirect_call_stmt_to_callee.
3924 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3925 make the this pointer explicit, adjust self-recursive calls and the
3926 call top make_direct. Return the resulting edge.
3927 (cgraph_edge::remove): Make this pointer explicit.
3928 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3929 (cgraph_edge::make_direct): Likewise, adjust call to
3930 resolve_speculation.
3931 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3932 call to set_call_stmt.
3933 (cgraph_update_edges_for_call_stmt_node): Update call to
3934 set_call_stmt and remove.
3935 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3936 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3937 (cgraph_node::create_edge_including_clones): Moved "first" definition
3938 of edge to the block where it was used. Adjusted calls to
3940 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3941 cgraph_edge::remove.
3942 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3943 make_direct and redirect_call_stmt_to_callee.
3944 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3945 resolve_speculation and make_direct.
3946 * ipa-inline-transform.c (inline_transform): Adjust call to
3947 redirect_call_stmt_to_callee.
3948 (check_speculations_1):: Adjust call to resolve_speculation.
3949 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3950 resolve-speculation.
3951 (inline_small_functions): Adjust call to resolve_speculation.
3952 (ipa_inline): Likewise.
3953 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3955 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3956 safe with regards to edge removal, adjust calls to
3957 redirect_call_stmt_to_callee.
3958 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3959 and redirect_call_stmt_to_callee.
3960 * multiple_target.c (create_dispatcher_calls): Adjust call to
3961 redirect_call_stmt_to_callee
3962 (redirect_to_specific_clone): Likewise.
3963 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3964 Adjust calls to cgraph_edge::remove.
3965 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3966 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3967 (expand_call_inline): Adjust call to cgraph_edge::remove.
3969 2020-01-09 Martin Liska <mliska@suse.cz>
3971 * params.opt: Set Optimization for
3972 param_max_speculative_devirt_maydefs.
3974 2020-01-09 Martin Sebor <msebor@redhat.com>
3978 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3980 2020-01-09 Martin Liska <mliska@suse.cz>
3982 * auto-profile.c (auto_profile): Use opt_for_fn
3984 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3985 (propagate_vals_across_arith_jfunc): Likewise.
3986 (hint_time_bonus): Likewise.
3987 (incorporate_penalties): Likewise.
3988 (good_cloning_opportunity_p): Likewise.
3989 (perform_estimation_of_a_value): Likewise.
3990 (estimate_local_effects): Likewise.
3991 (ipcp_propagate_stage): Likewise.
3992 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3993 (set_switch_stmt_execution_predicate): Likewise.
3994 (analyze_function_body): Likewise.
3995 * ipa-inline-analysis.c (offline_size): Likewise.
3996 * ipa-inline.c (early_inliner): Likewise.
3997 * ipa-prop.c (ipa_analyze_node): Likewise.
3998 (ipcp_transform_function): Likewise.
3999 * ipa-sra.c (process_scan_results): Likewise.
4000 (ipa_sra_summarize_function): Likewise.
4001 * params.opt: Rename ipcp-unit-growth to
4002 ipa-cp-unit-growth. Add Optimization for various
4003 IPA-related parameters.
4005 2020-01-09 Richard Biener <rguenther@suse.de>
4008 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4010 2020-01-09 Richard Biener <rguenther@suse.de>
4012 PR tree-optimization/93040
4013 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4015 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4017 * common/config/avr/avr-common.c (avr_option_optimization_table)
4018 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4020 2020-01-09 Martin Liska <mliska@suse.cz>
4022 * cgraphclones.c (symbol_table::materialize_all_clones):
4023 Use cgraph_node::dump_name.
4025 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4028 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4029 output_operand_lossage instead of gcc_unreachable.
4030 * doc/md.texi (riscv f constraint): Fix typo.
4033 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4034 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4035 CONST_SCALAR_INT_P instead of CONST_INT_P.
4036 (*subv<mode>4_1): Rename to ...
4037 (subv<mode>4_1): ... this.
4038 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4039 define_insn_and_split patterns.
4040 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4043 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4045 * vec.c (class selftest::count_dtor): New class.
4046 (selftest::test_auto_delete_vec): New test.
4047 (selftest::vec_c_tests): Call it.
4048 * vec.h (class auto_delete_vec): New class template.
4049 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4051 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4053 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4055 2020-01-08 Jim Wilson <jimw@sifive.com>
4057 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4058 use of TLS_MODEL_LOCAL_EXEC when not pic.
4060 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4062 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4065 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4068 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4069 *stack_protect_set_3 peephole2): Also check that the second
4070 insns source is general_operand.
4073 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4074 predicate for output operand instead of register_operand.
4075 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4076 memory destination and non-memory operands[2].
4078 2020-01-08 Martin Liska <mliska@suse.cz>
4080 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4081 ::dump_asm_name instead of (::name or ::asm_name).
4082 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4083 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4084 (analyze_functions): Likewise.
4085 (expand_all_functions): Likewise.
4086 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4087 (propagate_bits_across_jump_function): Likewise.
4088 (dump_profile_updates): Likewise.
4089 (ipcp_store_bits_results): Likewise.
4090 (ipcp_store_vr_results): Likewise.
4091 * ipa-devirt.c (dump_targets): Likewise.
4092 * ipa-fnsummary.c (analyze_function_body): Likewise.
4093 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4094 (process_hsa_functions): Likewise.
4095 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4096 (set_alias_uids): Likewise.
4097 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4098 * ipa-inline.c (recursive_inlining): Likewise.
4099 (inline_to_all_callers_1): Likewise.
4100 (ipa_inline): Likewise.
4101 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4102 (ipa_propagate_frequency): Likewise.
4103 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4104 (remove_described_reference): Likewise.
4105 * ipa-pure-const.c (worse_state): Likewise.
4106 (check_retval_uses): Likewise.
4107 (analyze_function): Likewise.
4108 (propagate_pure_const): Likewise.
4109 (propagate_nothrow): Likewise.
4110 (dump_malloc_lattice): Likewise.
4111 (propagate_malloc): Likewise.
4112 (pass_local_pure_const::execute): Likewise.
4113 * ipa-visibility.c (optimize_weakref): Likewise.
4114 (function_and_variable_visibility): Likewise.
4115 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4116 (ipa_discover_variable_flags): Likewise.
4117 * lto-streamer-out.c (output_function): Likewise.
4118 (output_constructor): Likewise.
4119 * tree-inline.c (copy_bb): Likewise.
4120 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4121 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4123 2020-01-08 Richard Biener <rguenther@suse.de>
4126 * tree-eh.c (sink_clobbers): Update virtual operands for
4127 the first and last stmt only. Add a dry-run capability.
4128 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4129 after CFG manipulations and in RPO order to catch all
4130 secondary opportunities reliably.
4132 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4135 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4137 2019-01-08 Richard Biener <rguenther@suse.de>
4140 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4141 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4142 virtual operand, also updating SSA use.
4143 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4144 Update stmt after resetting virtual operand.
4145 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4146 * gimple-iterator.c (gsi_remove): When not removing the stmt
4147 permanently do not delink immediate uses or mark the stmt modified.
4149 2020-01-08 Martin Liska <mliska@suse.cz>
4151 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4152 (ipa_call_context::estimate_size_and_time): Likewise.
4153 (inline_analyze_function): Likewise.
4155 2020-01-08 Martin Liska <mliska@suse.cz>
4157 * cgraph.c (cgraph_node::dump): Use systematically
4160 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4162 Add -nodevicespecs option for avr.
4165 * config/avr/avr.opt (-nodevicespecs): New driver option.
4166 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4167 "-specs=device-specs/..." if that option is not set.
4168 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4170 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4172 Implement 64-bit double functions for avr.
4175 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4176 --with-double-comparison.
4177 * doc/install.texi: Document them.
4178 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4179 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4180 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4181 * doc/invoke.texi (AVR Built-in Macros): Document them.
4182 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4183 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4184 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4186 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4189 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4190 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4191 when only building rm-profile multilibs.
4193 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4196 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4197 lattice for a value to check.
4198 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4199 finite propagation in self-recursive scc.
4201 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4203 * ipa-inline.c (caller_growth_limits): Restore the AND.
4205 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4207 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4208 (VEC_ALLREG_ALT): New iterator.
4209 (VEC_ALLREG_INT_MODE): New iterator.
4210 (VCMP_MODE): New iterator.
4211 (VCMP_MODE_INT): New iterator.
4212 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4213 (vec_cmp<u>v64qidi): New define_expand.
4214 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4215 (vec_cmpu<mode>di_exec): New define_expand.
4216 (vec_cmp<u>v64qidi_exec): New define_expand.
4217 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4218 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4219 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4220 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4221 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4222 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4223 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4224 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4225 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4226 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4228 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4229 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4231 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4233 * config/gcn/constraints.md (DA): Update description and match.
4235 (Db): New constraint.
4236 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4238 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4239 Implement 'Db' mixed immediate type.
4240 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4241 (addcv64si3_dup<exec_vcc>): Delete.
4242 (subcv64si3<exec_vcc>): Rework constraints.
4243 (addv64di3): Rework constraints.
4244 (addv64di3_exec): Rework constraints.
4245 (subv64di3): Rework constraints.
4246 (addv64di3_dup): Delete.
4247 (addv64di3_dup_exec): Delete.
4248 (addv64di3_zext): Rework constraints.
4249 (addv64di3_zext_exec): Rework constraints.
4250 (addv64di3_zext_dup): Rework constraints.
4251 (addv64di3_zext_dup_exec): Rework constraints.
4252 (addv64di3_zext_dup2): Rework constraints.
4253 (addv64di3_zext_dup2_exec): Rework constraints.
4254 (addv64di3_sext_dup2): Rework constraints.
4255 (addv64di3_sext_dup2_exec): Rework constraints.
4257 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4259 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4260 existing target checks.
4262 2020-01-07 Richard Biener <rguenther@suse.de>
4264 * doc/install.texi: Bump minimal supported MPC version.
4266 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4268 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4269 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4270 * langhooks.c: Include stor-layout.h.
4271 (lhd_simulate_enum_decl): New function.
4272 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4273 handle_arm_sve_h for the LTO frontend.
4274 (register_vector_type): Cope with null returns from pushdecl.
4276 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4278 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4279 (aarch64_sve::nvectors_if_data_type): Replace with...
4280 (aarch64_sve::builtin_type_p): ...this.
4281 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4282 (find_vector_type): Delete.
4283 (add_sve_type_attribute): New function.
4284 (lookup_sve_type_attribute): Likewise.
4285 (register_builtin_types): Add an "SVE type" attribute to each type.
4286 (register_tuple_type): Likewise.
4287 (svbool_type_p, nvectors_if_data_type): Delete.
4288 (mangle_builtin_type): Use lookup_sve_type_attribute.
4289 (builtin_type_p): Likewise. Add an overload that returns the
4290 number of constituent vector and predicate registers.
4291 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4292 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4293 instead of aarch64_sve_argument_p.
4294 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4295 (aarch64_pass_by_reference): Likewise.
4296 (aarch64_function_value_1): Likewise.
4297 (aarch64_return_in_memory): Likewise.
4298 (aarch64_layout_arg): Likewise.
4300 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4302 PR tree-optimization/93156
4303 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4304 least significant bit is always clear.
4306 PR tree-optimization/93118
4307 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4308 simplifier with two intermediate conversions.
4310 2020-01-07 Martin Liska <mliska@suse.cz>
4312 * params.opt: Add Optimization for various parameters.
4314 2020-01-07 Martin Liska <mliska@suse.cz>
4317 * doc/extend.texi: Explain cloning for target_clone
4320 2020-01-07 Martin Liska <mliska@suse.cz>
4322 PR tree-optimization/92860
4323 * common.opt: Make in Optimization option
4324 as it is affected by -O0, which is an Optimization
4326 * tree-inline.c (tree_inlinable_function_p):
4327 Use opt_for_fn for warn_inline.
4328 (expand_call_inline): Likewise.
4330 2020-01-07 Martin Liska <mliska@suse.cz>
4332 PR tree-optimization/92860
4333 * common.opt: Make flag_ree as optimization
4336 2020-01-07 Martin Liska <mliska@suse.cz>
4338 PR optimization/92860
4339 * params.opt: Mark param_min_crossjump_insns with Optimization
4342 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4344 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4345 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4347 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4349 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4350 helper function to return the valid addressing formats for a given
4351 hard register and mode.
4352 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4354 * config/rs6000/constraints.md (Q constraint): Update
4356 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4359 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4360 Use 'Q' for doing vector extract from memory.
4361 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4363 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4364 doing vector extract from memory.
4365 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4366 extract from memory.
4368 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4369 for the offset being 34-bits when -mcpu=future is used.
4371 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4373 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4374 instead of cmpib_comparison_operator in cmpib patterns.
4375 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4376 of cmpib_comparison_operator. Revise comment.
4378 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4380 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4381 in an IFN_DIV_POW2 node to be equal.
4383 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4385 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4386 (vect_check_scalar_mask): ...this.
4387 (vectorizable_store, vectorizable_load): Update call accordingly.
4388 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4389 argument in calls to conditional internal functions.
4391 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4393 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4394 '0' matching inputs.
4395 (subv64di3_exec): Likewise.
4397 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4399 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4400 * doc/md.texi (movstr): Likewise.
4402 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4404 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4407 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4409 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4411 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4412 to a temporary file and use move-if-change to update the real
4413 file where necessary.
4415 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4417 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4418 rather than Upa for CPY /M.
4420 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4422 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4425 2020-01-06 Martin Liska <mliska@suse.cz>
4427 PR tree-optimization/92860
4428 * params.opt: Mark param_max_combine_insns with Optimization
4431 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4434 * config/i386/i386.md (SWIDWI): New mode iterator.
4435 (DWI, dwi): Add TImode variants.
4436 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4437 <general_hilo_operand> instead of <general_operand>. Use
4438 CONST_SCALAR_INT_P instead of CONST_INT_P.
4439 (*addv<mode>4_1): Rename to ...
4440 (addv<mode>4_1): ... this.
4441 (QWI): New mode attribute.
4442 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4443 define_insn_and_split patterns.
4444 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4446 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4447 <general_hilo_operand> instead of <general_operand>.
4448 (*addcarry<mode>_1): New define_insn.
4449 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4451 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4453 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4454 Use "call" instead of "set".
4456 2020-01-03 Martin Jambor <mjambor@suse.cz>
4459 * ipa-cp.c (print_all_lattices): Skip functions without info.
4461 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4464 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4465 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4466 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4467 for 'e' simd clones.
4470 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4472 (mprefer-vector-width=): Add Save.
4473 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4474 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4475 (ix86_debug_options, ix86_function_specific_print): Adjust
4476 ix86_target_string callers.
4477 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4478 (ix86_valid_target_attribute_tree): Likewise.
4479 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4480 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4481 ix86_target_string caller.
4484 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4485 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4486 instead of gen_int_shift_amount + convert_modes.
4488 PR rtl-optimization/93088
4489 * loop-iv.c (find_single_def_src): Punt after looking through
4490 128 reg copies for regs with single definitions. Move definitions
4493 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4495 * config/arm/arm-c.c (arm_cpu_builtins): Define
4496 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4497 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4498 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4499 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4500 * config/arm/arm-tables.opt: Regenerated.
4501 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4502 arm_arch_i8mm and arm_arch_bf16 when enabled.
4503 * config/arm/arm.h (TARGET_I8MM): New macro.
4504 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4505 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4506 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4507 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4508 (v8_6_a_simd_variants): New.
4509 (v8_*_a_simd_variants): Add i8mm and bf16.
4510 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4512 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4515 * predict.c (compute_function_frequency): Don't call
4516 warn_function_cold on functions that already have cold attribute.
4518 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4521 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4522 COMDAT group function labels in .data.rel.ro.local section.
4523 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4526 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4527 comparison_operator in B and S integer comparisons. Likewise, use
4528 ordered_comparison_operator instead of cmpib_comparison_operator in
4530 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4532 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4534 Update copyright years.
4536 * gcc.c (process_command): Update copyright notice dates.
4537 * gcov-dump.c (print_version): Ditto.
4538 * gcov.c (print_version): Ditto.
4539 * gcov-tool.c (print_version): Ditto.
4540 * gengtype.c (create_file): Ditto.
4541 * doc/cpp.texi: Bump @copying's copyright year.
4542 * doc/cppinternals.texi: Ditto.
4543 * doc/gcc.texi: Ditto.
4544 * doc/gccint.texi: Ditto.
4545 * doc/gcov.texi: Ditto.
4546 * doc/install.texi: Ditto.
4547 * doc/invoke.texi: Ditto.
4549 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4551 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4554 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4556 PR tree-optimization/93098
4557 * match.pd (popcount): For shift amounts, use integer_onep
4558 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4559 tests. Make sure that precision is power of two larger than or equal
4560 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4561 instead of ULL suffixed constants. Formatting fixes.
4563 Copyright (C) 2020 Free Software Foundation, Inc.
4565 Copying and distribution of this file, with or without modification,
4566 are permitted in any medium without royalty provided the copyright
4567 notice and this notice are preserved.