1 2020-02-25 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/93908
4 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
7 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
9 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
11 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
13 * doc/install.texi (--enable-checking): Adjust wording.
15 2020-02-25 Richard Biener <rguenther@suse.de>
17 PR tree-optimization/93868
18 * tree-vect-slp.c (slp_copy_subtree): New function.
19 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
20 re-arranging stmts in it.
22 2020-02-25 Jakub Jelinek <jakub@redhat.com>
25 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
26 dummy function and remove it at the end.
29 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
30 paramter -> parameter.
31 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
32 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
34 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
36 * doc/install.texi (--enable-checking): Properly document current
38 (--enable-stage1-checking): Minor clarification about bootstrap.
40 2020-02-24 David Malcolm <dmalcolm@redhat.com>
43 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
44 -fanalyzer-checker=taint is also required.
45 (-fanalyzer-checker=): Note that providing this option enables the
46 given checker, and doing so may be required for checkers that are
49 2020-02-24 David Malcolm <dmalcolm@redhat.com>
51 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
52 significant control flow events; add a "3" which shows all
53 control flow events; the old "3" becomes "4".
55 2020-02-24 Jakub Jelinek <jakub@redhat.com>
57 PR tree-optimization/93582
58 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
59 pd.offset and pd.size to be counted in bits rather than bytes, add
60 support for maxsizei that is not a multiple of BITS_PER_UNIT and
61 handle bitfield stores and loads.
62 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
63 uncomparable quantities - bytes vs. bits. Allow push_partial_def
64 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
65 pd.offset/pd.size to be counted in bits rather than bytes.
66 Formatting fix. Rename shadowed len variable to buflen.
68 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
69 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
72 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
73 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
74 * opts-common.c (parse_options_from_collect_gcc_options): New function.
75 (prepend_xassembler_to_collect_as_options): Likewise.
76 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
77 (prepend_xassembler_to_collect_as_options): Likewise.
78 * lto-opts.c (lto_write_options): Stream assembler options
79 in COLLECT_AS_OPTIONS.
80 * lto-wrapper.c (xassembler_options_error): New static variable.
81 (get_options_from_collect_gcc_options): Move parsing options code to
82 parse_options_from_collect_gcc_options and call it.
83 (merge_and_complain): Validate -Xassembler options.
84 (append_compiler_options): Handle OPT_Xassembler.
85 (run_gcc): Append command line -Xassembler options to
87 * doc/invoke.texi: Add documentation about using Xassembler
90 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
92 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
94 (riscv_rtx_costs): Update cost model for LTGT.
96 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
98 PR rtl-optimization/93564
99 * ira-color.c (struct update_cost_queue_elem): New member start.
100 (queue_update_cost, get_next_update_cost): Add new arg start.
101 (allocnos_conflict_p): New function.
102 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
103 Add checking conflicts with allocnos_conflict_p.
104 (update_costs_from_prefs, restore_costs_from_copies): Adjust
105 update_costs_from_allocno calls.
106 (update_conflict_hard_regno_costs): Add checking conflicts with
107 allocnos_conflict_p. Adjust calls of queue_update_cost and
108 get_next_update_cost.
109 (assign_hard_reg): Adjust calls of queue_update_cost. Add
111 (bucket_allocno_compare_func): Restore previous version.
113 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
115 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
116 double-word size when handling aggregate return values.
117 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
118 that homogeneous SFmode and DFmode aggregates are passed and returned
119 in general registers.
121 2020-02-21 Jakub Jelinek <jakub@redhat.com>
124 * opts.c (print_filtered_help): Translate help before appending
125 messages to it rather than after that.
127 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
129 PR rtl-optimization/PR92989
130 * lra-lives.c (process_bb_lives): Restore the original order
131 of the bb liveness update. Call make_hard_regno_dead for each
132 register clobbered at the start of an EH receiver.
134 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
137 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
138 self-recursively generated.
140 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
143 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
146 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
148 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
149 Document new target supports option.
151 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
153 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
154 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
155 * config/arm/iterators.md (MATMUL): New iterator.
156 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
157 (mmla_sfx): New attribute.
158 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
159 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
160 (UNSPEC_MATMUL_US): New.
162 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
164 * config/arm/arm.md: Prevent scalar shifts from being used when big
167 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
168 Richard Biener <rguenther@suse.de>
170 PR tree-optimization/93586
171 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
172 after mismatched array refs; do not sure type size information to
173 recover from unmatched referneces with !flag_strict_aliasing_p.
175 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
177 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
178 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
179 (scatter_store<mode>): Rename to ...
180 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
181 (scatter<mode>_exec): Delete. Move contents ...
182 (mask_scatter_store<mode>): ... here, and rename that to ...
183 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
184 Remove mode conversion.
185 (mask_gather_load<mode>): Rename to ...
186 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
187 Remove mode conversion.
188 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
190 2020-02-21 Martin Jambor <mjambor@suse.cz>
192 PR tree-optimization/93845
193 * tree-sra.c (verify_sra_access_forest): Only test access size of
196 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
198 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
199 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
200 (addv64di3_exec): Likewise.
201 (subv64di3): Likewise.
202 (subv64di3_exec): Likewise.
203 (addv64di3_zext): Likewise.
204 (addv64di3_zext_exec): Likewise.
205 (addv64di3_zext_dup): Likewise.
206 (addv64di3_zext_dup_exec): Likewise.
207 (addv64di3_zext_dup2): Likewise.
208 (addv64di3_zext_dup2_exec): Likewise.
209 (addv64di3_sext_dup2): Likewise.
210 (addv64di3_sext_dup2_exec): Likewise.
211 (<expander>v64di3): Likewise.
212 (<expander>v64di3_exec): Likewise.
213 (*<reduc_op>_dpp_shr_v64di): Likewise.
214 (*plus_carry_dpp_shr_v64di): Likewise.
215 * config/gcn/gcn.md (adddi3): Likewise.
216 (addptrdi3): Likewise.
217 (<expander>di3): Likewise.
219 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
221 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
223 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
225 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
226 support. Use aarch64_emit_mult instead of emitting multiplication
227 instructions directly.
228 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
229 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
231 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
233 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
234 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
235 instead of emitting multiplication instructions directly.
236 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
237 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
238 (@aarch64_frecps<mode>): New expanders.
240 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
242 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
243 on and produce uint64_ts rather than ints.
244 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
245 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
247 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
249 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
250 an unused xmsk register when handling approximate rsqrt.
252 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
254 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
255 flag_finite_math_only condition.
257 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
260 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
261 to destination operand for shufps alternative.
262 (*vec_extractv2si_1): Ditto.
264 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
267 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
270 2020-02-20 Martin Liska <mliska@suse.cz>
273 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
275 2020-02-20 Martin Liska <mliska@suse.cz>
278 * common/config/avr/avr-common.c: Remote trailing "|".
280 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
282 * collect2.c (maybe_run_lto_and_relink): Fix typo in
285 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
287 PR tree-optimization/93767
288 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
289 access-size bias from the offset calculations for negative strides.
291 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
293 * collect2.c (c_file, o_file): Make const again.
294 (ldout,lderrout, dump_ld_file): Remove.
295 (tool_cleanup): Avoid calling not signal-safe functions.
296 (maybe_run_lto_and_relink): Avoid possible signal handler
297 access to unintialzed memory (lto_o_files).
298 (main): Avoid leaking temp files in $TMPDIR.
299 Initialize c_file/o_file with concat, which avoids exposing
300 uninitialized memory to signal handler, which calls unlink(!).
301 Avoid calling maybe_unlink when the main function returns,
302 since the atexit handler is already doing this.
303 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
305 2020-02-19 Martin Jambor <mjambor@suse.cz>
307 PR tree-optimization/93776
308 * tree-sra.c (create_access): Do not create zero size accesses.
309 (get_access_for_expr): Do not search for zero sized accesses.
311 2020-02-19 Martin Jambor <mjambor@suse.cz>
313 PR tree-optimization/93667
314 * tree-sra.c (scalarizable_type_p): Return false if record fields
315 do not follow wach other.
317 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
319 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
320 rather than fmv.x.s/fmv.s.x.
322 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
324 * config/aarch64/aarch64-simd-builtins.def
325 (intrinsic_vec_smult_lo_): New.
326 (intrinsic_vec_umult_lo_): Likewise.
327 (vec_widen_smult_hi_): Likewise.
328 (vec_widen_umult_hi_): Likewise.
329 * config/aarch64/aarch64-simd.md
330 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
331 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
332 (vmull_high_s16): Likewise.
333 (vmull_high_s32): Likewise.
334 (vmull_high_u8): Likewise.
335 (vmull_high_u16): Likewise.
336 (vmull_high_u32): Likewise.
337 (vmull_s8): Likewise.
338 (vmull_s16): Likewise.
339 (vmull_s32): Likewise.
340 (vmull_u8): Likewise.
341 (vmull_u16): Likewise.
342 (vmull_u32): Likewise.
344 2020-02-18 Martin Liska <mliska@suse.cz>
346 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
347 bootstrap by missing removal of invalid sanity check.
349 2020-02-18 Martin Liska <mliska@suse.cz>
352 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
353 Always compare LHS of gimple_assign.
355 2020-02-18 Martin Liska <mliska@suse.cz>
358 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
359 and return type of functions.
360 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
361 Drop MALLOC attribute for void functions.
362 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
363 malloc_state for a new VOID clone.
365 2020-02-18 Martin Liska <mliska@suse.cz>
368 * common.opt: Add -fprofile-reproducibility.
369 * doc/invoke.texi: Document it.
370 * value-prof.c (dump_histogram_value):
371 Document and support behavior for counters[0]
372 being a negative value.
373 (get_nth_most_common_value): Handle negative
374 counters[0] in respect to flag_profile_reproducible.
376 2020-02-18 Jakub Jelinek <jakub@redhat.com>
379 * cgraph.c (verify_speculative_call): Use speculative_id instead of
380 speculative_uid in messages. Remove trailing whitespace from error
381 message. Use num_speculative_call_targets instead of
382 num_speculative_targets in a message.
383 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
384 edge messages and stmt instead of cal_stmt in reference message.
386 PR tree-optimization/93780
387 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
388 before calling build_vector_type.
389 (execute_update_addresses_taken): Likewise.
392 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
393 typo, functoin -> function.
394 * tree.c (free_lang_data_in_decl): Fix comment typo,
395 functoin -> function.
396 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
398 2020-02-17 David Malcolm <dmalcolm@redhat.com>
400 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
402 (print_option_information): Don't call get_option_url if URLs
405 2020-02-17 Alexandre Oliva <oliva@adacore.com>
407 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
408 handling of register_common-less targets.
410 2020-02-17 Martin Liska <mliska@suse.cz>
413 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
415 2020-02-17 Martin Liska <mliska@suse.cz>
418 * config/rs6000/rs6000.c (rs6000_option_override_internal):
421 2020-02-17 Martin Liska <mliska@suse.cz>
424 * config/rx/elf.opt: Fix typo.
426 2020-02-17 Richard Biener <rguenther@suse.de>
429 * opts-global.c (print_ignored_options): Use inform and
432 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
435 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
437 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
440 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
441 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
443 2020-02-15 Jason Merrill <jason@redhat.com>
445 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
447 2020-02-15 Jakub Jelinek <jakub@redhat.com>
449 PR tree-optimization/93744
450 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
451 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
452 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
453 sure @2 in the first and @1 in the other patterns has no side-effects.
455 2020-02-15 David Malcolm <dmalcolm@redhat.com>
456 Bernd Edlinger <bernd.edlinger@hotmail.de>
460 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
461 * configure.ac (--with-diagnostics-urls): New configuration
462 option, based on --with-diagnostics-color.
463 (DIAGNOSTICS_URLS_DEFAULT): New define.
464 * config.h: Regenerate.
465 * configure: Regenerate.
466 * diagnostic.c (diagnostic_urls_init): Handle -1 for
467 DIAGNOSTICS_URLS_DEFAULT from configure-time
468 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
469 and TERM_URLS environment variable.
470 * diagnostic-url.h (diagnostic_url_format): New enum type.
471 (diagnostic_urls_enabled_p): rename to...
472 (determine_url_format): ... this, and change return type.
473 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
474 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
475 the linux console, and mingw.
476 (diagnostic_urls_enabled_p): rename to...
477 (determine_url_format): ... this, and adjust.
478 * pretty-print.h (pretty_printer::show_urls): rename to...
479 (pretty_printer::url_format): ... this, and change to enum.
480 * pretty-print.c (pretty_printer::pretty_printer,
481 pp_begin_url, pp_end_url, test_urls): Adjust.
482 * doc/install.texi (--with-diagnostics-urls): Document the new
483 configuration option.
484 (--with-diagnostics-color): Document the existing interaction
485 with GCC_COLORS better.
486 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
487 vindex reference. Update description of defaults based on the above.
488 (-fdiagnostics-color): Update description of how -fdiagnostics-color
489 interacts with GCC_COLORS.
491 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
494 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
495 conjunction with TARGET_GNU_TLS in early return.
497 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
499 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
500 the mode is not wider than UNITS_PER_WORD.
502 2020-02-14 Martin Jambor <mjambor@suse.cz>
504 PR tree-optimization/93516
505 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
506 access of the same type as the parent.
507 (propagate_subaccesses_from_lhs): Likewise.
509 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
512 * config/i386/avx512vbmi2intrin.h
513 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
514 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
515 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
516 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
517 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
518 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
519 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
520 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
521 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
522 of lacking a closing parenthesis.
523 * config/i386/avx512vbmi2vlintrin.h
524 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
525 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
526 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
527 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
528 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
529 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
530 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
531 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
532 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
533 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
534 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
535 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
536 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
537 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
538 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
539 _mm_shldi_epi32, _mm_mask_shldi_epi32,
540 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
541 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
543 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
546 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
547 the target function entry.
549 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
551 * common/config/arc/arc-common.c (arc_option_optimization_table):
552 Disable if-conversion step when optimized for size.
554 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
556 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
557 R12-R15 are always in ARCOMPACT16_REGS register class.
558 * config/arc/arc.opt (mq-class): Deprecate.
559 * config/arc/constraint.md ("q"): Remove dependency on mq-class
561 * doc/invoke.texi (mq-class): Update text.
562 * common/config/arc/arc-common.c (arc_option_optimization_table):
565 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
567 * config/arc/arc.c (arc_insn_cost): New function.
568 (TARGET_INSN_COST): Define.
569 * config/arc/arc.md (cost): New attribute.
570 (add_n): Use arc_nonmemory_operand.
571 (ashlsi3_insn): Likewise, also update constraints.
572 (ashrsi3_insn): Likewise.
574 (add_shift): Likewise.
575 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
577 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
579 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
581 (umulsidi_600): Likewise.
583 2020-02-13 Jakub Jelinek <jakub@redhat.com>
586 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
587 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
588 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
589 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
590 pass __A to the builtin followed by __W instead of __A followed by
592 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
593 _mm512_mask_popcnt_epi64): Likewise.
594 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
595 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
596 _mm256_mask_popcnt_epi64): Likewise.
598 PR tree-optimization/93582
599 * fold-const.h (shift_bytes_in_array_left,
600 shift_bytes_in_array_right): Declare.
601 * fold-const.c (shift_bytes_in_array_left,
602 shift_bytes_in_array_right): New function, moved from
603 gimple-ssa-store-merging.c, no longer static.
604 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
605 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
606 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
607 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
608 shift_bytes_in_array.
609 (verify_shift_bytes_in_array): Rename to ...
610 (verify_shift_bytes_in_array_left): ... this. Use
611 shift_bytes_in_array_left instead of shift_bytes_in_array.
612 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
613 instead of verify_shift_bytes_in_array.
614 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
615 / native_interpret_expr where the store covers all needed bits,
616 punt on PDP-endian, otherwise allow all involved offsets and sizes
617 not to be byte-aligned.
620 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
621 use const_0_to_255_operand predicate instead of immediate_operand.
622 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
623 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
624 vgf2p8affineinvqb_<mode><mask_name>,
625 vgf2p8affineqb_<mode><mask_name>): Drop mode from
626 const_0_to_255_operand predicated operands.
628 2020-02-12 Jeff Law <law@redhat.com>
630 * config/h8300/h8300.md (comparison shortening peepholes): Use
631 a mode iterator to merge the HImode and SImode peepholes.
633 2020-02-12 Jakub Jelinek <jakub@redhat.com>
636 * real.c (is_even): Make static. Function comment fix.
637 (is_halfway_below): Make static, don't assert R is not inf/nan,
638 instead return false for those. Small formatting fixes.
640 2020-02-12 Martin Sebor <msebor@redhat.com>
643 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
644 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
645 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
646 (strlen_check_and_optimize_call): Adjust callee name.
648 2020-02-12 Jeff Law <law@redhat.com>
650 * config/h8300/h8300.md (comparison shortening peepholes): Drop
651 (and (xor)) variant. Combine other two into single peephole.
653 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
655 PR rtl-optimization/93565
656 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
658 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
660 * config/aarch64/aarch64-simd.md
661 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
662 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
663 generating separate ADDV and zero_extend patterns.
664 * config/aarch64/iterators.md (VDQV_E): New iterator.
666 2020-02-12 Jeff Law <law@redhat.com>
668 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
669 expanders, splits, etc.
670 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
671 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
672 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
673 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
674 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
675 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
677 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
679 2020-02-12 Jakub Jelinek <jakub@redhat.com>
682 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
683 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
684 TARGET_AVX512DQ from condition.
685 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
686 instead of <mask_mode512bit_condition> in condition. If
687 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
689 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
692 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
695 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
697 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
699 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
700 where strlen is more legible.
701 (rs6000_builtin_vectorized_libmass): Ditto.
702 (rs6000_print_options_internal): Ditto.
704 2020-02-11 Martin Sebor <msebor@redhat.com>
706 PR tree-optimization/93683
707 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
709 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
711 * config/rs6000/predicates.md (cint34_operand): Rename the
712 -mprefixed-addr option to be -mprefixed.
713 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
714 the -mprefixed-addr option to be -mprefixed.
715 (OTHER_FUTURE_MASKS): Likewise.
716 (POWERPC_MASKS): Likewise.
717 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
718 the -mprefixed-addr option to be -mprefixed. Change error
719 messages to refer to -mprefixed.
720 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
722 (rs6000_legitimate_offset_address_p): Likewise.
723 (rs6000_mode_dependent_address): Likewise.
724 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
725 "-mprefixed" for target attributes and pragmas.
726 (address_to_insn_form): Rename the -mprefixed-addr option to be
728 (rs6000_adjust_insn_length): Likewise.
729 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
730 -mprefixed-addr option to be -mprefixed.
731 (ASM_OUTPUT_OPCODE): Likewise.
732 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
733 -mprefixed-addr option to be -mprefixed.
734 * config/rs6000/rs6000.opt (-mprefixed): Rename the
735 -mprefixed-addr option to be prefixed. Change the option from
736 being undocumented to being documented.
737 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
738 -mprefixed option. Update the -mpcrel documentation to mention
741 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
743 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
744 including FIRST_PSEUDO_REGISTER - 1.
745 * ira-color.c (print_hard_reg_set): Ditto.
747 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
749 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
750 (USTERNOP_QUALIFIERS): New define.
751 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
752 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
753 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
754 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
755 * config/arm/arm_neon.h (vusdot_s32): New.
756 (vusdot_lane_s32): New.
757 (vusdotq_lane_s32): New.
758 (vsudot_lane_s32): New.
759 (vsudotq_lane_s32): New.
760 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
761 * config/arm/iterators.md (DOTPROD_I8MM): New.
762 (sup, opsuffix): Add <us/su>.
763 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
764 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
766 2020-02-11 Richard Biener <rguenther@suse.de>
768 PR tree-optimization/93661
769 PR tree-optimization/93662
770 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
772 * tree-sra.c (get_access_for_expr): Likewise.
774 2020-02-10 Jakub Jelinek <jakub@redhat.com>
777 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
778 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
779 Change condition from TARGET_AVX2 to TARGET_AVX.
781 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
784 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
787 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
789 Try to generate zero-based comparisons.
790 * config/cris/cris.c (cris_reduce_compare): New function.
791 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
792 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
793 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
795 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
798 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
799 in Thumb state and also as a destination in Arm state. Add T16
802 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
804 * md.texi (Define Subst): Match closing paren in example.
806 2020-02-10 Jakub Jelinek <jakub@redhat.com>
810 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
811 arguments of strncmp.
813 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
816 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
817 but different source value.
818 (adjust_callers_for_value_intersection): New function.
819 (gather_edges_for_value): Adjust order of callers to let a
820 non-self-recursive caller be the first element.
821 (self_recursive_pass_through_p): Add a new parameter "simple", and
822 check generalized self-recursive pass-through jump function.
823 (self_recursive_agg_pass_through_p): Likewise.
824 (find_more_scalar_values_for_callers_subset): Compute value from
825 pass-through jump function for self-recursive.
826 (intersect_with_plats): Cleanup previous implementation code for value
827 itersection with self-recursive call edge.
828 (intersect_with_agg_replacements): Likewise.
829 (intersect_aggregates_with_edge): Deduce value from pass-through jump
830 function for self-recursive call edge. Cleanup previous implementation
831 code for value intersection with self-recursive call edge.
832 (decide_whether_version_node): Remove dead callers and adjust order
833 to let a non-self-recursive caller be the first element.
835 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
837 * recog.c: Move pass_split_before_sched2 code in front of
838 pass_split_before_regstack.
839 (pass_data_split_before_sched2): Rename pass to split3 from split4.
840 (pass_data_split_before_regstack): Rename pass to split4 from split3.
841 (rest_of_handle_split_before_sched2): Remove.
842 (pass_split_before_sched2::execute): Unconditionally call
844 (enable_split_before_sched2): New function.
845 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
846 (pass_split_before_regstack::gate): Ditto.
847 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
848 Update name check for renamed split4 pass.
849 * config/sh/sh.c (register_sh_passes): Update pass insertion
850 point for renamed split4 pass.
852 2020-02-09 Jakub Jelinek <jakub@redhat.com>
854 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
855 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
856 copying them around between host and target.
858 2020-02-08 Andrew Pinski <apinski@marvell.com>
861 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
862 STRICT_ALIGNMENT also.
864 2020-02-08 Jim Wilson <jimw@sifive.com>
867 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
869 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
870 Jakub Jelinek <jakub@redhat.com>
873 * config/i386/i386.h (CALL_USED_REGISTERS): Make
874 xmm16-xmm31 call-used even in 64-bit ms-abi.
876 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
878 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
879 (simd_ummla, simd_usmmla): Likewise.
880 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
881 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
884 2020-02-07 Richard Biener <rguenther@suse.de>
887 * tree-inline.c (fold_marked_statements): Do a PRE walk,
888 skipping unreachable regions.
889 (optimize_inline_calls): Skip folding stmts when we didn't
892 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
895 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
896 Don't return aggregates with only SFmode and DFmode in SSE
898 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
900 2020-02-07 Jakub Jelinek <jakub@redhat.com>
903 * config/rs6000/rs6000-logue.c
904 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
905 if it fails, move rs into end_addr and retry. Add
906 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
907 the insn pattern doesn't describe well what exactly happens to
911 * config/i386/predicates.md (avx_identity_operand): Remove.
912 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
913 (avx_<castmode><avxsizesuffix>_<castmode>,
914 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
915 a VEC_CONCAT of the operand and UNSPEC_CAST.
916 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
917 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
921 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
922 recog_data.insn if distance_non_agu_define changed it.
924 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
927 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
928 we only had X-FORM (reg+reg) addressing for vectors. Also before
929 ISA 3.0, we only had X-FORM addressing for scalars in the
930 traditional Altivec registers.
932 2020-02-06 <zhongyunde@huawei.com>
933 Vladimir Makarov <vmakarov@redhat.com>
935 PR rtl-optimization/93561
936 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
939 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
941 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
944 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
946 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
947 where the low and the high 32 bits are equal to each other specially,
948 with an rldimi instruction.
950 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
952 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
954 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
956 * config/arm/arm-tables.opt: Regenerate.
958 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
961 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
962 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
963 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
965 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
967 PR rtl-optimization/87763
968 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
970 2020-02-06 Delia Burduv <delia.burduv@arm.com>
972 * config/aarch64/aarch64-simd-builtins.def
973 (bfmlaq): New built-in function.
974 (bfmlalb): New built-in function.
975 (bfmlalt): New built-in function.
976 (bfmlalb_lane): New built-in function.
977 (bfmlalt_lane): New built-in function.
978 * config/aarch64/aarch64-simd.md
979 (aarch64_bfmmlaqv4sf): New pattern.
980 (aarch64_bfmlal<bt>v4sf): New pattern.
981 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
982 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
983 (vbfmlalbq_f32): New intrinsic.
984 (vbfmlaltq_f32): New intrinsic.
985 (vbfmlalbq_lane_f32): New intrinsic.
986 (vbfmlaltq_lane_f32): New intrinsic.
987 (vbfmlalbq_laneq_f32): New intrinsic.
988 (vbfmlaltq_laneq_f32): New intrinsic.
989 * config/aarch64/iterators.md (BF_MLA): New int iterator.
990 (bt): New int attribute.
992 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
994 * config/i386/i386.md (*pushtf): Emit "#" instead of
995 calling gcc_unreachable in insn output.
998 (*pushsf_rex64): Ditto for alternatives other than 1.
999 (*pushsf): Ditto for alternatives other than 1.
1001 2020-02-06 Martin Liska <mliska@suse.cz>
1003 PR gcov-profile/91971
1004 PR gcov-profile/93466
1005 * coverage.c (coverage_init): Revert mangling of
1006 path into filename. It can lead to huge filename length.
1007 Creation of subfolders seem more natural.
1009 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1012 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1013 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1014 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1016 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1019 * config/i386/predicates.md (avx_identity_operand): New predicate.
1020 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1021 define_insn_and_split.
1024 * omp-low.c (use_pointer_for_field): For nested constructs, also
1025 look for map clauses on target construct.
1026 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1027 taskreg_nesting_level.
1030 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1031 shared clause, call omp_notice_variable on outer context if any.
1033 2020-02-05 Jason Merrill <jason@redhat.com>
1036 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1037 non-zero address even if weak and not yet defined.
1039 2020-02-05 Martin Sebor <msebor@redhat.com>
1041 PR tree-optimization/92765
1042 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1043 * tree-ssa-strlen.c (compute_string_length): Remove.
1044 (determine_min_objsize): Remove.
1045 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1046 Avoid using type size as the upper bound on string length.
1047 (handle_builtin_string_cmp): Add an argument. Adjust.
1048 (strlen_check_and_optimize_call): Pass additional argument to
1049 handle_builtin_string_cmp.
1051 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1053 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1054 (*pushdi2_rex64 peephole2): Unconditionally split after
1056 (*ashl<mode>3_doubleword): Ditto.
1057 (*<shift_insn><mode>3_doubleword): Ditto.
1059 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1062 * config/rs6000/rs6000.c (get_vector_offset): Fix
1064 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1066 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1068 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1071 (Special Functions for Debugging the Analyzer): Update description
1072 of __analyzer_dump_exploded_nodes.
1074 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1077 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1078 include sets and not clobbers in the vzeroupper pattern.
1079 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1080 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1081 (*avx_vzeroupper_1): New define_insn_and_split.
1084 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1085 don't run when !optimize.
1086 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1089 2020-02-05 Richard Biener <rguenther@suse.de>
1092 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1093 checks before matching calls.
1095 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1097 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1098 function comment typo.
1101 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1102 simd_clone_create failed when i == 0, adjust clone->nargs by
1105 2020-02-05 Martin Liska <mliska@suse.cz>
1108 * doc/invoke.texi: Document that one should
1109 not combine ASLR and -fpch.
1111 2020-02-04 Richard Biener <rguenther@suse.de>
1113 PR tree-optimization/93538
1114 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1116 2020-02-04 Richard Biener <rguenther@suse.de>
1118 PR tree-optimization/91123
1119 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1120 (vn_walk_cb_data::last_vuse): New member.
1121 (vn_walk_cb_data::saved_operands): Likewsie.
1122 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1123 (vn_walk_cb_data::push_partial_def): Use finish.
1124 (vn_reference_lookup_2): Update last_vuse and use finish if
1125 we've saved operands.
1126 (vn_reference_lookup_3): Use finish and update calls to
1127 push_partial_defs everywhere. When translating through
1128 memcpy or aggregate copies save off operands and alias-set.
1129 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1130 operation for redundant store removal.
1132 2020-02-04 Richard Biener <rguenther@suse.de>
1134 PR tree-optimization/92819
1135 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1136 generating more stmts than before.
1138 2020-02-04 Martin Liska <mliska@suse.cz>
1140 * config/arm/arm.c (arm_gen_far_branch): Move the function
1141 outside of selftests.
1143 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1145 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1146 function to adjust PC-relative vector addresses.
1147 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1148 handle vectors with PC-relative addresses.
1150 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1152 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1154 (hard_reg_and_mode_to_addr_mask): Delete.
1155 (rs6000_adjust_vec_address): If the original vector address
1156 was REG+REG or REG+OFFSET and the element is not zero, do the add
1157 of the elements in the original address before adding the offset
1158 for the vector element. Use address_to_insn_form to validate the
1159 address using the register being loaded, rather than guessing
1160 whether the address is a DS-FORM or DQ-FORM address.
1162 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1164 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1165 to calculate the offset in memory from the start of a vector of a
1166 particular element. Add code to keep the element number in
1167 bounds if the element number is variable.
1168 (rs6000_adjust_vec_address): Move calculation of offset of the
1169 vector element to get_vector_offset.
1170 (rs6000_split_vec_extract_var): Do not do the initial AND of
1171 element here, move the code to get_vector_offset.
1173 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1175 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1178 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1180 * config/rs6000/constraints.md: Improve documentation.
1182 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1185 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1186 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1188 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1190 * config.gcc: Remove "carrizo" support.
1191 * config/gcn/gcn-opts.h (processor_type): Likewise.
1192 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1193 * config/gcn/gcn.opt (gpu_type): Likewise.
1194 * config/gcn/t-omp-device: Likewise.
1196 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1199 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1200 * config/arm/arm.c (arm_gen_far_branch): New function
1202 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1204 2020-02-03 Julian Brown <julian@codesourcery.com>
1205 Tobias Burnus <tobias@codesourcery.com>
1207 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1209 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1212 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1213 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1216 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1218 PR rtl-optimization/91333
1219 * ira-color.c (struct allocno_color_data): Add member
1221 (init_allocno_threads): Set the member up.
1222 (bucket_allocno_compare_func): Add compare hard reg
1225 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1227 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1229 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1230 * config.in: Regenerated.
1231 * configure: Regenerated.
1232 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1233 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1234 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1236 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1238 * configure: Regenerate.
1240 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1242 PR rtl-optimization/91333
1243 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1244 reg preferences comparison up.
1246 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1248 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1249 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1250 aarch64-sve-builtins-base.h.
1251 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1252 aarch64-sve-builtins-base.cc.
1253 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1254 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1256 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1257 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1258 (svcvtnt): New functions.
1259 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1260 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1261 (svcvtnt): New functions.
1262 (svcvt): Add a form that converts f32 to bf16.
1263 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1264 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1266 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1267 Treat B as bfloat16_t.
1268 (ternary_bfloat_lane_base): New class.
1269 (ternary_bfloat_def): Likewise.
1270 (ternary_bfloat): New shape.
1271 (ternary_bfloat_lane_def): New class.
1272 (ternary_bfloat_lane): New shape.
1273 (ternary_bfloat_lanex2_def): New class.
1274 (ternary_bfloat_lanex2): New shape.
1275 (ternary_bfloat_opt_n_def): New class.
1276 (ternary_bfloat_opt_n): New shape.
1277 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1278 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1279 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1280 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1281 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1282 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1283 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1284 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1285 the pattern off the narrow mode instead of the wider one.
1286 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1287 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1288 (sve_fp_op): Handle them.
1289 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1290 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1292 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1294 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1295 * config/aarch64/aarch64-modes.def (BF): Move definition before
1296 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1297 (SVE_MODES): Handle BF modes.
1298 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1300 (aarch64_full_sve_mode): Likewise.
1301 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1303 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1304 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1305 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1306 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1308 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1310 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1311 (TYPES_all_data): Add bf16.
1312 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1313 (register_tuple_type): Increase buffer size.
1314 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1315 (bf16): New type suffix.
1316 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1317 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1318 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1319 Change type from all_data to all_arith.
1320 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1323 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1324 Matthew Malcomson <matthew.malcomson@arm.com>
1325 Richard Sandiford <richard.sandiford@arm.com>
1327 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1328 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1329 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1330 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1331 __ARM_FEATURE_MATMUL_FP64.
1332 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1333 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1334 be disabled at the same time.
1335 (f32mm): New extension.
1336 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1337 (AARCH64_FL_F64MM): Bump to the next bit up.
1338 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1339 (TARGET_SVE_F64MM): New macros.
1340 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1341 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1342 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1343 (UNSPEC_ZIP2Q): New unspeccs.
1344 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1345 (optab, sur, perm_insn): Handle the new unspecs.
1346 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1347 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1348 TARGET_SVE_F64MM instead of separate tests.
1349 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1350 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1351 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1352 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1353 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1354 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1355 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1356 (TYPES_s_signed): New macro.
1357 (TYPES_s_integer): Use it.
1358 (TYPES_d_float): New macro.
1359 (TYPES_d_data): Use it.
1360 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1361 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1362 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1363 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1364 (svmmla): New shape.
1365 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1366 template parameters.
1367 (ternary_resize2_lane_base): Likewise.
1368 (ternary_resize2_base): New class.
1369 (ternary_qq_lane_base): Likewise.
1370 (ternary_intq_uintq_lane_def): Likewise.
1371 (ternary_intq_uintq_lane): New shape.
1372 (ternary_intq_uintq_opt_n_def): New class
1373 (ternary_intq_uintq_opt_n): New shape.
1374 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1375 (ternary_uintq_intq_def): New class.
1376 (ternary_uintq_intq): New shape.
1377 (ternary_uintq_intq_lane_def): New class.
1378 (ternary_uintq_intq_lane): New shape.
1379 (ternary_uintq_intq_opt_n_def): New class.
1380 (ternary_uintq_intq_opt_n): New shape.
1381 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1382 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1383 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1384 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1386 (svdotprod_lane_impl): ...this new class.
1387 (svmmla_impl, svusdot_impl): New classes.
1388 (svdot_lane): Update to use svdotprod_lane_impl.
1389 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1390 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1392 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1393 function, with no types defined.
1394 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1395 AARCH64_FL_I8MM functions.
1396 (svmmla): New AARCH64_FL_F32MM function.
1397 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1398 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1399 AARCH64_FL_F64MM function.
1400 (REQUIRED_EXTENSIONS):
1402 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1404 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1407 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1409 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1410 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1411 (*movti_internal): Do not check for
1412 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1413 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1414 just after check for TARGET_AVX.
1415 (*movdf_internal): Ditto.
1416 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1417 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1418 * config/i386/sse.md (mov<mode>_internal): Only check
1419 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1420 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1421 (<sse>_andnot<mode>3<mask_name>): Move check for
1422 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1423 (<code><mode>3<mask_name>): Ditto.
1424 (*andnot<mode>3): Ditto.
1425 (*andnottf3): Ditto.
1426 (*<code><mode>3): Ditto.
1427 (*<code>tf3): Ditto.
1428 (*andnot<VI:mode>3): Remove
1429 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1430 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1431 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1432 (sse4_1_blendv<ssemodesuffix>): Ditto.
1433 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1434 Explain that tune applies to 128bit instructions only.
1436 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1438 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1439 to definition of hsa_kernel_description. Parse assembly to find SGPR
1440 and VGPR count of kernel and store in hsa_kernel_description.
1442 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1444 PR rtl-optimization/91838
1445 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1446 to truncate if allowed or reject combination.
1448 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1450 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1451 (find_inv_vars_cb): Likewise.
1453 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1455 * calls.c (special_function_p): Split out the check for DECL_NAME
1456 being non-NULL and fndecl being extern at file scope into a
1457 new maybe_special_function_p and call it. Drop check for fndecl
1458 being non-NULL that was after a usage of DECL_NAME (fndecl).
1459 * tree.h (maybe_special_function_p): New inline function.
1461 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1463 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1464 (mask_gather_load<mode>): ... here, and zero-initialize the
1466 (maskload<mode>di): Zero-initialize the destination.
1469 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1472 * doc/analyzer.texi (Limitations): Note that constraints on
1473 floating-point values are currently ignored.
1475 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1478 * symtab.c (symtab_node::noninterposable_alias): If localalias
1479 already exists, but is not usable, append numbers after it until
1480 a unique name is found. Formatting fix.
1483 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1486 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1488 * config/gcn/gcn.c (print_operand): Handle LTGT.
1489 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1491 2020-01-30 Richard Biener <rguenther@suse.de>
1493 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1494 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1496 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1498 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1499 without a DECL in .data.rel.ro.local.
1501 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1504 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1508 * config/i386/sse.md
1509 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1510 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1511 any_extend code iterator instead of always zero_extend.
1512 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1513 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1514 Use any_extend code iterator instead of always zero_extend.
1515 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1516 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1517 Use any_extend code iterator instead of always zero_extend.
1518 (*sse2_pmovmskb_ext): New define_insn.
1519 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1522 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1523 (*popcountsi2_zext_falsedep): New define_insn.
1525 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1527 * config.in: Regenerated.
1528 * configure: Regenerated.
1530 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1533 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1534 LLVM's assembler changed the default in version 9.
1536 2020-01-24 Jeff Law <law@redhat.com>
1538 PR tree-optimization/89689
1539 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1541 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1545 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1547 PR rtl-optimization/87763
1548 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1549 simplification to handle subregs as well as bare regs.
1550 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1552 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1555 * ira.c (ira): Revert use of simplified LRA algorithm.
1557 2020-01-29 Martin Jambor <mjambor@suse.cz>
1559 PR tree-optimization/92706
1560 * tree-sra.c (struct access): Fields first_link, last_link,
1561 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1562 next_rhs_queued and grp_rhs_queued respectively, new fields
1563 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1564 (struct assign_link): Field next renamed to next_rhs, new field
1565 next_lhs. Updated comment.
1566 (work_queue_head): Renamed to rhs_work_queue_head.
1567 (lhs_work_queue_head): New variable.
1568 (add_link_to_lhs): New function.
1569 (relink_to_new_repr): Also relink LHS lists.
1570 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1571 (add_access_to_lhs_work_queue): New function.
1572 (pop_access_from_work_queue): Renamed to
1573 pop_access_from_rhs_work_queue.
1574 (pop_access_from_lhs_work_queue): New function.
1575 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1577 (child_would_conflict_in_lacc): Renamed to
1578 child_would_conflict_in_acc. Adjusted parameter names.
1579 (create_artificial_child_access): New parameter set_grp_read, use it.
1580 (subtree_mark_written_and_enqueue): Renamed to
1581 subtree_mark_written_and_rhs_enqueue.
1582 (propagate_subaccesses_across_link): Renamed to
1583 propagate_subaccesses_from_rhs.
1584 (propagate_subaccesses_from_lhs): New function.
1585 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1588 2020-01-29 Martin Jambor <mjambor@suse.cz>
1590 PR tree-optimization/92706
1591 * tree-sra.c (struct access): Adjust comment of
1592 grp_total_scalarization.
1593 (find_access_in_subtree): Look for single children spanning an entire
1595 (scalarizable_type_p): Allow register accesses, adjust callers.
1596 (completely_scalarize): Remove function.
1597 (scalarize_elem): Likewise.
1598 (create_total_scalarization_access): Likewise.
1599 (sort_and_splice_var_accesses): Do not track total scalarization
1601 (analyze_access_subtree): New parameter totally, adjust to new meaning
1602 of grp_total_scalarization.
1603 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1604 (can_totally_scalarize_forest_p): New function.
1605 (create_total_scalarization_access): Likewise.
1606 (create_total_access_and_reshape): Likewise.
1607 (total_should_skip_creating_access): Likewise.
1608 (totally_scalarize_subtree): Likewise.
1609 (analyze_all_variable_accesses): Perform total scalarization after
1610 subaccess propagation using the new functions above.
1611 (initialize_constant_pool_replacements): Output initializers by
1612 traversing the access tree.
1614 2020-01-29 Martin Jambor <mjambor@suse.cz>
1616 * tree-sra.c (verify_sra_access_forest): New function.
1617 (verify_all_sra_access_forests): Likewise.
1618 (create_artificial_child_access): Set parent.
1619 (analyze_all_variable_accesses): Call the verifier.
1621 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1623 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1624 if called on indirect edge.
1625 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1626 speculative call if needed.
1628 2020-01-29 Richard Biener <rguenther@suse.de>
1630 PR tree-optimization/93428
1631 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1632 permutation when the load node is created.
1633 (vect_analyze_slp_instance): Re-use it here.
1635 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1637 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1639 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1641 PR rtl-optimization/93272
1642 * ira-lives.c (process_out_of_region_eh_regs): New function.
1643 (process_bb_node_lives): Call it.
1645 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1647 * coverage.c (read_counts_file): Make error message lowercase.
1649 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1651 * profile-count.c (profile_quality_display_names): Fix ordering.
1653 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1656 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1657 hash only when edge is first within the sequence.
1658 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1659 (symbol_table::create_edge): Do not set target_prob.
1660 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1662 (cgraph_edge::make_speculative): Drop target_prob parameter.
1663 (cgraph_edge::speculative_call_info): Remove.
1664 (cgraph_edge::first_speculative_call_target): New member function.
1665 (update_call_stmt_hash_for_removing_direct_edge): New function.
1666 (cgraph_edge::resolve_speculation): Rewrite to new API.
1667 (cgraph_edge::speculative_call_for_target): New member function.
1668 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1669 multiple speculation targets.
1670 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1672 (verify_speculative_call): Verify that targets form an interval.
1673 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1674 (cgraph_edge::first_speculative_call_target): New member function.
1675 (cgraph_edge::next_speculative_call_target): New member function.
1676 (cgraph_edge::speculative_call_target_ref): New member function.
1677 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1678 (cgraph_edge): Remove target_prob.
1679 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1680 Fix handling of speculative calls.
1681 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1682 * ipa-fnsummary.c (analyze_function_body): Likewise.
1683 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1684 * ipa-profile.c (dump_histogram): Fix formating.
1685 (ipa_profile_generate_summary): Watch for overflows.
1686 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1687 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1688 (update_indirect_edges_after_inlining): Update to new API.
1689 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1691 * profile-count.h: (profile_probability::adjusted): New.
1692 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1693 updating of profile.
1694 * value-prof.c (gimple_ic_transform): Rename to ...
1695 (dump_ic_profile): ... this one; update dumping.
1696 (stream_in_histogram_value): Fix formating.
1697 (gimple_value_profile_transformations): Update.
1699 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1702 * config/i386/i386.md (*movoi_internal_avx): Remove
1703 TARGET_SSE_TYPELESS_STORES check.
1704 (*movti_internal): Prefer TARGET_AVX over
1705 TARGET_SSE_TYPELESS_STORES.
1706 (*movtf_internal): Likewise.
1707 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1708 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1709 from TARGET_SSE_TYPELESS_STORES.
1711 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1713 * diagnostic-core.h (warning_at): Rename overload to...
1714 (warning_meta): ...this.
1715 (emit_diagnostic_valist): Delete decl of overload taking
1716 diagnostic_metadata.
1717 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1718 (warning_at): Rename overload taking diagnostic_metadata to...
1719 (warning_meta): ...this.
1721 2020-01-28 Richard Biener <rguenther@suse.de>
1723 PR tree-optimization/93439
1724 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1725 * tree-cfg.c (move_sese_region_to_fn): ... here.
1726 (verify_types_in_gimple_reference): Verify used cliques are
1729 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1732 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1733 argument of a pointer to struct gcc_options and pass it to
1734 parse_mtune_ctrl_str.
1735 (ix86_function_specific_restore): Pass opts to
1736 set_ix86_tune_features.
1737 (ix86_option_override_internal): Likewise.
1738 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1739 gcc_options and use it for x_ix86_tune_ctrl_string.
1741 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1743 PR rtl-optimization/87763
1744 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1745 simplification to handle subregs as well as bare regs.
1746 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1748 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1750 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1751 for reduction chains that (now) include a call.
1753 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1755 PR tree-optimization/92822
1756 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1757 out the don't-care elements of a vector whose significant elements
1758 are duplicates, make the don't-care elements duplicates too.
1760 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1762 PR tree-optimization/93434
1763 * tree-predcom.c (split_data_refs_to_components): Record which
1764 components have had aliasing loads removed. Prevent store-store
1765 commoning for all such components.
1767 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1770 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1771 -1 or is_vshift is true, use new_vector with number of elts npatterns
1772 rather than new_unary_operation.
1774 PR tree-optimization/93454
1775 * gimple-fold.c (fold_array_ctor_reference): Perform
1776 elt_size.to_uhwi () just once, instead of calling it in every
1777 iteration. Punt if that value is above size of the temporary
1778 buffer. Decrease third native_encode_expr argument when
1779 bufoff + elt_sz is above size of buf.
1781 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1783 * config/mips/mips.c (mips_declare_object_name)
1784 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1786 2020-01-27 Martin Liska <mliska@suse.cz>
1788 PR gcov-profile/93403
1789 * tree-profile.c (gimple_init_gcov_profiler): Generate
1790 both __gcov_indirect_call_profiler_v4 and
1791 __gcov_indirect_call_profiler_v4_atomic.
1793 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1796 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1798 (@aarch64_split_simd_mov<mode>): Use it.
1799 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1800 Leave the vec_extract patterns to handle 2-element vectors.
1801 (aarch64_simd_mov_from_<mode>high): Likewise.
1802 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1803 (vec_extractv2dfv1df): Likewise.
1805 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1807 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1808 jump conditions for *compare_condjump<GPI:mode>.
1810 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1813 * digraph.cc (test_edge::test_edge): Specify template for base
1816 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1818 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1820 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1822 * config/arc/arc-protos.h (gen_mlo): Remove.
1823 (gen_mhi): Likewise.
1824 * config/arc/arc.c (AUX_MULHI): Define.
1825 (arc_must_save_reister): Special handling for r58/59.
1826 (arc_compute_frame_size): Consider mlo/mhi registers.
1827 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1829 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1830 mlo/mhi name selection.
1831 (arc_restore_callee_saves): Don't early restore blink when ISR.
1832 (arc_expand_prologue): Add mlo/mhi saving.
1833 (arc_expand_epilogue): Add mlo/mhi restoring.
1836 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1837 numbering when MUL64 option is used.
1838 (DWARF2_FRAME_REG_OUT): Define.
1839 * config/arc/arc.md (arc600_stall): New pattern.
1840 (VUNSPEC_ARC_ARC600_STALL): Define.
1841 (mulsi64): Use correct mlo/mhi registers.
1842 (mulsi_600): Clean it up.
1843 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1845 (mhi_operand): Likewise.
1847 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1848 Petro Karashchenko <petro.karashchenko@ring.com>
1850 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1851 attributes if needed.
1852 (prepare_move_operands): Generate special unspec instruction for
1854 (arc_isuncached_mem_p): Propagate uncached attribute to each
1856 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1857 (VUNSPEC_ARC_STDI): Likewise.
1858 (ALLI): New mode iterator.
1859 (mALLI): New mode attribute.
1860 (lddi): New instruction pattern.
1862 (stdidi_split): Split instruction for architectures which are not
1863 supporting ll64 option.
1864 (lddidi_split): Likewise.
1866 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1868 PR rtl-optimization/92989
1869 * lra-lives.c (process_bb_lives): Update the live-in set before
1870 processing additional clobbers.
1872 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1874 PR rtl-optimization/93170
1875 * cselib.c (cselib_invalidate_regno_val): New function, split out
1877 (cselib_invalidate_regno): ...here.
1878 (cselib_invalidated_by_call_p): New function.
1879 (cselib_process_insn): Iterate over all the hard-register entries in
1880 REG_VALUES and invalidate any that cross call-clobbered registers.
1882 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1884 * dojump.c (split_comparison): Use HONOR_NANS rather than
1885 HONOR_SNANS when splitting LTGT.
1887 2020-01-27 Martin Liska <mliska@suse.cz>
1890 * opts.c (print_filtered_help): Exclude language-specific
1891 options from --help=common unless enabled in all FEs.
1893 2020-01-27 Martin Liska <mliska@suse.cz>
1895 * opts.c (print_help): Exclude params from
1896 all except --help=param.
1898 2020-01-27 Martin Liska <mliska@suse.cz>
1901 * config/i386/i386-features.c (make_resolver_func):
1902 Align the code with ppc64 target implementation.
1903 Do not generate a unique name for resolver function.
1905 2020-01-27 Richard Biener <rguenther@suse.de>
1907 PR tree-optimization/93397
1908 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1909 converted reduction chain SLP graph adjustment.
1911 2020-01-26 Marek Polacek <polacek@redhat.com>
1914 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1917 2020-01-26 Jason Merrill <jason@redhat.com>
1920 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1923 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1925 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1926 (rx_setmem): Likewise.
1928 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1931 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1932 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1933 drop <di> from constraint of last operand.
1936 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1937 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1938 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1940 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1943 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1946 2020-01-24 Jeff Law <law@redhat.com>
1948 PR tree-optimization/92788
1949 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1952 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1955 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1956 *avx_vperm_broadcast_<mode>,
1957 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1958 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1959 Move before avx2_perm<mode>/avx512f_perm<mode>.
1962 * simplify-rtx.c (simplify_const_unary_operation,
1963 simplify_const_binary_operation): Punt for mode precision above
1964 MAX_BITSIZE_MODE_ANY_INT.
1966 2020-01-24 Andrew Pinski <apinski@marvell.com>
1968 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1971 2020-01-24 Jeff Law <law@redhat.com>
1974 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1975 for REGs. Call output_operand_lossage to get more reasonable
1978 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1980 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1981 gcn_fp_compare_operator.
1982 (vec_cmpu<mode>di): Use gcn_compare_operator.
1983 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1984 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1985 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1986 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1987 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1988 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1989 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1990 gcn_fp_compare_operator.
1991 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1992 gcn_fp_compare_operator.
1993 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1994 gcn_fp_compare_operator.
1995 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1996 gcn_fp_compare_operator.
1998 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2000 * doc/install.texi (Cross-Compiler-Specific Options): Document
2001 `--with-toolexeclibdir' option.
2003 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2005 * target.def (flags_regnum): Also mention effect on delay slot filling.
2006 * doc/tm.texi: Regenerate.
2008 2020-01-23 Jeff Law <law@redhat.com>
2010 PR translation/90162
2011 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2013 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2016 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2019 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2021 PR rtl-optimization/93402
2022 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2025 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2027 * config.in: Regenerated.
2028 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2029 for TARGET_LIBC_GNUSTACK.
2030 * configure: Regenerated.
2031 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2032 found to be 2.31 or greater.
2034 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2036 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2038 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2039 (mips_asm_file_end): New function. Delegate to
2040 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2041 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2043 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2046 * config/i386/i386-modes.def (POImode): New mode.
2047 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2048 * config/i386/i386.md (DPWI): New mode attribute.
2049 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2051 (QPWI): ... this. Use POI instead of OI for TImode.
2052 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2053 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2056 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2059 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2061 (speculation_tracker_rev): New pattern.
2062 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2063 Use speculation_tracker_rev to track the inverse condition.
2065 2020-01-23 Richard Biener <rguenther@suse.de>
2067 PR tree-optimization/93381
2068 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2069 alias-set of the def as argument and record the first one.
2070 (vn_walk_cb_data::first_set): New member.
2071 (vn_reference_lookup_3): Pass the alias-set of the current def
2072 to push_partial_def. Fix alias-set used in the aggregate copy
2074 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2075 * real.c (clear_significand_below): Fix out-of-bound access.
2077 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2080 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2081 New define_insn patterns.
2083 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2085 * doc/sourcebuild.texi (check-function-bodies): Add an
2086 optional target/xfail selector.
2088 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2090 PR rtl-optimization/93124
2091 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2092 bare USE and CLOBBER insns.
2094 2020-01-22 Andrew Pinski <apinski@marvell.com>
2096 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2098 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2101 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2102 diagnostic_manager into "ana" namespace.
2103 * selftest-run-tests.c (selftest::run_tests): Update for move of
2104 selftest::run_analyzer_selftests to
2105 ana::selftest::run_analyzer_selftests.
2107 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2109 * cfgexpand.c (union_stack_vars): Update the size.
2111 2020-01-22 Richard Biener <rguenther@suse.de>
2113 PR tree-optimization/93381
2114 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2115 throughout, handle all conversions the same.
2117 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2120 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2121 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2122 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2123 Call force_reg on high_in2 unconditionally.
2125 2020-01-22 Martin Liska <mliska@suse.cz>
2127 PR tree-optimization/92924
2128 * profile.c (compute_value_histograms): Divide
2131 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2134 * output.h (assemble_name_resolve): Declare.
2135 * varasm.c (assemble_name_resolve): New function.
2136 (assemble_name): Use it.
2137 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2139 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2141 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2142 update_web_docs_git instead of update_web_docs_svn.
2144 2020-01-21 Andrew Pinski <apinski@marvell.com>
2147 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2148 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2149 (*tlsgd_small_<mode>): Likewise.
2150 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2151 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2152 register. Convert that register back to dest using convert_mode.
2154 2020-01-21 Jim Wilson <jimw@sifive.com>
2156 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2159 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2160 Uros Bizjak <ubizjak@gmail.com>
2163 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2165 (legitimize_tls_address): Do GNU2 TLS address computation in
2166 ptr_mode and zero-extend result to Pmode.
2167 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2168 :P with :PTR and Pmode with ptr_mode.
2169 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2170 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2171 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2173 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2176 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2177 the last two operands are CONST_INT_P before using them as such.
2179 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2181 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2182 to get the integer element types.
2184 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2186 * config/aarch64/aarch64-sve-builtins.h
2187 (function_expander::convert_to_pmode): Declare.
2188 * config/aarch64/aarch64-sve-builtins.cc
2189 (function_expander::convert_to_pmode): New function.
2190 (function_expander::get_contiguous_base): Use it.
2191 (function_expander::prepare_gather_address_operands): Likewise.
2192 * config/aarch64/aarch64-sve-builtins-sve2.cc
2193 (svwhilerw_svwhilewr_impl::expand): Likewise.
2195 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2198 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2199 cfun->machine->label_is_assembled.
2200 (aarch64_print_patchable_function_entry): New.
2201 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2202 * config/aarch64/aarch64.h (struct machine_function): New field,
2205 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2208 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2211 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2214 * cgraph.c (cgraph_edge::resolve_speculation,
2215 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2216 call_stmt_site_hash.
2218 2020-01-21 Martin Liska <mliska@suse.cz>
2220 * config/rs6000/rs6000.c (common_mode_defined): Remove
2223 2020-01-21 Richard Biener <rguenther@suse.de>
2225 PR tree-optimization/92328
2226 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2227 type when value-numbering same-sized store by inserting a
2229 (eliminate_dom_walker::eliminate_stmt): When eliminating
2230 a redundant store handle bit-reinterpretation of the same value.
2232 2020-01-21 Andrew Pinski <apinski@marvel.com>
2235 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2237 (prepare_block_for_update): This. Use a worklist instead of
2240 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2242 * gcc/config/arm/arm.c (clear_operation_p):
2243 Initialise last_regno, skip first iteration
2244 based on the first_set value and use ints instead
2245 of the unnecessary HOST_WIDE_INTs.
2247 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2250 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2251 compare_mode other than SFmode or DFmode.
2253 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2256 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2257 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2258 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2260 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2262 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2264 2020-01-20 Andrew Pinski <apinski@marvell.com>
2267 * targhooks.c (default_print_patchable_function_entry): Use
2268 output_asm_insn to emit the nop instruction.
2270 2020-01-20 Fangrui Song <maskray@google.com>
2273 * targhooks.c (default_print_patchable_function_entry): Align to
2276 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2279 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2280 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2281 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2282 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2283 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2284 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2285 Remove the {q} suffix from lea.
2286 (*tls_dynamic_gnu2_call_64): Renamed to ...
2287 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2288 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2289 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2290 Pass Pmode to gen_tls_dynamic_gnu2_64.
2292 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2294 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2296 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2298 * config/aarch64/aarch64-sve-builtins-base.cc
2299 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2301 2020-01-20 Richard Biener <rguenther@suse.de>
2304 * dwarf2out.c (prune_unused_types): Unconditionally mark
2305 called function DIEs.
2307 2020-01-20 Martin Liska <mliska@suse.cz>
2309 PR tree-optimization/93199
2310 * tree-eh.c (struct leh_state): Add
2311 new field outer_non_cleanup.
2312 (cleanup_is_dead_in): Pass leh_state instead
2313 of eh_region. Add a checking that state->outer_non_cleanup
2314 points to outer non-clean up region.
2315 (lower_try_finally): Record outer_non_cleanup
2317 (lower_catch): Likewise.
2318 (lower_eh_filter): Likewise.
2319 (lower_eh_must_not_throw): Likewise.
2320 (lower_cleanup): Likewise.
2322 2020-01-20 Richard Biener <rguenther@suse.de>
2324 PR tree-optimization/93094
2325 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2326 (vect_transform_loop): Likewise.
2327 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2328 loop_vectorized_call to vect_transform_loop.
2329 * tree-vect-loop.c (vect_transform_loop): Pass down
2330 loop_vectorized_call to vect_loop_versioning.
2331 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2332 the earlier discovered loop_vectorized_call.
2334 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2336 * doc/contribute.texi: Update for SVN -> Git transition.
2337 * doc/install.texi: Likewise.
2339 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2342 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2343 speculative targets.
2344 (verify_speculative_call): New function
2345 (cgraph_node::verify_node): Use it.
2346 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2349 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2352 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2353 (cgraph_edge::make_direct): Remove all indirect targets.
2354 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2355 (cgraph_node::verify_node): Verify that only one call_stmt or
2356 lto_stmt_uid is set.
2357 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2359 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2360 (lto_output_ref): Simplify streaming of stmt.
2361 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2363 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2365 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2366 Mark parameter unused.
2368 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2370 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2372 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2374 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2376 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2378 * Makefile.in: Add coroutine-passes.o.
2379 * builtin-types.def (BT_CONST_SIZE): New.
2380 (BT_FN_BOOL_PTR): New.
2381 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2382 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2383 * coroutine-builtins.def: New file.
2384 * coroutine-passes.cc: New file.
2385 * function.h (struct GTY function): Add a bit to indicate that the
2386 function is a coroutine component.
2387 * internal-fn.c (expand_CO_FRAME): New.
2388 (expand_CO_YIELD): New.
2389 (expand_CO_SUSPN): New.
2390 (expand_CO_ACTOR): New.
2391 * internal-fn.def (CO_ACTOR): New.
2395 * passes.def: Add pass_coroutine_lower_builtins,
2396 pass_coroutine_early_expand_ifns.
2397 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2398 (make_pass_coroutine_early_expand_ifns): New.
2399 * doc/invoke.texi: Document the fcoroutines command line
2402 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2404 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2407 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2408 after checking the argument is a REG. Don't use REGNO (reg)
2409 again to set last_regno, reuse regno variable instead.
2411 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2413 * doc/analyzer.texi (Limitations): Add note about NaN.
2415 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2416 Sudakshina Das <sudi.das@arm.com>
2418 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2419 and valid immediate.
2420 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2421 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2422 * config/arm/constraints.md (Pg): New.
2423 * config/arm/predicates.md (long_shift_imm): New.
2424 (arm_reg_or_long_shift_imm): Likewise.
2425 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2426 (thumb2_lsll): Likewise.
2429 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2430 Sudakshina Das <sudi.das@arm.com>
2432 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2433 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2434 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2435 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2436 * config/arm/thumb2.md (thumb2_asrl): New.
2437 (thumb2_lsll): Likewise.
2439 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2441 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2444 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2446 * gdbinit.in (help-gcc-hooks): New command.
2447 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2448 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2451 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2453 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2454 correct target macro.
2456 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2458 * config/aarch64/aarch64-protos.h
2459 (aarch64_sve_ld1ro_operand_p): New.
2460 * config/aarch64/aarch64-sve-builtins-base.cc
2461 (class load_replicate): New.
2462 (class svld1ro_impl): New.
2463 (class svld1rq_impl): Change to inherit from load_replicate.
2464 (svld1ro): New sve intrinsic function base.
2465 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2466 New DEF_SVE_FUNCTION.
2467 * config/aarch64/aarch64-sve-builtins-base.h
2468 (svld1ro): New decl.
2469 * config/aarch64/aarch64-sve-builtins.cc
2470 (function_expander::add_mem_operand): Modify assert to allow
2472 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2474 * config/aarch64/aarch64.c
2475 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2476 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2477 (aarch64_sve_ld1ro_operand_p): New.
2478 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2479 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2480 * config/aarch64/predicates.md
2481 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2483 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2485 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2486 Introduce this ACLE specified predefined macro.
2487 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2488 (fp): Disabling this disables f64mm.
2489 (simd): Disabling this disables f64mm.
2490 (fp16): Disabling this disables f64mm.
2491 (sve): Disabling this disables f64mm.
2492 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2493 (AARCH64_ISA_F64MM): New.
2494 (TARGET_F64MM): New.
2495 * doc/invoke.texi (f64mm): Document new option.
2497 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2499 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2500 (neoversen1_tunings): Likewise.
2502 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2505 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2506 Add assert to ensure prolog has been emitted.
2507 (aarch64_split_atomic_op): Likewise.
2508 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2509 Use epilogue_completed rather than reload_completed.
2510 (aarch64_atomic_exchange<mode>): Likewise.
2511 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2512 (atomic_nand<mode>): Likewise.
2513 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2514 (atomic_fetch_nand<mode>): Likewise.
2515 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2516 (atomic_nand_fetch<mode>): Likewise.
2518 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2521 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2523 (REVERSE_CONDITION): Delete.
2524 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2525 (CCFP_CCFPE): Likewise.
2526 (e): New mode attribute.
2527 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2528 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2529 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2530 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2531 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2532 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2533 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2534 name of generator from gen_ccmpdi to gen_ccmpccdi.
2535 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2536 the previous comparison but aren't able to, use the new ccmp_rev
2539 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2541 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2542 than testing directly for INTEGER_CST.
2543 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2545 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2547 PR tree-optimization/93292
2548 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2549 get_vectype_for_scalar_type returns NULL.
2551 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2553 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2554 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2556 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2558 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2560 * params.opt: (max-predicted-iterations): Set bounds.
2561 * predict.c (real_almost_one, real_br_prob_base,
2562 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2563 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2564 probabilities; do not truncate to reg_br_prob_bases.
2565 (estimate_loops_at_level): Pass max_cyclic_prob.
2566 (estimate_loops): Compute max_cyclic_prob.
2567 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2569 * profile-count.c (profile_probability::to_sreal): New.
2570 * profile-count.h (class sreal): Move up in file.
2571 (profile_probability::to_sreal): Declare.
2573 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2576 (arm_invalid_conversion): New function for target hook.
2577 (arm_invalid_unary_op): New function for target hook.
2578 (arm_invalid_binary_op): New function for target hook.
2580 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2582 * config.gcc: Add arm_bf16.h.
2583 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2584 (arm_simd_builtin_std_type): Add BFmode.
2585 (arm_init_simd_builtin_types): Define element types for vector types.
2586 (arm_init_bf16_types): New function.
2587 (arm_init_builtins): Add arm_init_bf16_types function call.
2588 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2589 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2590 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2591 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2592 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2593 (arm_mangle_type): Add __bf16.
2594 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2595 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2596 arm_bf16_ptr_type_node.
2597 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2598 define_split between ARM registers.
2599 * config/arm/arm_bf16.h: New file.
2600 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2601 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2603 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2604 * config/arm/vfp.md: Add BFmode to movhf patterns.
2606 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2607 Andre Vieira <andre.simoesdiasvieira@arm.com>
2609 * config/arm/arm-cpus.in (mve, mve_float): New features.
2610 (dsp, mve, mve.fp): New options.
2611 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2612 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2613 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2615 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2616 Thomas Preud'homme <thomas.preudhomme@arm.com>
2618 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2620 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2621 error for using -mcmse when targeting Armv8.1-M Mainline.
2623 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2624 Thomas Preud'homme <thomas.preudhomme@arm.com>
2626 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2627 address in r4 when targeting Armv8.1-M Mainline.
2628 (nonsecure_call_value_internal): Likewise.
2629 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2630 a register match_operand again. Emit BLXNS when targeting
2632 (nonsecure_call_value_reg_thumb2): Likewise.
2634 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2635 Thomas Preud'homme <thomas.preudhomme@arm.com>
2637 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2638 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2639 variable as true when floating-point ABI is not hard. Replace
2640 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2641 Generate VLSTM and VLLDM instruction respectively before and
2642 after a function call to cmse_nonsecure_call function.
2643 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2644 (VUNSPEC_VLLDM): Likewise.
2645 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2646 (lazy_load_multiple_insn): Likewise.
2648 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2649 Thomas Preud'homme <thomas.preudhomme@arm.com>
2651 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2652 (arm_emit_vfp_multi_reg_pop): Likewise.
2653 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2654 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2655 restore callee-saved VFP registers.
2657 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2658 Thomas Preud'homme <thomas.preudhomme@arm.com>
2660 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2661 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2662 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2663 callee-saved GPRs as well as clear ip register before doing a nonsecure
2664 call then restore callee-saved GPRs after it when targeting
2666 (arm_reorg): Adapt to function rename.
2668 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2669 Thomas Preud'homme <thomas.preudhomme@arm.com>
2671 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2672 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2673 clear_vfp_multiple pattern based on a new vfp parameter.
2674 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2675 targeting Armv8.1-M Mainline.
2676 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2677 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2678 whether VFP registers are available before looking call_used_regs for a
2680 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2681 of prototype of clear_operation_p.
2682 (clear_vfp_multiple_operation): New predicate.
2683 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2684 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2686 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2687 Thomas Preud'homme <thomas.preudhomme@arm.com>
2689 * config/arm/arm-protos.h (clear_operation_p): Declare.
2690 * config/arm/arm.c (clear_operation_p): New function.
2691 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2692 targeting Armv8.1-M Mainline or successor.
2693 (output_return_instruction): Only output APSR register clearing if
2694 Armv8.1-M Mainline instructions not available.
2695 (thumb_exit): Likewise.
2696 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2697 * config/arm/thumb2.md (clear_apsr): New define_insn.
2698 (clear_multiple): Likewise.
2699 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2701 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2702 Thomas Preud'homme <thomas.preudhomme@arm.com>
2704 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2705 (use_return_insn): Also return false for Armv8.1-M Mainline.
2706 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2707 Mainline instructions are available.
2708 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2709 when targeting Armv8.1-M Mainline Security Extensions.
2710 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2711 Mainline entry function.
2712 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2713 targeting Armv8.1-M Mainline or successor.
2714 (arm_expand_epilogue): Fix indentation of caller-saved register
2715 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2717 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2718 (FP_SYSREGS): Likewise.
2719 (enum vfp_sysregs_encoding): Define enum.
2720 (fp_sysreg_names): Declare.
2721 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2722 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2723 (pop_fpsysreg_insn): Likewise.
2725 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2726 Thomas Preud'homme <thomas.preudhomme@arm.com>
2728 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2729 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2730 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2731 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2732 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2733 (ARMv8_1m_main): New feature group.
2734 (armv8.1-m.main): New architecture.
2735 * config/arm/arm-tables.opt: Regenerate.
2736 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2737 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2738 (arm_options_perform_arch_sanity_checks): Error out when targeting
2739 Armv8.1-M Mainline Security Extensions.
2740 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2742 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2744 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2745 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2746 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2747 aarch64_bfdot_laneq): New.
2748 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2749 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2750 vbfdotq_laneq_f32): New.
2751 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2752 VBFMLA_W, VBF): New.
2753 (isquadop): Add V4BF, V8BF.
2755 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2757 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2758 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2759 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2760 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2761 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2762 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2763 usdot_laneq, sudot_lane,sudot_laneq): New.
2764 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2765 (aarch64_<sur>dot_lane): New.
2766 * config/aarch64/arm_neon.h (vusdot_s32): New.
2768 (vusdot_lane_s32): New.
2769 (vsudot_lane_s32): New.
2770 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2771 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2773 2020-01-16 Martin Liska <mliska@suse.cz>
2775 * value-prof.c (dump_histogram_value): Fix
2776 obvious spacing issue.
2778 2020-01-16 Andrew Pinski <apinski@marvell.com>
2780 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2781 !storage_order_barrier_p.
2783 2020-01-16 Andrew Pinski <apinski@marvell.com>
2785 * sched-int.h (_dep): Add unused bit-field field for the padding.
2786 * sched-deps.c (init_dep_1): Init unused field.
2788 2020-01-16 Andrew Pinski <apinski@marvell.com>
2790 * optabs.h (create_expand_operand): Initialize target field also.
2792 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2794 PR tree-optimization/92429
2795 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2796 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2798 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2801 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2803 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2804 aarch64_sve_int_mode to each mode.
2806 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2808 * doc/analyzer.texi (Overview): Add note about
2809 -fdump-ipa-analyzer.
2811 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2813 PR tree-optimization/93231
2814 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2815 input_type is unsigned. Use tree_to_shwi for shift constant.
2816 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2817 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2818 inputs more efficiently.
2820 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2822 * config/i386/i386.md (*movsf_internal): Do not require
2823 SSE2 ISA for alternatives 14 and 15.
2825 2020-01-15 Richard Biener <rguenther@suse.de>
2828 * tree-eh.c (sink_clobbers): If we already visited the destination
2829 block do not defer insertion.
2830 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2831 the purpose of defered insertion.
2833 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2835 * BASE-VER: Bump to 10.0.1.
2837 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2839 PR tree-optimization/93247
2840 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2841 type of the stmt that we're going to vectorize.
2843 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2845 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2846 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2849 2020-01-15 Martin Liska <mliska@suse.cz>
2851 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2852 2 calls of streamer_read_hwi in a function call.
2854 2020-01-15 Richard Biener <rguenther@suse.de>
2856 * alias.c (record_alias_subset): Avoid redundant work when
2857 subset is already recorded.
2859 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2861 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2862 the analyzer options provide CWE identifiers.
2864 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2866 * tree-diagnostic-path.cc (path_summary::event_range::print):
2867 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2868 using get_pure_location.
2870 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2872 PR tree-optimization/93262
2873 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2874 perform head trimming only if the last argument is constant,
2875 either all ones, or larger or equal to head trim, in the latter
2876 case decrease the last argument by head_trim.
2878 PR tree-optimization/93249
2879 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2880 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2881 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2882 perform head trim unless we can prove there are no '\0' chars
2883 from the source among the first head_trim chars.
2885 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2887 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2889 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2892 * config/i386/sse.md
2893 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2894 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2895 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2896 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2897 just a single alternative instead of two, make operands 1 and 2
2900 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2903 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2906 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2908 * Makefile.in (lang_opt_files): Add analyzer.opt.
2909 (ANALYZER_OBJS): New.
2910 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2911 tristate.o and ANALYZER_OBJS.
2912 (TEXI_GCCINT_FILES): Add analyzer.texi.
2913 * common.opt (-fanalyzer): New driver option.
2914 * config.in: Regenerate.
2915 * configure: Regenerate.
2916 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2917 (gccdepdir): Also create depdir for "analyzer" subdir.
2918 * digraph.cc: New file.
2919 * digraph.h: New file.
2920 * doc/analyzer.texi: New file.
2921 * doc/gccint.texi ("Static Analyzer") New menu item.
2922 (analyzer.texi): Include it.
2923 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2924 ("Warning Options"): Add static analysis warnings to the list.
2925 (-Wno-analyzer-double-fclose): New option.
2926 (-Wno-analyzer-double-free): New option.
2927 (-Wno-analyzer-exposure-through-output-file): New option.
2928 (-Wno-analyzer-file-leak): New option.
2929 (-Wno-analyzer-free-of-non-heap): New option.
2930 (-Wno-analyzer-malloc-leak): New option.
2931 (-Wno-analyzer-possible-null-argument): New option.
2932 (-Wno-analyzer-possible-null-dereference): New option.
2933 (-Wno-analyzer-null-argument): New option.
2934 (-Wno-analyzer-null-dereference): New option.
2935 (-Wno-analyzer-stale-setjmp-buffer): New option.
2936 (-Wno-analyzer-tainted-array-index): New option.
2937 (-Wno-analyzer-use-after-free): New option.
2938 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2939 (-Wno-analyzer-use-of-uninitialized-value): New option.
2940 (-Wanalyzer-too-complex): New option.
2941 (-fanalyzer-call-summaries): New warning.
2942 (-fanalyzer-checker=): New warning.
2943 (-fanalyzer-fine-grained): New warning.
2944 (-fno-analyzer-state-merge): New warning.
2945 (-fno-analyzer-state-purge): New warning.
2946 (-fanalyzer-transitivity): New warning.
2947 (-fanalyzer-verbose-edges): New warning.
2948 (-fanalyzer-verbose-state-changes): New warning.
2949 (-fanalyzer-verbosity=): New warning.
2950 (-fdump-analyzer): New warning.
2951 (-fdump-analyzer-callgraph): New warning.
2952 (-fdump-analyzer-exploded-graph): New warning.
2953 (-fdump-analyzer-exploded-nodes): New warning.
2954 (-fdump-analyzer-exploded-nodes-2): New warning.
2955 (-fdump-analyzer-exploded-nodes-3): New warning.
2956 (-fdump-analyzer-supergraph): New warning.
2957 * doc/sourcebuild.texi (dg-require-dot): New.
2958 (dg-check-dot): New.
2959 * gdbinit.in (break-on-saved-diagnostic): New command.
2960 * graphviz.cc: New file.
2961 * graphviz.h: New file.
2962 * ordered-hash-map-tests.cc: New file.
2963 * ordered-hash-map.h: New file.
2964 * passes.def (pass_analyzer): Add before
2965 pass_ipa_whole_program_visibility.
2966 * selftest-run-tests.c (selftest::run_tests): Call
2967 selftest::ordered_hash_map_tests_cc_tests.
2968 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2970 * shortest-paths.h: New file.
2971 * timevar.def (TV_ANALYZER): New timevar.
2972 (TV_ANALYZER_SUPERGRAPH): Likewise.
2973 (TV_ANALYZER_STATE_PURGE): Likewise.
2974 (TV_ANALYZER_PLAN): Likewise.
2975 (TV_ANALYZER_SCC): Likewise.
2976 (TV_ANALYZER_WORKLIST): Likewise.
2977 (TV_ANALYZER_DUMP): Likewise.
2978 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2979 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2980 * tree-pass.h (make_pass_analyzer): New decl.
2981 * tristate.cc: New file.
2982 * tristate.h: New file.
2984 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2987 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2988 alternatives 9 and 10.
2990 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2992 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2993 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2994 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2995 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2996 (selftest::hash_map_tests_c_tests): Call it.
2997 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2998 New static constant, using the value of = H::empty_zero_p.
2999 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3000 from default_hash_traits <Value>.
3001 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3003 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3004 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3005 calls to mark_empty with !Descriptor::empty_zero_p.
3006 (hash_table::empty_slow): Conditionalize the memset call with a
3007 check that Descriptor::empty_zero_p; otherwise, loop through the
3008 entries calling mark_empty on them.
3009 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3010 (pointer_hash::empty_zero_p): Likewise.
3011 (pair_hash::empty_zero_p): Likewise.
3012 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3014 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3015 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3016 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3017 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3018 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3019 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3020 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3021 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3023 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3026 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3028 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3029 fix typo on return value.
3031 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3034 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3036 (cgraph_edge::make_speculative): Add param for setting speculative_id
3038 (cgraph_edge::speculative_call_info): Update comments and find reference
3039 by speculative_id for multiple indirect targets.
3040 (cgraph_edge::resolve_speculation): Decrease the speculations
3041 for indirect edge, drop it's speculative if not direct target
3042 left. Update comments.
3043 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3044 (cgraph_node::dump): Print num_speculative_call_targets.
3045 (cgraph_node::verify_node): Don't report error if speculative
3046 edge not include statement.
3047 (cgraph_edge::num_speculative_call_targets_p): New function.
3048 * cgraph.h (int common_target_id): Remove.
3049 (int common_target_probability): Remove.
3050 (num_speculative_call_targets): New variable.
3051 (make_speculative): Add param for setting speculative_id.
3052 (cgraph_edge::num_speculative_call_targets_p): New declare.
3053 (target_prob): New variable.
3054 (speculative_id): New variable.
3055 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3056 call summaries for multiple speculative call targets.
3057 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3058 * ipa-profile.c (struct speculative_call_target): New struct.
3059 (class speculative_call_summary): New class.
3060 (class speculative_call_summaries): New class.
3061 (call_sums): New variable.
3062 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3063 (ipa_profile_write_edge_summary): New function.
3064 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3065 (ipa_profile_dump_all_summaries): New function.
3066 (ipa_profile_read_edge_summary): New function.
3067 (ipa_profile_read_summary_section): New function.
3068 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3069 (ipa_profile): Generate num_speculative_call_targets from
3071 * ipa-ref.h (speculative_id): New variable.
3072 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3073 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3074 common_target_probability. Stream out speculative_id and
3075 num_speculative_call_targets.
3076 (input_edge): Likewise.
3077 * predict.c (dump_prediction): Remove edges count assert to be
3079 * symtab.c (symtab_node::create_reference): Init speculative_id.
3080 (symtab_node::clone_references): Clone speculative_id.
3081 (symtab_node::clone_referring): Clone speculative_id.
3082 (symtab_node::clone_reference): Clone speculative_id.
3083 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3084 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3085 if indirect call contains multiple speculative targets.
3086 * value-prof.h (check_ic_target): Remove.
3087 * value-prof.c (gimple_value_profile_transformations):
3088 Use void function gimple_ic_transform.
3089 * value-prof.c (gimple_ic_transform): Handle topn case.
3090 Fix comment typos. Change it to a void function.
3092 2020-01-13 Andrew Pinski <apinski@marvell.com>
3094 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3095 (octeontx2t98): New define.
3096 (octeontx2t96): New define.
3097 (octeontx2t93): New define.
3098 (octeontx2f95): New define.
3099 (octeontx2f95n): New define.
3100 (octeontx2f95mm): New define.
3101 * config/aarch64/aarch64-tune.md: Regenerate.
3102 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3104 2020-01-13 Jason Merrill <jason@redhat.com>
3106 PR c++/33799 - destroy return value if local cleanup throws.
3107 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3109 2020-01-13 Martin Liska <mliska@suse.cz>
3111 * ipa-cp.c (get_max_overall_size): Use newly
3112 renamed param param_ipa_cp_unit_growth.
3113 * params.opt: Remove legacy param name.
3115 2020-01-13 Martin Sebor <msebor@redhat.com>
3117 PR tree-optimization/93213
3118 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3119 stores to be eliminated.
3121 2020-01-13 Martin Liska <mliska@suse.cz>
3123 * opts.c (print_help): Do not print CL_PARAM
3124 and CL_WARNING for CL_OPTIMIZATION.
3126 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3129 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3130 depending on optimization settings.
3132 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3134 PR tree-optimization/90838
3135 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3136 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3137 argument rather than to initialize temporary for targets that
3138 don't use the mode argument at all. Initialize ctzval to avoid
3141 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3143 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3144 * tree-core.h: Document it.
3145 * gimplify.c (gimplify_omp_workshare): Set it.
3146 * omp-low.c (lower_omp_target): Use it.
3147 * tree-pretty-print.c (dump_omp_clause): Print it.
3149 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3150 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3152 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3154 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3155 * common.opt (fdiagnostics-path-format=): New option.
3156 (diagnostic_path_format): New enum.
3157 (fdiagnostics-show-path-depths): New option.
3158 * coretypes.h (diagnostic_event_id_t): New forward decl.
3159 * diagnostic-color.c (color_dict): Add "path".
3160 * diagnostic-event-id.h: New file.
3161 * diagnostic-format-json.cc (json_from_expanded_location): Make
3163 (json_end_diagnostic): Call context->make_json_for_path if it
3164 exists and the diagnostic has a path.
3165 (diagnostic_output_format_init): Clear context->print_path.
3166 * diagnostic-path.h: New file.
3167 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3168 when printing a run of events in a diagnostic_path so that they
3169 all get the same color.
3170 (layout::m_diagnostic_path_p): New field.
3171 (layout::layout): Initialize it.
3172 (layout::print_any_labels): Don't colorize the label text for an
3173 event in a diagnostic_path.
3174 (gcc_rich_location::add_location_if_nearby): Add
3175 "restrict_to_current_line_spans" and "label" params. Pass the
3176 former to layout.maybe_add_location_range; pass the latter
3177 when calling add_range.
3178 * diagnostic.c: Include "diagnostic-path.h".
3179 (diagnostic_initialize): Initialize context->path_format and
3180 context->show_path_depths.
3181 (diagnostic_show_any_path): New function.
3182 (diagnostic_path::interprocedural_p): New function.
3183 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3184 (simple_diagnostic_path::num_events): New function.
3185 (simple_diagnostic_path::get_event): New function.
3186 (simple_diagnostic_path::add_event): New function.
3187 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3188 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3189 (debug): New overload taking a diagnostic_path *.
3190 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3191 * diagnostic.h (enum diagnostic_path_format): New enum.
3192 (json::value): New forward decl.
3193 (diagnostic_context::path_format): New field.
3194 (diagnostic_context::show_path_depths): New field.
3195 (diagnostic_context::print_path): New callback field.
3196 (diagnostic_context::make_json_for_path): New callback field.
3197 (diagnostic_show_any_path): New decl.
3198 (json_from_expanded_location): New decl.
3199 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3200 (-fdiagnostics-show-path-depths): New option.
3201 (-fdiagnostics-color): Add "path" to description of default
3202 GCC_COLORS; describe it.
3203 (-fdiagnostics-format=json): Document how diagnostic paths are
3204 represented in the JSON output format.
3205 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3206 Add optional params "restrict_to_current_line_spans" and "label".
3207 * opts.c (common_handle_option): Handle
3208 OPT_fdiagnostics_path_format_ and
3209 OPT_fdiagnostics_show_path_depths.
3210 * pretty-print.c: Include "diagnostic-event-id.h".
3211 (pp_format): Implement "%@" format code for printing
3212 diagnostic_event_id_t *.
3213 (selftest::test_pp_format): Add tests for "%@".
3214 * selftest-run-tests.c (selftest::run_tests): Call
3215 selftest::tree_diagnostic_path_cc_tests.
3216 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3217 * toplev.c (general_init): Initialize global_dc->path_format and
3218 global_dc->show_path_depths.
3219 * tree-diagnostic-path.cc: New file.
3220 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3221 non-static. Drop "diagnostic" param in favor of storing the
3222 original value of "where" and re-using it.
3223 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3224 maybe_unwind_expanded_macro_loc.
3225 (tree_diagnostics_defaults): Initialize context->print_path and
3226 context->make_json_for_path.
3227 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3229 (default_tree_make_json_for_path): New decl.
3230 (maybe_unwind_expanded_macro_loc): New decl.
3232 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3234 PR tree-optimization/93210
3235 * fold-const.h (native_encode_initializer,
3236 can_native_interpret_type_p): Declare.
3237 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3239 (native_encode_initializer): New function, moved from dwarf2out.c.
3240 Adjust to native_encode_expr compatible arguments, including dry-run
3241 and partial extraction modes. Don't handle STRING_CST.
3242 (can_native_interpret_type_p): No longer static.
3243 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3244 offset / BITS_PER_UNIT fits into int and don't call it if
3245 can_native_interpret_type_p fails. If suboff is NULL and for
3246 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3247 native_encode_initializer.
3248 (fold_const_aggregate_ref_1): Formatting fix.
3249 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3250 (tree_add_const_value_attribute): Adjust caller.
3252 PR tree-optimization/90838
3253 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3254 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3255 CTZ_DEFINED_VALUE_AT_ZERO.
3257 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3260 * lra-constraints.c (match_reload): Permit input operands have the
3261 same mode as output while other input operands have a different
3264 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3266 PR tree-optimization/90838
3267 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3268 (check_ctz_string): Likewise.
3269 (optimize_count_trailing_zeroes): Likewise.
3270 (simplify_count_trailing_zeroes): Likewise.
3271 (pass_forwprop::execute): Try ctz simplification.
3272 * match.pd: Add matching for ctz idioms.
3274 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3276 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3278 (aarch64_invalid_unary_op): New function for target hook.
3279 (aarch64_invalid_binary_op): New function for target hook.
3281 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3283 * config.gcc: Add arm_bf16.h.
3284 * config/aarch64/aarch64-builtins.c
3285 (aarch64_simd_builtin_std_type): Add BFmode.
3286 (aarch64_init_simd_builtin_types): Define element types for vector
3288 (aarch64_init_bf16_types): New function.
3289 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3290 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3292 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3293 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3295 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3296 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3297 * config/aarch64/aarch64.c
3298 (aarch64_classify_vector_mode): Add support for BF types.
3299 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3300 (aarch64_vq_mode): Add support for BF types.
3301 (aarch64_simd_container_mode): Add support for BF types.
3302 (aarch64_mangle_type): Add support for BF scalar type.
3303 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3304 * config/aarch64/arm_bf16.h: New file.
3305 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3306 * config/aarch64/iterators.md: Add BF types to mode attributes.
3307 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3309 2020-01-10 Jason Merrill <jason@redhat.com>
3311 PR c++/93173 - incorrect tree sharing.
3312 * gimplify.c (copy_if_shared): No longer static.
3313 * gimplify.h: Declare it.
3315 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3317 * doc/invoke.texi (-msve-vector-bits=): Document that
3318 -msve-vector-bits=128 now generates VL-specific code for
3319 little-endian targets.
3320 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3321 build_vector_type_for_mode to construct the data vector types.
3322 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3323 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3324 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3325 for 128-bit vectors.
3327 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3329 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3332 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3334 * config/aarch64/aarch64-builtins.c
3335 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3336 rather than checking the number of elements and the element mode.
3338 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3340 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3341 get_related_vectype_for_scalar_type rather than build_vector_type
3342 to create the index type for a conditional reduction.
3344 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3346 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3347 for any type of gather or scatter, including strided accesses.
3349 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3351 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3354 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3356 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3358 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3359 parameter and its use to reset DR_OFFSET's.
3360 (vect_transform_loop): Remove orig_drs_init argument.
3361 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3362 member of dr_vec_info rather than the offset of the associated
3363 data_reference's innermost_loop_behavior.
3364 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3365 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3366 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3367 get_dr_vinfo_offset.
3368 (vectorizable_store): Likewise.
3369 (vectorizable_load): Likewise.
3371 2020-01-10 Richard Biener <rguenther@suse.de>
3373 * gimple-ssa-store-merging
3374 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3376 2020-01-10 Martin Liska <mliska@suse.cz>
3379 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3380 encapsulation that was there before r280040.
3382 2020-01-10 Richard Biener <rguenther@suse.de>
3385 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3386 sequences to avoid walking them again for secondary opportunities.
3387 (pass_lower_eh_dispatch::execute): Instead actually insert
3390 2020-01-10 Richard Biener <rguenther@suse.de>
3393 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3394 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3395 avoid quadraticness.
3397 2020-01-10 Martin Jambor <mjambor@suse.cz>
3399 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3400 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3401 to get param_ipa_sra_max_replacements.
3402 (param_splitting_across_edge): Pass the caller to
3403 pull_accesses_from_callee.
3405 2020-01-10 Martin Jambor <mjambor@suse.cz>
3407 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3408 * ipa-cp.c (max_new_size): Removed.
3409 (orig_overall_size): New variable.
3410 (get_max_overall_size): New function.
3411 (estimate_local_effects): Use it. Adjust dump.
3412 (decide_about_value): Likewise.
3413 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3414 orig_overall_size. Adjust dump.
3415 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3417 2020-01-10 Martin Jambor <mjambor@suse.cz>
3419 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3420 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3421 instead of param_ipa_max_agg_items.
3422 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3423 optimization info for the callee.
3425 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3427 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3428 markers if debug_inline_points is false.
3430 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3432 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3434 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3435 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3436 aarch64-sve-builtins-sve2.h.
3437 (aarch64-sve-builtins-sve2.o): New rule.
3438 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3439 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3440 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3441 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3442 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3443 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3445 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3446 instructions that are handled here.
3447 (@cond_asrd<mode>): Generalize to...
3448 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3449 (*cond_asrd<mode>_2): Generalize to...
3450 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3451 (*cond_asrd<mode>_z): Generalize to...
3452 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3453 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3454 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3455 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3456 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3458 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3459 (@aarch64_scatter_stnt<mode>): Likewise.
3460 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3461 (@aarch64_mul_lane_<mode>): Likewise.
3462 (@aarch64_sve_suqadd<mode>_const): Likewise.
3463 (*<sur>h<addsub><mode>): Generalize to...
3464 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3466 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3467 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3468 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3469 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3470 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3471 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3472 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3473 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3474 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3475 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3476 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3477 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3478 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3479 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3480 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3481 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3482 (@aarch64_sve2_xar<mode>): Likewise.
3483 (@aarch64_sve2_bcax<mode>): Likewise.
3484 (*aarch64_sve2_eor3<mode>): Rename to...
3485 (@aarch64_sve2_eor3<mode>): ...this.
3486 (@aarch64_sve2_bsl<mode>): New expander.
3487 (@aarch64_sve2_nbsl<mode>): Likewise.
3488 (@aarch64_sve2_bsl1n<mode>): Likewise.
3489 (@aarch64_sve2_bsl2n<mode>): Likewise.
3490 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3491 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3492 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3493 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3494 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3495 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3496 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3497 (<su>mull<bt><Vwide>): Generalize to...
3498 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3500 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3501 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3502 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3503 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3504 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3505 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3506 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3507 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3508 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3509 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3510 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3511 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3512 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3513 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3514 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3515 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3516 (<SHRNB:r>shrnb<mode>): Generalize to...
3517 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3519 (<SHRNT:r>shrnt<mode>): Generalize to...
3520 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3522 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3523 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3524 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3525 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3526 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3527 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3528 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3529 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3530 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3531 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3532 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3533 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3534 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3535 (@aarch64_sve2_cvtnt<mode>): Likewise.
3536 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3537 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3538 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3539 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3540 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3541 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3542 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3543 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3544 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3545 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3546 (@aarch64_sve2_pmul<mode>): Likewise.
3547 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3548 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3549 (@aarch64_sve2_tbl2<mode>): Likewise.
3550 (@aarch64_sve2_tbx<mode>): Likewise.
3551 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3552 (@aarch64_sve2_histcnt<mode>): Likewise.
3553 (@aarch64_sve2_histseg<mode>): Likewise.
3554 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3555 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3556 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3557 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3558 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3559 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3560 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3561 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3562 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3563 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3564 (SVE2_PMULL_PAIR_I): New mode iterators.
3565 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3566 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3567 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3568 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3569 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3570 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3571 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3572 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3573 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3574 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3575 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3576 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3577 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3578 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3579 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3580 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3581 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3582 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3583 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3584 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3585 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3586 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3587 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3588 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3589 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3590 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3591 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3592 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3593 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3594 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3595 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3597 (VNARROW, Ventype): New mode attributes.
3598 (Vewtype): Handle VNx2DI. Fix typo in comment.
3599 (VDOUBLE): New mode attribute.
3600 (sve_lane_con): Handle VNx8HI.
3601 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3602 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3603 (sve_int_op, sve_int_op_rev): Handle the above codes.
3604 (sve_pred_int_rhs2_operand): Likewise.
3605 (MULLBT, SHRNB, SHRNT): Delete.
3606 (SVE_INT_SHIFT_IMM): New int iterator.
3607 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3608 and UNSPEC_WHILEHS for TARGET_SVE2.
3609 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3610 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3611 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3612 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3613 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3614 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3615 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3616 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3617 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3618 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3619 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3620 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3621 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3622 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3623 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3624 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3625 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3626 (optab): Handle the new unspecs.
3627 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3629 (lr): Handle the new unspecs.
3631 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3632 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3633 (sve_int_qsub_op): New int attributes.
3634 (sve_fp_op, rot): Handle the new unspecs.
3635 * config/aarch64/aarch64-sve-builtins.h
3636 (function_resolver::require_matching_pointer_type): Declare.
3637 (function_resolver::resolve_unary): Add an optional boolean argument.
3638 (function_resolver::finish_opt_n_resolution): Add an optional
3639 type_suffix_index argument.
3640 (gimple_folder::redirect_call): Declare.
3641 (gimple_expander::prepare_gather_address_operands): Add an optional
3643 * config/aarch64/aarch64-sve-builtins.cc: Include
3644 aarch64-sve-builtins-sve2.h.
3645 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3646 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3647 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3648 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3649 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3650 (TYPES_s_unsigned): Likewise.
3651 (TYPES_s_integer): Use TYPES_s_unsigned.
3652 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3653 (TYPES_sd_integer): Use them.
3654 (TYPES_d_unsigned): New macro.
3655 (TYPES_d_integer): Use it.
3656 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3657 (TYPES_cvt_narrow): Likewise.
3658 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3659 (preds_mx): New variable.
3660 (function_builder::add_overloaded_function): Allow the new feature
3661 set to be more restrictive than the original one.
3662 (function_resolver::infer_pointer_type): Remove qualifiers from
3663 the pointer type before printing it.
3664 (function_resolver::require_matching_pointer_type): New function.
3665 (function_resolver::resolve_sv_displacement): Handle functions
3666 that don't support 32-bit vector indices or svint32_t vector offsets.
3667 (function_resolver::finish_opt_n_resolution): Take the inferred type
3668 as a separate argument.
3669 (function_resolver::resolve_unary): Optionally treat all forms in
3670 the same way as normal merging functions.
3671 (gimple_folder::redirect_call): New function.
3672 (function_expander::prepare_gather_address_operands): Add an argument
3673 that says whether scaled forms are available. If they aren't,
3674 handle scaling of vector indices and don't add the extension and
3676 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3677 fall back to using cond_* instead.
3678 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3679 Split out the member variables into...
3680 (rtx_code_function_base): ...this new base class.
3681 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3682 (unspec_based_function): Split out the member variables into...
3683 (unspec_based_function_base): ...this new base class.
3684 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3685 (unspec_based_function_exact_insn): New class.
3686 (unspec_based_add_function, unspec_based_add_lane_function)
3687 (unspec_based_lane_function, unspec_based_pred_function)
3688 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3689 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3690 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3692 (unspec_based_fused_function): New class.
3693 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3694 (unspec_based_fused_lane_function): New class.
3695 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3697 (CODE_FOR_MODE1): New macro.
3698 (fixed_insn_function): New class.
3699 (while_comparison): Likewise.
3700 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3701 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3702 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3703 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3704 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3705 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3706 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3707 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3708 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3709 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3710 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3711 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3712 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3713 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3714 Also add an initial argument for unary_convert_narrowt, regardless
3715 of the predication type.
3716 (build_32_64): Allow loads and stores to specify MODE_none.
3717 (build_sv_index64, build_sv_uint_offset): New functions.
3718 (long_type_suffix): New function.
3719 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3720 (binary_imm_long_base, load_gather_sv_base): Likewise.
3721 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3722 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3723 (unary_narrowb_base, unary_narrowt_base): Likewise.
3724 (binary_long_lane_def, binary_long_lane): New shape.
3725 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3726 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3727 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3728 (binary_to_uint_def, binary_to_uint): Likewise.
3729 (binary_wide_def, binary_wide): Likewise.
3730 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3731 (compare_def, compare): Likewise.
3732 (compare_ptr_def, compare_ptr): Likewise.
3733 (load_ext_gather_index_restricted_def,
3734 load_ext_gather_index_restricted): Likewise.
3735 (load_ext_gather_offset_restricted_def,
3736 load_ext_gather_offset_restricted): Likewise.
3737 (load_gather_sv_def): Inherit from load_gather_sv_base.
3738 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3739 (shift_left_imm_def, shift_left_imm): Likewise.
3740 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3741 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3742 (store_scatter_index_restricted_def,
3743 store_scatter_index_restricted): Likewise.
3744 (store_scatter_offset_restricted_def,
3745 store_scatter_offset_restricted): Likewise.
3746 (tbl_tuple_def, tbl_tuple): Likewise.
3747 (ternary_long_lane_def, ternary_long_lane): Likewise.
3748 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3749 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3750 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3751 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3752 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3753 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3754 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3755 (ternary_uint_def, ternary_uint): Likewise.
3756 (unary_convert): Fix typo in comment.
3757 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3758 (unary_long_def, unary_long): Likewise.
3759 (unary_narrowb_def, unary_narrowb): Likewise.
3760 (unary_narrowt_def, unary_narrowt): Likewise.
3761 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3762 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3763 (unary_to_int_def, unary_to_int): Likewise.
3764 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3765 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3766 (svasrd_impl): Delete.
3767 (svcadd_impl::expand): Handle integer operations too.
3768 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3769 new functions to derive the unspec numbers.
3770 (svmla_svmls_lane_impl): Replace with...
3771 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3772 integer operations too.
3773 (svwhile_impl): Rename to...
3774 (svwhilelx_impl): ...this and inherit from while_comparison.
3775 (svasrd): Use unspec_based_function.
3776 (svmla_lane): Use svmla_lane_impl.
3777 (svmls_lane): Use svmls_lane_impl.
3778 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3779 (svwhilele, svwhilelt): Use svwhilelx_impl.
3780 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3781 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3782 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3783 * config/aarch64/aarch64-sve-builtins.def: Include
3784 aarch64-sve-builtins-sve2.def.
3786 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3788 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3789 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3790 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3791 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3792 immediates as well as vector ones.
3793 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3794 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3795 (aarch64_sve_qsub_immediate): Update calls accordingly.
3797 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3799 * config/aarch64/aarch64-sve2.md: Add banner comments.
3800 (<su>mulh<r>s<mode>3): Move further up file.
3801 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3802 (*aarch64_sve2_sra<mode>): Move further down file.
3803 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3805 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3807 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3809 (while_optab_cmp): Handle them.
3810 * config/aarch64/aarch64-sve.md
3811 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3812 and add a "@" marker.
3813 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3814 instead of gen_aarch64_sve2_while_ptest.
3815 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3817 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3819 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3820 (UNSPEC_WHILELE): ...this.
3821 (UNSPEC_WHILE_LO): Rename to...
3822 (UNSPEC_WHILELO): ...this.
3823 (UNSPEC_WHILE_LS): Rename to...
3824 (UNSPEC_WHILELS): ...this.
3825 (UNSPEC_WHILE_LT): Rename to...
3826 (UNSPEC_WHILELT): ...this.
3827 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3828 (cmp_op, while_optab_cmp): Likewise.
3829 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3830 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3831 (svwhilelt): Likewise.
3833 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3835 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3836 (unary_to_uint): Define.
3837 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3838 (unary_count): Rename to...
3839 (unary_to_uint_def, unary_to_uint): ...this.
3840 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3842 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3844 * config/aarch64/aarch64-sve-builtins-functions.h
3845 (code_for_mode_function): New class.
3846 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3847 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3848 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3849 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3850 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3852 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3854 * config/aarch64/iterators.md (addsub): New code attribute.
3855 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3857 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3858 in the asm string and attributes. Fix indentation.
3859 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3861 (@aarch64_sve_<optab><mode>): ...this.
3862 * config/aarch64/aarch64-sve-builtins.h
3863 (function_expander::expand_signed_unpred_op): Delete.
3864 * config/aarch64/aarch64-sve-builtins.cc
3865 (function_expander::expand_signed_unpred_op): Likewise.
3866 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3867 try using code_for_aarch64_sve instead.
3868 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3869 (svqsub_impl): Likewise.
3870 (svqadd, svqsub): Use rtx_code_function instead.
3872 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3874 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3875 (HADDSUB, sur, addsub): Remove them.
3877 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3879 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3880 functions the same way, rather than singling out those that
3881 aren't mapped directly to optabs.
3883 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3885 * target.def (compatible_vector_types_p): New target hook.
3886 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3887 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3888 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3889 * doc/tm.texi: Regenerate.
3890 * gimple-expr.c: Include target.h.
3891 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3892 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3894 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3895 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3896 Use the original predicate if it already has a suitable type.
3898 2020-01-09 Martin Jambor <mjambor@suse.cz>
3900 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3901 resolve_speculation and redirect_call_stmt_to_callee static. Change
3902 return type of set_call_stmt to cgraph_edge *.
3903 * auto-profile.c (afdo_indirect_call): Adjust call to
3904 redirect_call_stmt_to_callee.
3905 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3906 make the this pointer explicit, adjust self-recursive calls and the
3907 call top make_direct. Return the resulting edge.
3908 (cgraph_edge::remove): Make this pointer explicit.
3909 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3910 (cgraph_edge::make_direct): Likewise, adjust call to
3911 resolve_speculation.
3912 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3913 call to set_call_stmt.
3914 (cgraph_update_edges_for_call_stmt_node): Update call to
3915 set_call_stmt and remove.
3916 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3917 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3918 (cgraph_node::create_edge_including_clones): Moved "first" definition
3919 of edge to the block where it was used. Adjusted calls to
3921 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3922 cgraph_edge::remove.
3923 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3924 make_direct and redirect_call_stmt_to_callee.
3925 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3926 resolve_speculation and make_direct.
3927 * ipa-inline-transform.c (inline_transform): Adjust call to
3928 redirect_call_stmt_to_callee.
3929 (check_speculations_1):: Adjust call to resolve_speculation.
3930 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3931 resolve-speculation.
3932 (inline_small_functions): Adjust call to resolve_speculation.
3933 (ipa_inline): Likewise.
3934 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3936 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3937 safe with regards to edge removal, adjust calls to
3938 redirect_call_stmt_to_callee.
3939 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3940 and redirect_call_stmt_to_callee.
3941 * multiple_target.c (create_dispatcher_calls): Adjust call to
3942 redirect_call_stmt_to_callee
3943 (redirect_to_specific_clone): Likewise.
3944 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3945 Adjust calls to cgraph_edge::remove.
3946 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3947 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3948 (expand_call_inline): Adjust call to cgraph_edge::remove.
3950 2020-01-09 Martin Liska <mliska@suse.cz>
3952 * params.opt: Set Optimization for
3953 param_max_speculative_devirt_maydefs.
3955 2020-01-09 Martin Sebor <msebor@redhat.com>
3959 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3961 2020-01-09 Martin Liska <mliska@suse.cz>
3963 * auto-profile.c (auto_profile): Use opt_for_fn
3965 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3966 (propagate_vals_across_arith_jfunc): Likewise.
3967 (hint_time_bonus): Likewise.
3968 (incorporate_penalties): Likewise.
3969 (good_cloning_opportunity_p): Likewise.
3970 (perform_estimation_of_a_value): Likewise.
3971 (estimate_local_effects): Likewise.
3972 (ipcp_propagate_stage): Likewise.
3973 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3974 (set_switch_stmt_execution_predicate): Likewise.
3975 (analyze_function_body): Likewise.
3976 * ipa-inline-analysis.c (offline_size): Likewise.
3977 * ipa-inline.c (early_inliner): Likewise.
3978 * ipa-prop.c (ipa_analyze_node): Likewise.
3979 (ipcp_transform_function): Likewise.
3980 * ipa-sra.c (process_scan_results): Likewise.
3981 (ipa_sra_summarize_function): Likewise.
3982 * params.opt: Rename ipcp-unit-growth to
3983 ipa-cp-unit-growth. Add Optimization for various
3984 IPA-related parameters.
3986 2020-01-09 Richard Biener <rguenther@suse.de>
3989 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3991 2020-01-09 Richard Biener <rguenther@suse.de>
3993 PR tree-optimization/93040
3994 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3996 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3998 * common/config/avr/avr-common.c (avr_option_optimization_table)
3999 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4001 2020-01-09 Martin Liska <mliska@suse.cz>
4003 * cgraphclones.c (symbol_table::materialize_all_clones):
4004 Use cgraph_node::dump_name.
4006 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4009 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4010 output_operand_lossage instead of gcc_unreachable.
4011 * doc/md.texi (riscv f constraint): Fix typo.
4014 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4015 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4016 CONST_SCALAR_INT_P instead of CONST_INT_P.
4017 (*subv<mode>4_1): Rename to ...
4018 (subv<mode>4_1): ... this.
4019 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4020 define_insn_and_split patterns.
4021 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4024 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4026 * vec.c (class selftest::count_dtor): New class.
4027 (selftest::test_auto_delete_vec): New test.
4028 (selftest::vec_c_tests): Call it.
4029 * vec.h (class auto_delete_vec): New class template.
4030 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4032 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4034 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4036 2020-01-08 Jim Wilson <jimw@sifive.com>
4038 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4039 use of TLS_MODEL_LOCAL_EXEC when not pic.
4041 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4043 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4046 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4049 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4050 *stack_protect_set_3 peephole2): Also check that the second
4051 insns source is general_operand.
4054 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4055 predicate for output operand instead of register_operand.
4056 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4057 memory destination and non-memory operands[2].
4059 2020-01-08 Martin Liska <mliska@suse.cz>
4061 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4062 ::dump_asm_name instead of (::name or ::asm_name).
4063 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4064 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4065 (analyze_functions): Likewise.
4066 (expand_all_functions): Likewise.
4067 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4068 (propagate_bits_across_jump_function): Likewise.
4069 (dump_profile_updates): Likewise.
4070 (ipcp_store_bits_results): Likewise.
4071 (ipcp_store_vr_results): Likewise.
4072 * ipa-devirt.c (dump_targets): Likewise.
4073 * ipa-fnsummary.c (analyze_function_body): Likewise.
4074 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4075 (process_hsa_functions): Likewise.
4076 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4077 (set_alias_uids): Likewise.
4078 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4079 * ipa-inline.c (recursive_inlining): Likewise.
4080 (inline_to_all_callers_1): Likewise.
4081 (ipa_inline): Likewise.
4082 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4083 (ipa_propagate_frequency): Likewise.
4084 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4085 (remove_described_reference): Likewise.
4086 * ipa-pure-const.c (worse_state): Likewise.
4087 (check_retval_uses): Likewise.
4088 (analyze_function): Likewise.
4089 (propagate_pure_const): Likewise.
4090 (propagate_nothrow): Likewise.
4091 (dump_malloc_lattice): Likewise.
4092 (propagate_malloc): Likewise.
4093 (pass_local_pure_const::execute): Likewise.
4094 * ipa-visibility.c (optimize_weakref): Likewise.
4095 (function_and_variable_visibility): Likewise.
4096 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4097 (ipa_discover_variable_flags): Likewise.
4098 * lto-streamer-out.c (output_function): Likewise.
4099 (output_constructor): Likewise.
4100 * tree-inline.c (copy_bb): Likewise.
4101 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4102 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4104 2020-01-08 Richard Biener <rguenther@suse.de>
4107 * tree-eh.c (sink_clobbers): Update virtual operands for
4108 the first and last stmt only. Add a dry-run capability.
4109 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4110 after CFG manipulations and in RPO order to catch all
4111 secondary opportunities reliably.
4113 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4116 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4118 2019-01-08 Richard Biener <rguenther@suse.de>
4121 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4122 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4123 virtual operand, also updating SSA use.
4124 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4125 Update stmt after resetting virtual operand.
4126 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4127 * gimple-iterator.c (gsi_remove): When not removing the stmt
4128 permanently do not delink immediate uses or mark the stmt modified.
4130 2020-01-08 Martin Liska <mliska@suse.cz>
4132 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4133 (ipa_call_context::estimate_size_and_time): Likewise.
4134 (inline_analyze_function): Likewise.
4136 2020-01-08 Martin Liska <mliska@suse.cz>
4138 * cgraph.c (cgraph_node::dump): Use systematically
4141 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4143 Add -nodevicespecs option for avr.
4146 * config/avr/avr.opt (-nodevicespecs): New driver option.
4147 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4148 "-specs=device-specs/..." if that option is not set.
4149 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4151 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4153 Implement 64-bit double functions for avr.
4156 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4157 --with-double-comparison.
4158 * doc/install.texi: Document them.
4159 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4160 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4161 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4162 * doc/invoke.texi (AVR Built-in Macros): Document them.
4163 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4164 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4165 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4167 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4170 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4171 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4172 when only building rm-profile multilibs.
4174 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4177 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4178 lattice for a value to check.
4179 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4180 finite propagation in self-recursive scc.
4182 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4184 * ipa-inline.c (caller_growth_limits): Restore the AND.
4186 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4188 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4189 (VEC_ALLREG_ALT): New iterator.
4190 (VEC_ALLREG_INT_MODE): New iterator.
4191 (VCMP_MODE): New iterator.
4192 (VCMP_MODE_INT): New iterator.
4193 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4194 (vec_cmp<u>v64qidi): New define_expand.
4195 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4196 (vec_cmpu<mode>di_exec): New define_expand.
4197 (vec_cmp<u>v64qidi_exec): New define_expand.
4198 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4199 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4200 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4201 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4202 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4203 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4204 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4205 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4206 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4207 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4209 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4210 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4212 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4214 * config/gcn/constraints.md (DA): Update description and match.
4216 (Db): New constraint.
4217 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4219 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4220 Implement 'Db' mixed immediate type.
4221 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4222 (addcv64si3_dup<exec_vcc>): Delete.
4223 (subcv64si3<exec_vcc>): Rework constraints.
4224 (addv64di3): Rework constraints.
4225 (addv64di3_exec): Rework constraints.
4226 (subv64di3): Rework constraints.
4227 (addv64di3_dup): Delete.
4228 (addv64di3_dup_exec): Delete.
4229 (addv64di3_zext): Rework constraints.
4230 (addv64di3_zext_exec): Rework constraints.
4231 (addv64di3_zext_dup): Rework constraints.
4232 (addv64di3_zext_dup_exec): Rework constraints.
4233 (addv64di3_zext_dup2): Rework constraints.
4234 (addv64di3_zext_dup2_exec): Rework constraints.
4235 (addv64di3_sext_dup2): Rework constraints.
4236 (addv64di3_sext_dup2_exec): Rework constraints.
4238 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4240 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4241 existing target checks.
4243 2020-01-07 Richard Biener <rguenther@suse.de>
4245 * doc/install.texi: Bump minimal supported MPC version.
4247 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4249 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4250 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4251 * langhooks.c: Include stor-layout.h.
4252 (lhd_simulate_enum_decl): New function.
4253 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4254 handle_arm_sve_h for the LTO frontend.
4255 (register_vector_type): Cope with null returns from pushdecl.
4257 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4259 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4260 (aarch64_sve::nvectors_if_data_type): Replace with...
4261 (aarch64_sve::builtin_type_p): ...this.
4262 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4263 (find_vector_type): Delete.
4264 (add_sve_type_attribute): New function.
4265 (lookup_sve_type_attribute): Likewise.
4266 (register_builtin_types): Add an "SVE type" attribute to each type.
4267 (register_tuple_type): Likewise.
4268 (svbool_type_p, nvectors_if_data_type): Delete.
4269 (mangle_builtin_type): Use lookup_sve_type_attribute.
4270 (builtin_type_p): Likewise. Add an overload that returns the
4271 number of constituent vector and predicate registers.
4272 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4273 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4274 instead of aarch64_sve_argument_p.
4275 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4276 (aarch64_pass_by_reference): Likewise.
4277 (aarch64_function_value_1): Likewise.
4278 (aarch64_return_in_memory): Likewise.
4279 (aarch64_layout_arg): Likewise.
4281 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4283 PR tree-optimization/93156
4284 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4285 least significant bit is always clear.
4287 PR tree-optimization/93118
4288 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4289 simplifier with two intermediate conversions.
4291 2020-01-07 Martin Liska <mliska@suse.cz>
4293 * params.opt: Add Optimization for various parameters.
4295 2020-01-07 Martin Liska <mliska@suse.cz>
4298 * doc/extend.texi: Explain cloning for target_clone
4301 2020-01-07 Martin Liska <mliska@suse.cz>
4303 PR tree-optimization/92860
4304 * common.opt: Make in Optimization option
4305 as it is affected by -O0, which is an Optimization
4307 * tree-inline.c (tree_inlinable_function_p):
4308 Use opt_for_fn for warn_inline.
4309 (expand_call_inline): Likewise.
4311 2020-01-07 Martin Liska <mliska@suse.cz>
4313 PR tree-optimization/92860
4314 * common.opt: Make flag_ree as optimization
4317 2020-01-07 Martin Liska <mliska@suse.cz>
4319 PR optimization/92860
4320 * params.opt: Mark param_min_crossjump_insns with Optimization
4323 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4325 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4326 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4328 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4330 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4331 helper function to return the valid addressing formats for a given
4332 hard register and mode.
4333 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4335 * config/rs6000/constraints.md (Q constraint): Update
4337 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4340 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4341 Use 'Q' for doing vector extract from memory.
4342 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4344 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4345 doing vector extract from memory.
4346 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4347 extract from memory.
4349 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4350 for the offset being 34-bits when -mcpu=future is used.
4352 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4354 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4355 instead of cmpib_comparison_operator in cmpib patterns.
4356 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4357 of cmpib_comparison_operator. Revise comment.
4359 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4361 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4362 in an IFN_DIV_POW2 node to be equal.
4364 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4366 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4367 (vect_check_scalar_mask): ...this.
4368 (vectorizable_store, vectorizable_load): Update call accordingly.
4369 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4370 argument in calls to conditional internal functions.
4372 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4374 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4375 '0' matching inputs.
4376 (subv64di3_exec): Likewise.
4378 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4380 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4381 * doc/md.texi (movstr): Likewise.
4383 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4385 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4388 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4390 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4392 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4393 to a temporary file and use move-if-change to update the real
4394 file where necessary.
4396 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4398 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4399 rather than Upa for CPY /M.
4401 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4403 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4406 2020-01-06 Martin Liska <mliska@suse.cz>
4408 PR tree-optimization/92860
4409 * params.opt: Mark param_max_combine_insns with Optimization
4412 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4415 * config/i386/i386.md (SWIDWI): New mode iterator.
4416 (DWI, dwi): Add TImode variants.
4417 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4418 <general_hilo_operand> instead of <general_operand>. Use
4419 CONST_SCALAR_INT_P instead of CONST_INT_P.
4420 (*addv<mode>4_1): Rename to ...
4421 (addv<mode>4_1): ... this.
4422 (QWI): New mode attribute.
4423 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4424 define_insn_and_split patterns.
4425 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4427 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4428 <general_hilo_operand> instead of <general_operand>.
4429 (*addcarry<mode>_1): New define_insn.
4430 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4432 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4434 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4435 Use "call" instead of "set".
4437 2020-01-03 Martin Jambor <mjambor@suse.cz>
4440 * ipa-cp.c (print_all_lattices): Skip functions without info.
4442 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4445 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4446 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4447 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4448 for 'e' simd clones.
4451 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4453 (mprefer-vector-width=): Add Save.
4454 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4455 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4456 (ix86_debug_options, ix86_function_specific_print): Adjust
4457 ix86_target_string callers.
4458 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4459 (ix86_valid_target_attribute_tree): Likewise.
4460 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4461 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4462 ix86_target_string caller.
4465 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4466 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4467 instead of gen_int_shift_amount + convert_modes.
4469 PR rtl-optimization/93088
4470 * loop-iv.c (find_single_def_src): Punt after looking through
4471 128 reg copies for regs with single definitions. Move definitions
4474 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4476 * config/arm/arm-c.c (arm_cpu_builtins): Define
4477 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4478 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4479 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4480 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4481 * config/arm/arm-tables.opt: Regenerated.
4482 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4483 arm_arch_i8mm and arm_arch_bf16 when enabled.
4484 * config/arm/arm.h (TARGET_I8MM): New macro.
4485 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4486 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4487 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4488 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4489 (v8_6_a_simd_variants): New.
4490 (v8_*_a_simd_variants): Add i8mm and bf16.
4491 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4493 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4496 * predict.c (compute_function_frequency): Don't call
4497 warn_function_cold on functions that already have cold attribute.
4499 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4502 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4503 COMDAT group function labels in .data.rel.ro.local section.
4504 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4507 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4508 comparison_operator in B and S integer comparisons. Likewise, use
4509 ordered_comparison_operator instead of cmpib_comparison_operator in
4511 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4513 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4515 Update copyright years.
4517 * gcc.c (process_command): Update copyright notice dates.
4518 * gcov-dump.c (print_version): Ditto.
4519 * gcov.c (print_version): Ditto.
4520 * gcov-tool.c (print_version): Ditto.
4521 * gengtype.c (create_file): Ditto.
4522 * doc/cpp.texi: Bump @copying's copyright year.
4523 * doc/cppinternals.texi: Ditto.
4524 * doc/gcc.texi: Ditto.
4525 * doc/gccint.texi: Ditto.
4526 * doc/gcov.texi: Ditto.
4527 * doc/install.texi: Ditto.
4528 * doc/invoke.texi: Ditto.
4530 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4532 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4535 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4537 PR tree-optimization/93098
4538 * match.pd (popcount): For shift amounts, use integer_onep
4539 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4540 tests. Make sure that precision is power of two larger than or equal
4541 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4542 instead of ULL suffixed constants. Formatting fixes.
4544 Copyright (C) 2020 Free Software Foundation, Inc.
4546 Copying and distribution of this file, with or without modification,
4547 are permitted in any medium without royalty provided the copyright
4548 notice and this notice are preserved.