1 2020-02-24 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/93582
4 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
5 pd.offset and pd.size to be counted in bits rather than bytes, add
6 support for maxsizei that is not a multiple of BITS_PER_UNIT and
7 handle bitfield stores and loads.
8 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
9 uncomparable quantities - bytes vs. bits. Allow push_partial_def
10 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
11 pd.offset/pd.size to be counted in bits rather than bytes.
12 Formatting fix. Rename shadowed len variable to buflen.
14 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
18 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
19 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
20 * opts-common.c (parse_options_from_collect_gcc_options): New function.
21 (prepend_xassembler_to_collect_as_options): Likewise.
22 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
23 (prepend_xassembler_to_collect_as_options): Likewise.
24 * lto-opts.c (lto_write_options): Stream assembler options
25 in COLLECT_AS_OPTIONS.
26 * lto-wrapper.c (xassembler_options_error): New static variable.
27 (get_options_from_collect_gcc_options): Move parsing options code to
28 parse_options_from_collect_gcc_options and call it.
29 (merge_and_complain): Validate -Xassembler options.
30 (append_compiler_options): Handle OPT_Xassembler.
31 (run_gcc): Append command line -Xassembler options to
33 * doc/invoke.texi: Add documentation about using Xassembler
36 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
38 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
40 (riscv_rtx_costs): Update cost model for LTGT.
42 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
44 PR rtl-optimization/93564
45 * ira-color.c (struct update_cost_queue_elem): New member start.
46 (queue_update_cost, get_next_update_cost): Add new arg start.
47 (allocnos_conflict_p): New function.
48 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
49 Add checking conflicts with allocnos_conflict_p.
50 (update_costs_from_prefs, restore_costs_from_copies): Adjust
51 update_costs_from_allocno calls.
52 (update_conflict_hard_regno_costs): Add checking conflicts with
53 allocnos_conflict_p. Adjust calls of queue_update_cost and
55 (assign_hard_reg): Adjust calls of queue_update_cost. Add
57 (bucket_allocno_compare_func): Restore previous version.
59 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
61 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
62 double-word size when handling aggregate return values.
63 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
64 that homogeneous SFmode and DFmode aggregates are passed and returned
67 2020-02-21 Jakub Jelinek <jakub@redhat.com>
70 * opts.c (print_filtered_help): Translate help before appending
71 messages to it rather than after that.
73 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
75 PR rtl-optimization/PR92989
76 * lra-lives.c (process_bb_lives): Restore the original order
77 of the bb liveness update. Call make_hard_regno_dead for each
78 register clobbered at the start of an EH receiver.
80 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
83 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
84 self-recursively generated.
86 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
89 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
92 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
94 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
95 Document new target supports option.
97 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
99 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
100 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
101 * config/arm/iterators.md (MATMUL): New iterator.
102 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
103 (mmla_sfx): New attribute.
104 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
105 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
106 (UNSPEC_MATMUL_US): New.
108 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
110 * config/arm/arm.md: Prevent scalar shifts from being used when big
113 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
114 Richard Biener <rguenther@suse.de>
116 PR tree-optimization/93586
117 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
118 after mismatched array refs; do not sure type size information to
119 recover from unmatched referneces with !flag_strict_aliasing_p.
121 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
123 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
124 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
125 (scatter_store<mode>): Rename to ...
126 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
127 (scatter<mode>_exec): Delete. Move contents ...
128 (mask_scatter_store<mode>): ... here, and rename that to ...
129 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
130 Remove mode conversion.
131 (mask_gather_load<mode>): Rename to ...
132 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
133 Remove mode conversion.
134 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
136 2020-02-21 Martin Jambor <mjambor@suse.cz>
138 PR tree-optimization/93845
139 * tree-sra.c (verify_sra_access_forest): Only test access size of
142 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
144 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
145 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
146 (addv64di3_exec): Likewise.
147 (subv64di3): Likewise.
148 (subv64di3_exec): Likewise.
149 (addv64di3_zext): Likewise.
150 (addv64di3_zext_exec): Likewise.
151 (addv64di3_zext_dup): Likewise.
152 (addv64di3_zext_dup_exec): Likewise.
153 (addv64di3_zext_dup2): Likewise.
154 (addv64di3_zext_dup2_exec): Likewise.
155 (addv64di3_sext_dup2): Likewise.
156 (addv64di3_sext_dup2_exec): Likewise.
157 (<expander>v64di3): Likewise.
158 (<expander>v64di3_exec): Likewise.
159 (*<reduc_op>_dpp_shr_v64di): Likewise.
160 (*plus_carry_dpp_shr_v64di): Likewise.
161 * config/gcn/gcn.md (adddi3): Likewise.
162 (addptrdi3): Likewise.
163 (<expander>di3): Likewise.
165 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
167 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
169 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
171 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
172 support. Use aarch64_emit_mult instead of emitting multiplication
173 instructions directly.
174 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
175 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
177 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
179 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
180 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
181 instead of emitting multiplication instructions directly.
182 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
183 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
184 (@aarch64_frecps<mode>): New expanders.
186 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
188 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
189 on and produce uint64_ts rather than ints.
190 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
191 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
193 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
195 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
196 an unused xmsk register when handling approximate rsqrt.
198 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
200 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
201 flag_finite_math_only condition.
203 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
206 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
207 to destination operand for shufps alternative.
208 (*vec_extractv2si_1): Ditto.
210 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
213 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
216 2020-02-20 Martin Liska <mliska@suse.cz>
219 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
221 2020-02-20 Martin Liska <mliska@suse.cz>
224 * common/config/avr/avr-common.c: Remote trailing "|".
226 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
228 * collect2.c (maybe_run_lto_and_relink): Fix typo in
231 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
233 PR tree-optimization/93767
234 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
235 access-size bias from the offset calculations for negative strides.
237 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
239 * collect2.c (c_file, o_file): Make const again.
240 (ldout,lderrout, dump_ld_file): Remove.
241 (tool_cleanup): Avoid calling not signal-safe functions.
242 (maybe_run_lto_and_relink): Avoid possible signal handler
243 access to unintialzed memory (lto_o_files).
244 (main): Avoid leaking temp files in $TMPDIR.
245 Initialize c_file/o_file with concat, which avoids exposing
246 uninitialized memory to signal handler, which calls unlink(!).
247 Avoid calling maybe_unlink when the main function returns,
248 since the atexit handler is already doing this.
249 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
251 2020-02-19 Martin Jambor <mjambor@suse.cz>
253 PR tree-optimization/93776
254 * tree-sra.c (create_access): Do not create zero size accesses.
255 (get_access_for_expr): Do not search for zero sized accesses.
257 2020-02-19 Martin Jambor <mjambor@suse.cz>
259 PR tree-optimization/93667
260 * tree-sra.c (scalarizable_type_p): Return false if record fields
261 do not follow wach other.
263 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
265 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
266 rather than fmv.x.s/fmv.s.x.
268 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
270 * config/aarch64/aarch64-simd-builtins.def
271 (intrinsic_vec_smult_lo_): New.
272 (intrinsic_vec_umult_lo_): Likewise.
273 (vec_widen_smult_hi_): Likewise.
274 (vec_widen_umult_hi_): Likewise.
275 * config/aarch64/aarch64-simd.md
276 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
277 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
278 (vmull_high_s16): Likewise.
279 (vmull_high_s32): Likewise.
280 (vmull_high_u8): Likewise.
281 (vmull_high_u16): Likewise.
282 (vmull_high_u32): Likewise.
283 (vmull_s8): Likewise.
284 (vmull_s16): Likewise.
285 (vmull_s32): Likewise.
286 (vmull_u8): Likewise.
287 (vmull_u16): Likewise.
288 (vmull_u32): Likewise.
290 2020-02-18 Martin Liska <mliska@suse.cz>
292 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
293 bootstrap by missing removal of invalid sanity check.
295 2020-02-18 Martin Liska <mliska@suse.cz>
298 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
299 Always compare LHS of gimple_assign.
301 2020-02-18 Martin Liska <mliska@suse.cz>
304 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
305 and return type of functions.
306 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
307 Drop MALLOC attribute for void functions.
308 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
309 malloc_state for a new VOID clone.
311 2020-02-18 Martin Liska <mliska@suse.cz>
314 * common.opt: Add -fprofile-reproducibility.
315 * doc/invoke.texi: Document it.
316 * value-prof.c (dump_histogram_value):
317 Document and support behavior for counters[0]
318 being a negative value.
319 (get_nth_most_common_value): Handle negative
320 counters[0] in respect to flag_profile_reproducible.
322 2020-02-18 Jakub Jelinek <jakub@redhat.com>
325 * cgraph.c (verify_speculative_call): Use speculative_id instead of
326 speculative_uid in messages. Remove trailing whitespace from error
327 message. Use num_speculative_call_targets instead of
328 num_speculative_targets in a message.
329 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
330 edge messages and stmt instead of cal_stmt in reference message.
332 PR tree-optimization/93780
333 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
334 before calling build_vector_type.
335 (execute_update_addresses_taken): Likewise.
338 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
339 typo, functoin -> function.
340 * tree.c (free_lang_data_in_decl): Fix comment typo,
341 functoin -> function.
342 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
344 2020-02-17 David Malcolm <dmalcolm@redhat.com>
346 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
348 (print_option_information): Don't call get_option_url if URLs
351 2020-02-17 Alexandre Oliva <oliva@adacore.com>
353 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
354 handling of register_common-less targets.
356 2020-02-17 Martin Liska <mliska@suse.cz>
359 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
361 2020-02-17 Martin Liska <mliska@suse.cz>
364 * config/rs6000/rs6000.c (rs6000_option_override_internal):
367 2020-02-17 Martin Liska <mliska@suse.cz>
370 * config/rx/elf.opt: Fix typo.
372 2020-02-17 Richard Biener <rguenther@suse.de>
375 * opts-global.c (print_ignored_options): Use inform and
378 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
381 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
383 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
386 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
387 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
389 2020-02-15 Jason Merrill <jason@redhat.com>
391 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
393 2020-02-15 Jakub Jelinek <jakub@redhat.com>
395 PR tree-optimization/93744
396 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
397 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
398 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
399 sure @2 in the first and @1 in the other patterns has no side-effects.
401 2020-02-15 David Malcolm <dmalcolm@redhat.com>
402 Bernd Edlinger <bernd.edlinger@hotmail.de>
406 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
407 * configure.ac (--with-diagnostics-urls): New configuration
408 option, based on --with-diagnostics-color.
409 (DIAGNOSTICS_URLS_DEFAULT): New define.
410 * config.h: Regenerate.
411 * configure: Regenerate.
412 * diagnostic.c (diagnostic_urls_init): Handle -1 for
413 DIAGNOSTICS_URLS_DEFAULT from configure-time
414 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
415 and TERM_URLS environment variable.
416 * diagnostic-url.h (diagnostic_url_format): New enum type.
417 (diagnostic_urls_enabled_p): rename to...
418 (determine_url_format): ... this, and change return type.
419 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
420 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
421 the linux console, and mingw.
422 (diagnostic_urls_enabled_p): rename to...
423 (determine_url_format): ... this, and adjust.
424 * pretty-print.h (pretty_printer::show_urls): rename to...
425 (pretty_printer::url_format): ... this, and change to enum.
426 * pretty-print.c (pretty_printer::pretty_printer,
427 pp_begin_url, pp_end_url, test_urls): Adjust.
428 * doc/install.texi (--with-diagnostics-urls): Document the new
429 configuration option.
430 (--with-diagnostics-color): Document the existing interaction
431 with GCC_COLORS better.
432 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
433 vindex reference. Update description of defaults based on the above.
434 (-fdiagnostics-color): Update description of how -fdiagnostics-color
435 interacts with GCC_COLORS.
437 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
440 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
441 conjunction with TARGET_GNU_TLS in early return.
443 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
445 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
446 the mode is not wider than UNITS_PER_WORD.
448 2020-02-14 Martin Jambor <mjambor@suse.cz>
450 PR tree-optimization/93516
451 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
452 access of the same type as the parent.
453 (propagate_subaccesses_from_lhs): Likewise.
455 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
458 * config/i386/avx512vbmi2intrin.h
459 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
460 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
461 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
462 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
463 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
464 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
465 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
466 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
467 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
468 of lacking a closing parenthesis.
469 * config/i386/avx512vbmi2vlintrin.h
470 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
471 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
472 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
473 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
474 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
475 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
476 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
477 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
478 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
479 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
480 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
481 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
482 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
483 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
484 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
485 _mm_shldi_epi32, _mm_mask_shldi_epi32,
486 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
487 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
489 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
492 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
493 the target function entry.
495 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
497 * common/config/arc/arc-common.c (arc_option_optimization_table):
498 Disable if-conversion step when optimized for size.
500 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
502 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
503 R12-R15 are always in ARCOMPACT16_REGS register class.
504 * config/arc/arc.opt (mq-class): Deprecate.
505 * config/arc/constraint.md ("q"): Remove dependency on mq-class
507 * doc/invoke.texi (mq-class): Update text.
508 * common/config/arc/arc-common.c (arc_option_optimization_table):
511 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
513 * config/arc/arc.c (arc_insn_cost): New function.
514 (TARGET_INSN_COST): Define.
515 * config/arc/arc.md (cost): New attribute.
516 (add_n): Use arc_nonmemory_operand.
517 (ashlsi3_insn): Likewise, also update constraints.
518 (ashrsi3_insn): Likewise.
520 (add_shift): Likewise.
521 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
523 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
525 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
527 (umulsidi_600): Likewise.
529 2020-02-13 Jakub Jelinek <jakub@redhat.com>
532 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
533 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
534 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
535 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
536 pass __A to the builtin followed by __W instead of __A followed by
538 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
539 _mm512_mask_popcnt_epi64): Likewise.
540 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
541 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
542 _mm256_mask_popcnt_epi64): Likewise.
544 PR tree-optimization/93582
545 * fold-const.h (shift_bytes_in_array_left,
546 shift_bytes_in_array_right): Declare.
547 * fold-const.c (shift_bytes_in_array_left,
548 shift_bytes_in_array_right): New function, moved from
549 gimple-ssa-store-merging.c, no longer static.
550 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
551 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
552 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
553 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
554 shift_bytes_in_array.
555 (verify_shift_bytes_in_array): Rename to ...
556 (verify_shift_bytes_in_array_left): ... this. Use
557 shift_bytes_in_array_left instead of shift_bytes_in_array.
558 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
559 instead of verify_shift_bytes_in_array.
560 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
561 / native_interpret_expr where the store covers all needed bits,
562 punt on PDP-endian, otherwise allow all involved offsets and sizes
563 not to be byte-aligned.
566 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
567 use const_0_to_255_operand predicate instead of immediate_operand.
568 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
569 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
570 vgf2p8affineinvqb_<mode><mask_name>,
571 vgf2p8affineqb_<mode><mask_name>): Drop mode from
572 const_0_to_255_operand predicated operands.
574 2020-02-12 Jeff Law <law@redhat.com>
576 * config/h8300/h8300.md (comparison shortening peepholes): Use
577 a mode iterator to merge the HImode and SImode peepholes.
579 2020-02-12 Jakub Jelinek <jakub@redhat.com>
582 * real.c (is_even): Make static. Function comment fix.
583 (is_halfway_below): Make static, don't assert R is not inf/nan,
584 instead return false for those. Small formatting fixes.
586 2020-02-12 Martin Sebor <msebor@redhat.com>
589 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
590 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
591 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
592 (strlen_check_and_optimize_call): Adjust callee name.
594 2020-02-12 Jeff Law <law@redhat.com>
596 * config/h8300/h8300.md (comparison shortening peepholes): Drop
597 (and (xor)) variant. Combine other two into single peephole.
599 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
601 PR rtl-optimization/93565
602 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
604 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
606 * config/aarch64/aarch64-simd.md
607 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
608 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
609 generating separate ADDV and zero_extend patterns.
610 * config/aarch64/iterators.md (VDQV_E): New iterator.
612 2020-02-12 Jeff Law <law@redhat.com>
614 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
615 expanders, splits, etc.
616 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
617 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
618 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
619 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
620 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
621 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
623 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
625 2020-02-12 Jakub Jelinek <jakub@redhat.com>
628 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
629 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
630 TARGET_AVX512DQ from condition.
631 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
632 instead of <mask_mode512bit_condition> in condition. If
633 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
635 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
638 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
641 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
643 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
645 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
646 where strlen is more legible.
647 (rs6000_builtin_vectorized_libmass): Ditto.
648 (rs6000_print_options_internal): Ditto.
650 2020-02-11 Martin Sebor <msebor@redhat.com>
652 PR tree-optimization/93683
653 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
655 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
657 * config/rs6000/predicates.md (cint34_operand): Rename the
658 -mprefixed-addr option to be -mprefixed.
659 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
660 the -mprefixed-addr option to be -mprefixed.
661 (OTHER_FUTURE_MASKS): Likewise.
662 (POWERPC_MASKS): Likewise.
663 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
664 the -mprefixed-addr option to be -mprefixed. Change error
665 messages to refer to -mprefixed.
666 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
668 (rs6000_legitimate_offset_address_p): Likewise.
669 (rs6000_mode_dependent_address): Likewise.
670 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
671 "-mprefixed" for target attributes and pragmas.
672 (address_to_insn_form): Rename the -mprefixed-addr option to be
674 (rs6000_adjust_insn_length): Likewise.
675 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
676 -mprefixed-addr option to be -mprefixed.
677 (ASM_OUTPUT_OPCODE): Likewise.
678 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
679 -mprefixed-addr option to be -mprefixed.
680 * config/rs6000/rs6000.opt (-mprefixed): Rename the
681 -mprefixed-addr option to be prefixed. Change the option from
682 being undocumented to being documented.
683 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
684 -mprefixed option. Update the -mpcrel documentation to mention
687 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
689 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
690 including FIRST_PSEUDO_REGISTER - 1.
691 * ira-color.c (print_hard_reg_set): Ditto.
693 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
695 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
696 (USTERNOP_QUALIFIERS): New define.
697 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
698 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
699 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
700 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
701 * config/arm/arm_neon.h (vusdot_s32): New.
702 (vusdot_lane_s32): New.
703 (vusdotq_lane_s32): New.
704 (vsudot_lane_s32): New.
705 (vsudotq_lane_s32): New.
706 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
707 * config/arm/iterators.md (DOTPROD_I8MM): New.
708 (sup, opsuffix): Add <us/su>.
709 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
710 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
712 2020-02-11 Richard Biener <rguenther@suse.de>
714 PR tree-optimization/93661
715 PR tree-optimization/93662
716 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
718 * tree-sra.c (get_access_for_expr): Likewise.
720 2020-02-10 Jakub Jelinek <jakub@redhat.com>
723 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
724 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
725 Change condition from TARGET_AVX2 to TARGET_AVX.
727 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
730 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
733 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
735 Try to generate zero-based comparisons.
736 * config/cris/cris.c (cris_reduce_compare): New function.
737 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
738 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
739 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
741 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
744 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
745 in Thumb state and also as a destination in Arm state. Add T16
748 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
750 * md.texi (Define Subst): Match closing paren in example.
752 2020-02-10 Jakub Jelinek <jakub@redhat.com>
756 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
757 arguments of strncmp.
759 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
762 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
763 but different source value.
764 (adjust_callers_for_value_intersection): New function.
765 (gather_edges_for_value): Adjust order of callers to let a
766 non-self-recursive caller be the first element.
767 (self_recursive_pass_through_p): Add a new parameter "simple", and
768 check generalized self-recursive pass-through jump function.
769 (self_recursive_agg_pass_through_p): Likewise.
770 (find_more_scalar_values_for_callers_subset): Compute value from
771 pass-through jump function for self-recursive.
772 (intersect_with_plats): Cleanup previous implementation code for value
773 itersection with self-recursive call edge.
774 (intersect_with_agg_replacements): Likewise.
775 (intersect_aggregates_with_edge): Deduce value from pass-through jump
776 function for self-recursive call edge. Cleanup previous implementation
777 code for value intersection with self-recursive call edge.
778 (decide_whether_version_node): Remove dead callers and adjust order
779 to let a non-self-recursive caller be the first element.
781 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
783 * recog.c: Move pass_split_before_sched2 code in front of
784 pass_split_before_regstack.
785 (pass_data_split_before_sched2): Rename pass to split3 from split4.
786 (pass_data_split_before_regstack): Rename pass to split4 from split3.
787 (rest_of_handle_split_before_sched2): Remove.
788 (pass_split_before_sched2::execute): Unconditionally call
790 (enable_split_before_sched2): New function.
791 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
792 (pass_split_before_regstack::gate): Ditto.
793 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
794 Update name check for renamed split4 pass.
795 * config/sh/sh.c (register_sh_passes): Update pass insertion
796 point for renamed split4 pass.
798 2020-02-09 Jakub Jelinek <jakub@redhat.com>
800 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
801 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
802 copying them around between host and target.
804 2020-02-08 Andrew Pinski <apinski@marvell.com>
807 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
808 STRICT_ALIGNMENT also.
810 2020-02-08 Jim Wilson <jimw@sifive.com>
813 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
815 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
816 Jakub Jelinek <jakub@redhat.com>
819 * config/i386/i386.h (CALL_USED_REGISTERS): Make
820 xmm16-xmm31 call-used even in 64-bit ms-abi.
822 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
824 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
825 (simd_ummla, simd_usmmla): Likewise.
826 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
827 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
830 2020-02-07 Richard Biener <rguenther@suse.de>
833 * tree-inline.c (fold_marked_statements): Do a PRE walk,
834 skipping unreachable regions.
835 (optimize_inline_calls): Skip folding stmts when we didn't
838 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
841 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
842 Don't return aggregates with only SFmode and DFmode in SSE
844 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
846 2020-02-07 Jakub Jelinek <jakub@redhat.com>
849 * config/rs6000/rs6000-logue.c
850 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
851 if it fails, move rs into end_addr and retry. Add
852 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
853 the insn pattern doesn't describe well what exactly happens to
857 * config/i386/predicates.md (avx_identity_operand): Remove.
858 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
859 (avx_<castmode><avxsizesuffix>_<castmode>,
860 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
861 a VEC_CONCAT of the operand and UNSPEC_CAST.
862 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
863 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
867 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
868 recog_data.insn if distance_non_agu_define changed it.
870 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
873 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
874 we only had X-FORM (reg+reg) addressing for vectors. Also before
875 ISA 3.0, we only had X-FORM addressing for scalars in the
876 traditional Altivec registers.
878 2020-02-06 <zhongyunde@huawei.com>
879 Vladimir Makarov <vmakarov@redhat.com>
881 PR rtl-optimization/93561
882 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
885 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
887 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
890 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
892 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
893 where the low and the high 32 bits are equal to each other specially,
894 with an rldimi instruction.
896 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
898 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
900 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
902 * config/arm/arm-tables.opt: Regenerate.
904 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
907 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
908 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
909 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
911 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
913 PR rtl-optimization/87763
914 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
916 2020-02-06 Delia Burduv <delia.burduv@arm.com>
918 * config/aarch64/aarch64-simd-builtins.def
919 (bfmlaq): New built-in function.
920 (bfmlalb): New built-in function.
921 (bfmlalt): New built-in function.
922 (bfmlalb_lane): New built-in function.
923 (bfmlalt_lane): New built-in function.
924 * config/aarch64/aarch64-simd.md
925 (aarch64_bfmmlaqv4sf): New pattern.
926 (aarch64_bfmlal<bt>v4sf): New pattern.
927 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
928 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
929 (vbfmlalbq_f32): New intrinsic.
930 (vbfmlaltq_f32): New intrinsic.
931 (vbfmlalbq_lane_f32): New intrinsic.
932 (vbfmlaltq_lane_f32): New intrinsic.
933 (vbfmlalbq_laneq_f32): New intrinsic.
934 (vbfmlaltq_laneq_f32): New intrinsic.
935 * config/aarch64/iterators.md (BF_MLA): New int iterator.
936 (bt): New int attribute.
938 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
940 * config/i386/i386.md (*pushtf): Emit "#" instead of
941 calling gcc_unreachable in insn output.
944 (*pushsf_rex64): Ditto for alternatives other than 1.
945 (*pushsf): Ditto for alternatives other than 1.
947 2020-02-06 Martin Liska <mliska@suse.cz>
949 PR gcov-profile/91971
950 PR gcov-profile/93466
951 * coverage.c (coverage_init): Revert mangling of
952 path into filename. It can lead to huge filename length.
953 Creation of subfolders seem more natural.
955 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
958 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
959 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
960 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
962 2020-02-06 Jakub Jelinek <jakub@redhat.com>
965 * config/i386/predicates.md (avx_identity_operand): New predicate.
966 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
967 define_insn_and_split.
970 * omp-low.c (use_pointer_for_field): For nested constructs, also
971 look for map clauses on target construct.
972 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
973 taskreg_nesting_level.
976 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
977 shared clause, call omp_notice_variable on outer context if any.
979 2020-02-05 Jason Merrill <jason@redhat.com>
982 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
983 non-zero address even if weak and not yet defined.
985 2020-02-05 Martin Sebor <msebor@redhat.com>
987 PR tree-optimization/92765
988 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
989 * tree-ssa-strlen.c (compute_string_length): Remove.
990 (determine_min_objsize): Remove.
991 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
992 Avoid using type size as the upper bound on string length.
993 (handle_builtin_string_cmp): Add an argument. Adjust.
994 (strlen_check_and_optimize_call): Pass additional argument to
995 handle_builtin_string_cmp.
997 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
999 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1000 (*pushdi2_rex64 peephole2): Unconditionally split after
1002 (*ashl<mode>3_doubleword): Ditto.
1003 (*<shift_insn><mode>3_doubleword): Ditto.
1005 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1008 * config/rs6000/rs6000.c (get_vector_offset): Fix
1010 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1012 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1014 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1017 (Special Functions for Debugging the Analyzer): Update description
1018 of __analyzer_dump_exploded_nodes.
1020 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1023 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1024 include sets and not clobbers in the vzeroupper pattern.
1025 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1026 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1027 (*avx_vzeroupper_1): New define_insn_and_split.
1030 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1031 don't run when !optimize.
1032 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1035 2020-02-05 Richard Biener <rguenther@suse.de>
1038 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1039 checks before matching calls.
1041 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1043 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1044 function comment typo.
1047 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1048 simd_clone_create failed when i == 0, adjust clone->nargs by
1051 2020-02-05 Martin Liska <mliska@suse.cz>
1054 * doc/invoke.texi: Document that one should
1055 not combine ASLR and -fpch.
1057 2020-02-04 Richard Biener <rguenther@suse.de>
1059 PR tree-optimization/93538
1060 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1062 2020-02-04 Richard Biener <rguenther@suse.de>
1064 PR tree-optimization/91123
1065 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1066 (vn_walk_cb_data::last_vuse): New member.
1067 (vn_walk_cb_data::saved_operands): Likewsie.
1068 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1069 (vn_walk_cb_data::push_partial_def): Use finish.
1070 (vn_reference_lookup_2): Update last_vuse and use finish if
1071 we've saved operands.
1072 (vn_reference_lookup_3): Use finish and update calls to
1073 push_partial_defs everywhere. When translating through
1074 memcpy or aggregate copies save off operands and alias-set.
1075 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1076 operation for redundant store removal.
1078 2020-02-04 Richard Biener <rguenther@suse.de>
1080 PR tree-optimization/92819
1081 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1082 generating more stmts than before.
1084 2020-02-04 Martin Liska <mliska@suse.cz>
1086 * config/arm/arm.c (arm_gen_far_branch): Move the function
1087 outside of selftests.
1089 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1091 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1092 function to adjust PC-relative vector addresses.
1093 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1094 handle vectors with PC-relative addresses.
1096 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1098 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1100 (hard_reg_and_mode_to_addr_mask): Delete.
1101 (rs6000_adjust_vec_address): If the original vector address
1102 was REG+REG or REG+OFFSET and the element is not zero, do the add
1103 of the elements in the original address before adding the offset
1104 for the vector element. Use address_to_insn_form to validate the
1105 address using the register being loaded, rather than guessing
1106 whether the address is a DS-FORM or DQ-FORM address.
1108 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1110 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1111 to calculate the offset in memory from the start of a vector of a
1112 particular element. Add code to keep the element number in
1113 bounds if the element number is variable.
1114 (rs6000_adjust_vec_address): Move calculation of offset of the
1115 vector element to get_vector_offset.
1116 (rs6000_split_vec_extract_var): Do not do the initial AND of
1117 element here, move the code to get_vector_offset.
1119 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1121 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1124 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1126 * config/rs6000/constraints.md: Improve documentation.
1128 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1131 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1132 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1134 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1136 * config.gcc: Remove "carrizo" support.
1137 * config/gcn/gcn-opts.h (processor_type): Likewise.
1138 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1139 * config/gcn/gcn.opt (gpu_type): Likewise.
1140 * config/gcn/t-omp-device: Likewise.
1142 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1145 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1146 * config/arm/arm.c (arm_gen_far_branch): New function
1148 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1150 2020-02-03 Julian Brown <julian@codesourcery.com>
1151 Tobias Burnus <tobias@codesourcery.com>
1153 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1155 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1158 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1159 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1162 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1164 PR rtl-optimization/91333
1165 * ira-color.c (struct allocno_color_data): Add member
1167 (init_allocno_threads): Set the member up.
1168 (bucket_allocno_compare_func): Add compare hard reg
1171 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1173 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1175 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1176 * config.in: Regenerated.
1177 * configure: Regenerated.
1178 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1179 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1180 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1182 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1184 * configure: Regenerate.
1186 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1188 PR rtl-optimization/91333
1189 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1190 reg preferences comparison up.
1192 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1194 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1195 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1196 aarch64-sve-builtins-base.h.
1197 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1198 aarch64-sve-builtins-base.cc.
1199 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1200 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1202 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1203 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1204 (svcvtnt): New functions.
1205 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1206 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1207 (svcvtnt): New functions.
1208 (svcvt): Add a form that converts f32 to bf16.
1209 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1210 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1212 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1213 Treat B as bfloat16_t.
1214 (ternary_bfloat_lane_base): New class.
1215 (ternary_bfloat_def): Likewise.
1216 (ternary_bfloat): New shape.
1217 (ternary_bfloat_lane_def): New class.
1218 (ternary_bfloat_lane): New shape.
1219 (ternary_bfloat_lanex2_def): New class.
1220 (ternary_bfloat_lanex2): New shape.
1221 (ternary_bfloat_opt_n_def): New class.
1222 (ternary_bfloat_opt_n): New shape.
1223 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1224 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1225 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1226 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1227 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1228 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1229 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1230 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1231 the pattern off the narrow mode instead of the wider one.
1232 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1233 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1234 (sve_fp_op): Handle them.
1235 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1236 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1238 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1240 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1241 * config/aarch64/aarch64-modes.def (BF): Move definition before
1242 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1243 (SVE_MODES): Handle BF modes.
1244 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1246 (aarch64_full_sve_mode): Likewise.
1247 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1249 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1250 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1251 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1252 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1254 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1256 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1257 (TYPES_all_data): Add bf16.
1258 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1259 (register_tuple_type): Increase buffer size.
1260 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1261 (bf16): New type suffix.
1262 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1263 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1264 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1265 Change type from all_data to all_arith.
1266 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1269 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1270 Matthew Malcomson <matthew.malcomson@arm.com>
1271 Richard Sandiford <richard.sandiford@arm.com>
1273 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1274 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1275 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1276 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1277 __ARM_FEATURE_MATMUL_FP64.
1278 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1279 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1280 be disabled at the same time.
1281 (f32mm): New extension.
1282 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1283 (AARCH64_FL_F64MM): Bump to the next bit up.
1284 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1285 (TARGET_SVE_F64MM): New macros.
1286 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1287 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1288 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1289 (UNSPEC_ZIP2Q): New unspeccs.
1290 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1291 (optab, sur, perm_insn): Handle the new unspecs.
1292 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1293 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1294 TARGET_SVE_F64MM instead of separate tests.
1295 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1296 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1297 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1298 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1299 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1300 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1301 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1302 (TYPES_s_signed): New macro.
1303 (TYPES_s_integer): Use it.
1304 (TYPES_d_float): New macro.
1305 (TYPES_d_data): Use it.
1306 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1307 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1308 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1309 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1310 (svmmla): New shape.
1311 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1312 template parameters.
1313 (ternary_resize2_lane_base): Likewise.
1314 (ternary_resize2_base): New class.
1315 (ternary_qq_lane_base): Likewise.
1316 (ternary_intq_uintq_lane_def): Likewise.
1317 (ternary_intq_uintq_lane): New shape.
1318 (ternary_intq_uintq_opt_n_def): New class
1319 (ternary_intq_uintq_opt_n): New shape.
1320 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1321 (ternary_uintq_intq_def): New class.
1322 (ternary_uintq_intq): New shape.
1323 (ternary_uintq_intq_lane_def): New class.
1324 (ternary_uintq_intq_lane): New shape.
1325 (ternary_uintq_intq_opt_n_def): New class.
1326 (ternary_uintq_intq_opt_n): New shape.
1327 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1328 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1329 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1330 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1332 (svdotprod_lane_impl): ...this new class.
1333 (svmmla_impl, svusdot_impl): New classes.
1334 (svdot_lane): Update to use svdotprod_lane_impl.
1335 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1336 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1338 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1339 function, with no types defined.
1340 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1341 AARCH64_FL_I8MM functions.
1342 (svmmla): New AARCH64_FL_F32MM function.
1343 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1344 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1345 AARCH64_FL_F64MM function.
1346 (REQUIRED_EXTENSIONS):
1348 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1350 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1353 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1355 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1356 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1357 (*movti_internal): Do not check for
1358 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1359 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1360 just after check for TARGET_AVX.
1361 (*movdf_internal): Ditto.
1362 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1363 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1364 * config/i386/sse.md (mov<mode>_internal): Only check
1365 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1366 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1367 (<sse>_andnot<mode>3<mask_name>): Move check for
1368 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1369 (<code><mode>3<mask_name>): Ditto.
1370 (*andnot<mode>3): Ditto.
1371 (*andnottf3): Ditto.
1372 (*<code><mode>3): Ditto.
1373 (*<code>tf3): Ditto.
1374 (*andnot<VI:mode>3): Remove
1375 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1376 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1377 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1378 (sse4_1_blendv<ssemodesuffix>): Ditto.
1379 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1380 Explain that tune applies to 128bit instructions only.
1382 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1384 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1385 to definition of hsa_kernel_description. Parse assembly to find SGPR
1386 and VGPR count of kernel and store in hsa_kernel_description.
1388 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1390 PR rtl-optimization/91838
1391 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1392 to truncate if allowed or reject combination.
1394 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1396 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1397 (find_inv_vars_cb): Likewise.
1399 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1401 * calls.c (special_function_p): Split out the check for DECL_NAME
1402 being non-NULL and fndecl being extern at file scope into a
1403 new maybe_special_function_p and call it. Drop check for fndecl
1404 being non-NULL that was after a usage of DECL_NAME (fndecl).
1405 * tree.h (maybe_special_function_p): New inline function.
1407 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1409 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1410 (mask_gather_load<mode>): ... here, and zero-initialize the
1412 (maskload<mode>di): Zero-initialize the destination.
1415 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1418 * doc/analyzer.texi (Limitations): Note that constraints on
1419 floating-point values are currently ignored.
1421 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1424 * symtab.c (symtab_node::noninterposable_alias): If localalias
1425 already exists, but is not usable, append numbers after it until
1426 a unique name is found. Formatting fix.
1429 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1432 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1434 * config/gcn/gcn.c (print_operand): Handle LTGT.
1435 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1437 2020-01-30 Richard Biener <rguenther@suse.de>
1439 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1440 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1442 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1444 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1445 without a DECL in .data.rel.ro.local.
1447 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1450 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1454 * config/i386/sse.md
1455 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1456 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1457 any_extend code iterator instead of always zero_extend.
1458 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1459 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1460 Use any_extend code iterator instead of always zero_extend.
1461 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1462 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1463 Use any_extend code iterator instead of always zero_extend.
1464 (*sse2_pmovmskb_ext): New define_insn.
1465 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1468 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1469 (*popcountsi2_zext_falsedep): New define_insn.
1471 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1473 * config.in: Regenerated.
1474 * configure: Regenerated.
1476 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1479 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1480 LLVM's assembler changed the default in version 9.
1482 2020-01-24 Jeff Law <law@redhat.com>
1484 PR tree-optimization/89689
1485 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1487 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1491 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1493 PR rtl-optimization/87763
1494 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1495 simplification to handle subregs as well as bare regs.
1496 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1498 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1501 * ira.c (ira): Revert use of simplified LRA algorithm.
1503 2020-01-29 Martin Jambor <mjambor@suse.cz>
1505 PR tree-optimization/92706
1506 * tree-sra.c (struct access): Fields first_link, last_link,
1507 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1508 next_rhs_queued and grp_rhs_queued respectively, new fields
1509 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1510 (struct assign_link): Field next renamed to next_rhs, new field
1511 next_lhs. Updated comment.
1512 (work_queue_head): Renamed to rhs_work_queue_head.
1513 (lhs_work_queue_head): New variable.
1514 (add_link_to_lhs): New function.
1515 (relink_to_new_repr): Also relink LHS lists.
1516 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1517 (add_access_to_lhs_work_queue): New function.
1518 (pop_access_from_work_queue): Renamed to
1519 pop_access_from_rhs_work_queue.
1520 (pop_access_from_lhs_work_queue): New function.
1521 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1523 (child_would_conflict_in_lacc): Renamed to
1524 child_would_conflict_in_acc. Adjusted parameter names.
1525 (create_artificial_child_access): New parameter set_grp_read, use it.
1526 (subtree_mark_written_and_enqueue): Renamed to
1527 subtree_mark_written_and_rhs_enqueue.
1528 (propagate_subaccesses_across_link): Renamed to
1529 propagate_subaccesses_from_rhs.
1530 (propagate_subaccesses_from_lhs): New function.
1531 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1534 2020-01-29 Martin Jambor <mjambor@suse.cz>
1536 PR tree-optimization/92706
1537 * tree-sra.c (struct access): Adjust comment of
1538 grp_total_scalarization.
1539 (find_access_in_subtree): Look for single children spanning an entire
1541 (scalarizable_type_p): Allow register accesses, adjust callers.
1542 (completely_scalarize): Remove function.
1543 (scalarize_elem): Likewise.
1544 (create_total_scalarization_access): Likewise.
1545 (sort_and_splice_var_accesses): Do not track total scalarization
1547 (analyze_access_subtree): New parameter totally, adjust to new meaning
1548 of grp_total_scalarization.
1549 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1550 (can_totally_scalarize_forest_p): New function.
1551 (create_total_scalarization_access): Likewise.
1552 (create_total_access_and_reshape): Likewise.
1553 (total_should_skip_creating_access): Likewise.
1554 (totally_scalarize_subtree): Likewise.
1555 (analyze_all_variable_accesses): Perform total scalarization after
1556 subaccess propagation using the new functions above.
1557 (initialize_constant_pool_replacements): Output initializers by
1558 traversing the access tree.
1560 2020-01-29 Martin Jambor <mjambor@suse.cz>
1562 * tree-sra.c (verify_sra_access_forest): New function.
1563 (verify_all_sra_access_forests): Likewise.
1564 (create_artificial_child_access): Set parent.
1565 (analyze_all_variable_accesses): Call the verifier.
1567 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1569 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1570 if called on indirect edge.
1571 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1572 speculative call if needed.
1574 2020-01-29 Richard Biener <rguenther@suse.de>
1576 PR tree-optimization/93428
1577 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1578 permutation when the load node is created.
1579 (vect_analyze_slp_instance): Re-use it here.
1581 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1583 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1585 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1587 PR rtl-optimization/93272
1588 * ira-lives.c (process_out_of_region_eh_regs): New function.
1589 (process_bb_node_lives): Call it.
1591 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1593 * coverage.c (read_counts_file): Make error message lowercase.
1595 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1597 * profile-count.c (profile_quality_display_names): Fix ordering.
1599 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1602 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1603 hash only when edge is first within the sequence.
1604 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1605 (symbol_table::create_edge): Do not set target_prob.
1606 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1608 (cgraph_edge::make_speculative): Drop target_prob parameter.
1609 (cgraph_edge::speculative_call_info): Remove.
1610 (cgraph_edge::first_speculative_call_target): New member function.
1611 (update_call_stmt_hash_for_removing_direct_edge): New function.
1612 (cgraph_edge::resolve_speculation): Rewrite to new API.
1613 (cgraph_edge::speculative_call_for_target): New member function.
1614 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1615 multiple speculation targets.
1616 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1618 (verify_speculative_call): Verify that targets form an interval.
1619 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1620 (cgraph_edge::first_speculative_call_target): New member function.
1621 (cgraph_edge::next_speculative_call_target): New member function.
1622 (cgraph_edge::speculative_call_target_ref): New member function.
1623 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1624 (cgraph_edge): Remove target_prob.
1625 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1626 Fix handling of speculative calls.
1627 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1628 * ipa-fnsummary.c (analyze_function_body): Likewise.
1629 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1630 * ipa-profile.c (dump_histogram): Fix formating.
1631 (ipa_profile_generate_summary): Watch for overflows.
1632 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1633 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1634 (update_indirect_edges_after_inlining): Update to new API.
1635 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1637 * profile-count.h: (profile_probability::adjusted): New.
1638 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1639 updating of profile.
1640 * value-prof.c (gimple_ic_transform): Rename to ...
1641 (dump_ic_profile): ... this one; update dumping.
1642 (stream_in_histogram_value): Fix formating.
1643 (gimple_value_profile_transformations): Update.
1645 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1648 * config/i386/i386.md (*movoi_internal_avx): Remove
1649 TARGET_SSE_TYPELESS_STORES check.
1650 (*movti_internal): Prefer TARGET_AVX over
1651 TARGET_SSE_TYPELESS_STORES.
1652 (*movtf_internal): Likewise.
1653 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1654 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1655 from TARGET_SSE_TYPELESS_STORES.
1657 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1659 * diagnostic-core.h (warning_at): Rename overload to...
1660 (warning_meta): ...this.
1661 (emit_diagnostic_valist): Delete decl of overload taking
1662 diagnostic_metadata.
1663 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1664 (warning_at): Rename overload taking diagnostic_metadata to...
1665 (warning_meta): ...this.
1667 2020-01-28 Richard Biener <rguenther@suse.de>
1669 PR tree-optimization/93439
1670 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1671 * tree-cfg.c (move_sese_region_to_fn): ... here.
1672 (verify_types_in_gimple_reference): Verify used cliques are
1675 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1678 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1679 argument of a pointer to struct gcc_options and pass it to
1680 parse_mtune_ctrl_str.
1681 (ix86_function_specific_restore): Pass opts to
1682 set_ix86_tune_features.
1683 (ix86_option_override_internal): Likewise.
1684 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1685 gcc_options and use it for x_ix86_tune_ctrl_string.
1687 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1689 PR rtl-optimization/87763
1690 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1691 simplification to handle subregs as well as bare regs.
1692 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1694 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1696 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1697 for reduction chains that (now) include a call.
1699 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1701 PR tree-optimization/92822
1702 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1703 out the don't-care elements of a vector whose significant elements
1704 are duplicates, make the don't-care elements duplicates too.
1706 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1708 PR tree-optimization/93434
1709 * tree-predcom.c (split_data_refs_to_components): Record which
1710 components have had aliasing loads removed. Prevent store-store
1711 commoning for all such components.
1713 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1716 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1717 -1 or is_vshift is true, use new_vector with number of elts npatterns
1718 rather than new_unary_operation.
1720 PR tree-optimization/93454
1721 * gimple-fold.c (fold_array_ctor_reference): Perform
1722 elt_size.to_uhwi () just once, instead of calling it in every
1723 iteration. Punt if that value is above size of the temporary
1724 buffer. Decrease third native_encode_expr argument when
1725 bufoff + elt_sz is above size of buf.
1727 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1729 * config/mips/mips.c (mips_declare_object_name)
1730 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1732 2020-01-27 Martin Liska <mliska@suse.cz>
1734 PR gcov-profile/93403
1735 * tree-profile.c (gimple_init_gcov_profiler): Generate
1736 both __gcov_indirect_call_profiler_v4 and
1737 __gcov_indirect_call_profiler_v4_atomic.
1739 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1742 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1744 (@aarch64_split_simd_mov<mode>): Use it.
1745 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1746 Leave the vec_extract patterns to handle 2-element vectors.
1747 (aarch64_simd_mov_from_<mode>high): Likewise.
1748 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1749 (vec_extractv2dfv1df): Likewise.
1751 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1753 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1754 jump conditions for *compare_condjump<GPI:mode>.
1756 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1759 * digraph.cc (test_edge::test_edge): Specify template for base
1762 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1764 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1766 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1768 * config/arc/arc-protos.h (gen_mlo): Remove.
1769 (gen_mhi): Likewise.
1770 * config/arc/arc.c (AUX_MULHI): Define.
1771 (arc_must_save_reister): Special handling for r58/59.
1772 (arc_compute_frame_size): Consider mlo/mhi registers.
1773 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1775 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1776 mlo/mhi name selection.
1777 (arc_restore_callee_saves): Don't early restore blink when ISR.
1778 (arc_expand_prologue): Add mlo/mhi saving.
1779 (arc_expand_epilogue): Add mlo/mhi restoring.
1782 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1783 numbering when MUL64 option is used.
1784 (DWARF2_FRAME_REG_OUT): Define.
1785 * config/arc/arc.md (arc600_stall): New pattern.
1786 (VUNSPEC_ARC_ARC600_STALL): Define.
1787 (mulsi64): Use correct mlo/mhi registers.
1788 (mulsi_600): Clean it up.
1789 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1791 (mhi_operand): Likewise.
1793 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1794 Petro Karashchenko <petro.karashchenko@ring.com>
1796 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1797 attributes if needed.
1798 (prepare_move_operands): Generate special unspec instruction for
1800 (arc_isuncached_mem_p): Propagate uncached attribute to each
1802 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1803 (VUNSPEC_ARC_STDI): Likewise.
1804 (ALLI): New mode iterator.
1805 (mALLI): New mode attribute.
1806 (lddi): New instruction pattern.
1808 (stdidi_split): Split instruction for architectures which are not
1809 supporting ll64 option.
1810 (lddidi_split): Likewise.
1812 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1814 PR rtl-optimization/92989
1815 * lra-lives.c (process_bb_lives): Update the live-in set before
1816 processing additional clobbers.
1818 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1820 PR rtl-optimization/93170
1821 * cselib.c (cselib_invalidate_regno_val): New function, split out
1823 (cselib_invalidate_regno): ...here.
1824 (cselib_invalidated_by_call_p): New function.
1825 (cselib_process_insn): Iterate over all the hard-register entries in
1826 REG_VALUES and invalidate any that cross call-clobbered registers.
1828 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1830 * dojump.c (split_comparison): Use HONOR_NANS rather than
1831 HONOR_SNANS when splitting LTGT.
1833 2020-01-27 Martin Liska <mliska@suse.cz>
1836 * opts.c (print_filtered_help): Exclude language-specific
1837 options from --help=common unless enabled in all FEs.
1839 2020-01-27 Martin Liska <mliska@suse.cz>
1841 * opts.c (print_help): Exclude params from
1842 all except --help=param.
1844 2020-01-27 Martin Liska <mliska@suse.cz>
1847 * config/i386/i386-features.c (make_resolver_func):
1848 Align the code with ppc64 target implementation.
1849 Do not generate a unique name for resolver function.
1851 2020-01-27 Richard Biener <rguenther@suse.de>
1853 PR tree-optimization/93397
1854 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1855 converted reduction chain SLP graph adjustment.
1857 2020-01-26 Marek Polacek <polacek@redhat.com>
1860 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1863 2020-01-26 Jason Merrill <jason@redhat.com>
1866 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1869 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1871 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1872 (rx_setmem): Likewise.
1874 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1877 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1878 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1879 drop <di> from constraint of last operand.
1882 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1883 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1884 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1886 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1889 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1892 2020-01-24 Jeff Law <law@redhat.com>
1894 PR tree-optimization/92788
1895 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1898 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1901 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1902 *avx_vperm_broadcast_<mode>,
1903 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1904 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1905 Move before avx2_perm<mode>/avx512f_perm<mode>.
1908 * simplify-rtx.c (simplify_const_unary_operation,
1909 simplify_const_binary_operation): Punt for mode precision above
1910 MAX_BITSIZE_MODE_ANY_INT.
1912 2020-01-24 Andrew Pinski <apinski@marvell.com>
1914 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1917 2020-01-24 Jeff Law <law@redhat.com>
1920 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1921 for REGs. Call output_operand_lossage to get more reasonable
1924 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1926 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1927 gcn_fp_compare_operator.
1928 (vec_cmpu<mode>di): Use gcn_compare_operator.
1929 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1930 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1931 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1932 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1933 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1934 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1935 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1936 gcn_fp_compare_operator.
1937 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1938 gcn_fp_compare_operator.
1939 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1940 gcn_fp_compare_operator.
1941 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1942 gcn_fp_compare_operator.
1944 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1946 * doc/install.texi (Cross-Compiler-Specific Options): Document
1947 `--with-toolexeclibdir' option.
1949 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1951 * target.def (flags_regnum): Also mention effect on delay slot filling.
1952 * doc/tm.texi: Regenerate.
1954 2020-01-23 Jeff Law <law@redhat.com>
1956 PR translation/90162
1957 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1959 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1962 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1965 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1967 PR rtl-optimization/93402
1968 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1971 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1973 * config.in: Regenerated.
1974 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1975 for TARGET_LIBC_GNUSTACK.
1976 * configure: Regenerated.
1977 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1978 found to be 2.31 or greater.
1980 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1982 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1984 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1985 (mips_asm_file_end): New function. Delegate to
1986 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1987 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1989 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1992 * config/i386/i386-modes.def (POImode): New mode.
1993 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1994 * config/i386/i386.md (DPWI): New mode attribute.
1995 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1997 (QPWI): ... this. Use POI instead of OI for TImode.
1998 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1999 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2002 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2005 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2007 (speculation_tracker_rev): New pattern.
2008 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2009 Use speculation_tracker_rev to track the inverse condition.
2011 2020-01-23 Richard Biener <rguenther@suse.de>
2013 PR tree-optimization/93381
2014 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2015 alias-set of the def as argument and record the first one.
2016 (vn_walk_cb_data::first_set): New member.
2017 (vn_reference_lookup_3): Pass the alias-set of the current def
2018 to push_partial_def. Fix alias-set used in the aggregate copy
2020 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2021 * real.c (clear_significand_below): Fix out-of-bound access.
2023 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2026 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2027 New define_insn patterns.
2029 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2031 * doc/sourcebuild.texi (check-function-bodies): Add an
2032 optional target/xfail selector.
2034 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2036 PR rtl-optimization/93124
2037 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2038 bare USE and CLOBBER insns.
2040 2020-01-22 Andrew Pinski <apinski@marvell.com>
2042 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2044 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2047 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2048 diagnostic_manager into "ana" namespace.
2049 * selftest-run-tests.c (selftest::run_tests): Update for move of
2050 selftest::run_analyzer_selftests to
2051 ana::selftest::run_analyzer_selftests.
2053 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2055 * cfgexpand.c (union_stack_vars): Update the size.
2057 2020-01-22 Richard Biener <rguenther@suse.de>
2059 PR tree-optimization/93381
2060 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2061 throughout, handle all conversions the same.
2063 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2066 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2067 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2068 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2069 Call force_reg on high_in2 unconditionally.
2071 2020-01-22 Martin Liska <mliska@suse.cz>
2073 PR tree-optimization/92924
2074 * profile.c (compute_value_histograms): Divide
2077 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2080 * output.h (assemble_name_resolve): Declare.
2081 * varasm.c (assemble_name_resolve): New function.
2082 (assemble_name): Use it.
2083 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2085 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2087 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2088 update_web_docs_git instead of update_web_docs_svn.
2090 2020-01-21 Andrew Pinski <apinski@marvell.com>
2093 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2094 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2095 (*tlsgd_small_<mode>): Likewise.
2096 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2097 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2098 register. Convert that register back to dest using convert_mode.
2100 2020-01-21 Jim Wilson <jimw@sifive.com>
2102 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2105 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2106 Uros Bizjak <ubizjak@gmail.com>
2109 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2111 (legitimize_tls_address): Do GNU2 TLS address computation in
2112 ptr_mode and zero-extend result to Pmode.
2113 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2114 :P with :PTR and Pmode with ptr_mode.
2115 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2116 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2117 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2119 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2122 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2123 the last two operands are CONST_INT_P before using them as such.
2125 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2127 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2128 to get the integer element types.
2130 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2132 * config/aarch64/aarch64-sve-builtins.h
2133 (function_expander::convert_to_pmode): Declare.
2134 * config/aarch64/aarch64-sve-builtins.cc
2135 (function_expander::convert_to_pmode): New function.
2136 (function_expander::get_contiguous_base): Use it.
2137 (function_expander::prepare_gather_address_operands): Likewise.
2138 * config/aarch64/aarch64-sve-builtins-sve2.cc
2139 (svwhilerw_svwhilewr_impl::expand): Likewise.
2141 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2144 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2145 cfun->machine->label_is_assembled.
2146 (aarch64_print_patchable_function_entry): New.
2147 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2148 * config/aarch64/aarch64.h (struct machine_function): New field,
2151 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2154 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2157 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2160 * cgraph.c (cgraph_edge::resolve_speculation,
2161 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2162 call_stmt_site_hash.
2164 2020-01-21 Martin Liska <mliska@suse.cz>
2166 * config/rs6000/rs6000.c (common_mode_defined): Remove
2169 2020-01-21 Richard Biener <rguenther@suse.de>
2171 PR tree-optimization/92328
2172 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2173 type when value-numbering same-sized store by inserting a
2175 (eliminate_dom_walker::eliminate_stmt): When eliminating
2176 a redundant store handle bit-reinterpretation of the same value.
2178 2020-01-21 Andrew Pinski <apinski@marvel.com>
2181 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2183 (prepare_block_for_update): This. Use a worklist instead of
2186 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2188 * gcc/config/arm/arm.c (clear_operation_p):
2189 Initialise last_regno, skip first iteration
2190 based on the first_set value and use ints instead
2191 of the unnecessary HOST_WIDE_INTs.
2193 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2196 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2197 compare_mode other than SFmode or DFmode.
2199 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2202 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2203 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2204 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2206 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2208 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2210 2020-01-20 Andrew Pinski <apinski@marvell.com>
2213 * targhooks.c (default_print_patchable_function_entry): Use
2214 output_asm_insn to emit the nop instruction.
2216 2020-01-20 Fangrui Song <maskray@google.com>
2219 * targhooks.c (default_print_patchable_function_entry): Align to
2222 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2225 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2226 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2227 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2228 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2229 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2230 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2231 Remove the {q} suffix from lea.
2232 (*tls_dynamic_gnu2_call_64): Renamed to ...
2233 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2234 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2235 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2236 Pass Pmode to gen_tls_dynamic_gnu2_64.
2238 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2240 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2242 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2244 * config/aarch64/aarch64-sve-builtins-base.cc
2245 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2247 2020-01-20 Richard Biener <rguenther@suse.de>
2250 * dwarf2out.c (prune_unused_types): Unconditionally mark
2251 called function DIEs.
2253 2020-01-20 Martin Liska <mliska@suse.cz>
2255 PR tree-optimization/93199
2256 * tree-eh.c (struct leh_state): Add
2257 new field outer_non_cleanup.
2258 (cleanup_is_dead_in): Pass leh_state instead
2259 of eh_region. Add a checking that state->outer_non_cleanup
2260 points to outer non-clean up region.
2261 (lower_try_finally): Record outer_non_cleanup
2263 (lower_catch): Likewise.
2264 (lower_eh_filter): Likewise.
2265 (lower_eh_must_not_throw): Likewise.
2266 (lower_cleanup): Likewise.
2268 2020-01-20 Richard Biener <rguenther@suse.de>
2270 PR tree-optimization/93094
2271 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2272 (vect_transform_loop): Likewise.
2273 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2274 loop_vectorized_call to vect_transform_loop.
2275 * tree-vect-loop.c (vect_transform_loop): Pass down
2276 loop_vectorized_call to vect_loop_versioning.
2277 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2278 the earlier discovered loop_vectorized_call.
2280 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2282 * doc/contribute.texi: Update for SVN -> Git transition.
2283 * doc/install.texi: Likewise.
2285 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2288 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2289 speculative targets.
2290 (verify_speculative_call): New function
2291 (cgraph_node::verify_node): Use it.
2292 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2295 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2298 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2299 (cgraph_edge::make_direct): Remove all indirect targets.
2300 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2301 (cgraph_node::verify_node): Verify that only one call_stmt or
2302 lto_stmt_uid is set.
2303 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2305 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2306 (lto_output_ref): Simplify streaming of stmt.
2307 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2309 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2311 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2312 Mark parameter unused.
2314 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2316 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2318 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2320 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2322 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2324 * Makefile.in: Add coroutine-passes.o.
2325 * builtin-types.def (BT_CONST_SIZE): New.
2326 (BT_FN_BOOL_PTR): New.
2327 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2328 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2329 * coroutine-builtins.def: New file.
2330 * coroutine-passes.cc: New file.
2331 * function.h (struct GTY function): Add a bit to indicate that the
2332 function is a coroutine component.
2333 * internal-fn.c (expand_CO_FRAME): New.
2334 (expand_CO_YIELD): New.
2335 (expand_CO_SUSPN): New.
2336 (expand_CO_ACTOR): New.
2337 * internal-fn.def (CO_ACTOR): New.
2341 * passes.def: Add pass_coroutine_lower_builtins,
2342 pass_coroutine_early_expand_ifns.
2343 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2344 (make_pass_coroutine_early_expand_ifns): New.
2345 * doc/invoke.texi: Document the fcoroutines command line
2348 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2350 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2353 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2354 after checking the argument is a REG. Don't use REGNO (reg)
2355 again to set last_regno, reuse regno variable instead.
2357 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2359 * doc/analyzer.texi (Limitations): Add note about NaN.
2361 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2362 Sudakshina Das <sudi.das@arm.com>
2364 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2365 and valid immediate.
2366 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2367 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2368 * config/arm/constraints.md (Pg): New.
2369 * config/arm/predicates.md (long_shift_imm): New.
2370 (arm_reg_or_long_shift_imm): Likewise.
2371 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2372 (thumb2_lsll): Likewise.
2375 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2376 Sudakshina Das <sudi.das@arm.com>
2378 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2379 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2380 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2381 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2382 * config/arm/thumb2.md (thumb2_asrl): New.
2383 (thumb2_lsll): Likewise.
2385 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2387 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2390 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2392 * gdbinit.in (help-gcc-hooks): New command.
2393 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2394 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2397 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2399 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2400 correct target macro.
2402 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2404 * config/aarch64/aarch64-protos.h
2405 (aarch64_sve_ld1ro_operand_p): New.
2406 * config/aarch64/aarch64-sve-builtins-base.cc
2407 (class load_replicate): New.
2408 (class svld1ro_impl): New.
2409 (class svld1rq_impl): Change to inherit from load_replicate.
2410 (svld1ro): New sve intrinsic function base.
2411 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2412 New DEF_SVE_FUNCTION.
2413 * config/aarch64/aarch64-sve-builtins-base.h
2414 (svld1ro): New decl.
2415 * config/aarch64/aarch64-sve-builtins.cc
2416 (function_expander::add_mem_operand): Modify assert to allow
2418 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2420 * config/aarch64/aarch64.c
2421 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2422 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2423 (aarch64_sve_ld1ro_operand_p): New.
2424 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2425 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2426 * config/aarch64/predicates.md
2427 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2429 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2431 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2432 Introduce this ACLE specified predefined macro.
2433 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2434 (fp): Disabling this disables f64mm.
2435 (simd): Disabling this disables f64mm.
2436 (fp16): Disabling this disables f64mm.
2437 (sve): Disabling this disables f64mm.
2438 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2439 (AARCH64_ISA_F64MM): New.
2440 (TARGET_F64MM): New.
2441 * doc/invoke.texi (f64mm): Document new option.
2443 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2445 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2446 (neoversen1_tunings): Likewise.
2448 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2451 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2452 Add assert to ensure prolog has been emitted.
2453 (aarch64_split_atomic_op): Likewise.
2454 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2455 Use epilogue_completed rather than reload_completed.
2456 (aarch64_atomic_exchange<mode>): Likewise.
2457 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2458 (atomic_nand<mode>): Likewise.
2459 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2460 (atomic_fetch_nand<mode>): Likewise.
2461 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2462 (atomic_nand_fetch<mode>): Likewise.
2464 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2467 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2469 (REVERSE_CONDITION): Delete.
2470 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2471 (CCFP_CCFPE): Likewise.
2472 (e): New mode attribute.
2473 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2474 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2475 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2476 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2477 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2478 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2479 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2480 name of generator from gen_ccmpdi to gen_ccmpccdi.
2481 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2482 the previous comparison but aren't able to, use the new ccmp_rev
2485 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2487 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2488 than testing directly for INTEGER_CST.
2489 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2491 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2493 PR tree-optimization/93292
2494 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2495 get_vectype_for_scalar_type returns NULL.
2497 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2499 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2500 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2502 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2504 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2506 * params.opt: (max-predicted-iterations): Set bounds.
2507 * predict.c (real_almost_one, real_br_prob_base,
2508 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2509 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2510 probabilities; do not truncate to reg_br_prob_bases.
2511 (estimate_loops_at_level): Pass max_cyclic_prob.
2512 (estimate_loops): Compute max_cyclic_prob.
2513 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2515 * profile-count.c (profile_probability::to_sreal): New.
2516 * profile-count.h (class sreal): Move up in file.
2517 (profile_probability::to_sreal): Declare.
2519 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2522 (arm_invalid_conversion): New function for target hook.
2523 (arm_invalid_unary_op): New function for target hook.
2524 (arm_invalid_binary_op): New function for target hook.
2526 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2528 * config.gcc: Add arm_bf16.h.
2529 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2530 (arm_simd_builtin_std_type): Add BFmode.
2531 (arm_init_simd_builtin_types): Define element types for vector types.
2532 (arm_init_bf16_types): New function.
2533 (arm_init_builtins): Add arm_init_bf16_types function call.
2534 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2535 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2536 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2537 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2538 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2539 (arm_mangle_type): Add __bf16.
2540 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2541 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2542 arm_bf16_ptr_type_node.
2543 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2544 define_split between ARM registers.
2545 * config/arm/arm_bf16.h: New file.
2546 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2547 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2549 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2550 * config/arm/vfp.md: Add BFmode to movhf patterns.
2552 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2553 Andre Vieira <andre.simoesdiasvieira@arm.com>
2555 * config/arm/arm-cpus.in (mve, mve_float): New features.
2556 (dsp, mve, mve.fp): New options.
2557 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2558 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2559 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2561 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2562 Thomas Preud'homme <thomas.preudhomme@arm.com>
2564 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2566 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2567 error for using -mcmse when targeting Armv8.1-M Mainline.
2569 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2570 Thomas Preud'homme <thomas.preudhomme@arm.com>
2572 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2573 address in r4 when targeting Armv8.1-M Mainline.
2574 (nonsecure_call_value_internal): Likewise.
2575 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2576 a register match_operand again. Emit BLXNS when targeting
2578 (nonsecure_call_value_reg_thumb2): Likewise.
2580 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2581 Thomas Preud'homme <thomas.preudhomme@arm.com>
2583 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2584 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2585 variable as true when floating-point ABI is not hard. Replace
2586 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2587 Generate VLSTM and VLLDM instruction respectively before and
2588 after a function call to cmse_nonsecure_call function.
2589 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2590 (VUNSPEC_VLLDM): Likewise.
2591 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2592 (lazy_load_multiple_insn): Likewise.
2594 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2595 Thomas Preud'homme <thomas.preudhomme@arm.com>
2597 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2598 (arm_emit_vfp_multi_reg_pop): Likewise.
2599 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2600 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2601 restore callee-saved VFP registers.
2603 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2604 Thomas Preud'homme <thomas.preudhomme@arm.com>
2606 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2607 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2608 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2609 callee-saved GPRs as well as clear ip register before doing a nonsecure
2610 call then restore callee-saved GPRs after it when targeting
2612 (arm_reorg): Adapt to function rename.
2614 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2615 Thomas Preud'homme <thomas.preudhomme@arm.com>
2617 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2618 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2619 clear_vfp_multiple pattern based on a new vfp parameter.
2620 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2621 targeting Armv8.1-M Mainline.
2622 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2623 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2624 whether VFP registers are available before looking call_used_regs for a
2626 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2627 of prototype of clear_operation_p.
2628 (clear_vfp_multiple_operation): New predicate.
2629 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2630 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2632 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2633 Thomas Preud'homme <thomas.preudhomme@arm.com>
2635 * config/arm/arm-protos.h (clear_operation_p): Declare.
2636 * config/arm/arm.c (clear_operation_p): New function.
2637 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2638 targeting Armv8.1-M Mainline or successor.
2639 (output_return_instruction): Only output APSR register clearing if
2640 Armv8.1-M Mainline instructions not available.
2641 (thumb_exit): Likewise.
2642 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2643 * config/arm/thumb2.md (clear_apsr): New define_insn.
2644 (clear_multiple): Likewise.
2645 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2647 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2648 Thomas Preud'homme <thomas.preudhomme@arm.com>
2650 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2651 (use_return_insn): Also return false for Armv8.1-M Mainline.
2652 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2653 Mainline instructions are available.
2654 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2655 when targeting Armv8.1-M Mainline Security Extensions.
2656 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2657 Mainline entry function.
2658 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2659 targeting Armv8.1-M Mainline or successor.
2660 (arm_expand_epilogue): Fix indentation of caller-saved register
2661 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2663 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2664 (FP_SYSREGS): Likewise.
2665 (enum vfp_sysregs_encoding): Define enum.
2666 (fp_sysreg_names): Declare.
2667 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2668 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2669 (pop_fpsysreg_insn): Likewise.
2671 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2672 Thomas Preud'homme <thomas.preudhomme@arm.com>
2674 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2675 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2676 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2677 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2678 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2679 (ARMv8_1m_main): New feature group.
2680 (armv8.1-m.main): New architecture.
2681 * config/arm/arm-tables.opt: Regenerate.
2682 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2683 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2684 (arm_options_perform_arch_sanity_checks): Error out when targeting
2685 Armv8.1-M Mainline Security Extensions.
2686 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2688 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2690 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2691 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2692 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2693 aarch64_bfdot_laneq): New.
2694 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2695 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2696 vbfdotq_laneq_f32): New.
2697 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2698 VBFMLA_W, VBF): New.
2699 (isquadop): Add V4BF, V8BF.
2701 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2703 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2704 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2705 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2706 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2707 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2708 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2709 usdot_laneq, sudot_lane,sudot_laneq): New.
2710 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2711 (aarch64_<sur>dot_lane): New.
2712 * config/aarch64/arm_neon.h (vusdot_s32): New.
2714 (vusdot_lane_s32): New.
2715 (vsudot_lane_s32): New.
2716 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2717 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2719 2020-01-16 Martin Liska <mliska@suse.cz>
2721 * value-prof.c (dump_histogram_value): Fix
2722 obvious spacing issue.
2724 2020-01-16 Andrew Pinski <apinski@marvell.com>
2726 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2727 !storage_order_barrier_p.
2729 2020-01-16 Andrew Pinski <apinski@marvell.com>
2731 * sched-int.h (_dep): Add unused bit-field field for the padding.
2732 * sched-deps.c (init_dep_1): Init unused field.
2734 2020-01-16 Andrew Pinski <apinski@marvell.com>
2736 * optabs.h (create_expand_operand): Initialize target field also.
2738 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2740 PR tree-optimization/92429
2741 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2742 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2744 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2747 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2749 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2750 aarch64_sve_int_mode to each mode.
2752 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2754 * doc/analyzer.texi (Overview): Add note about
2755 -fdump-ipa-analyzer.
2757 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2759 PR tree-optimization/93231
2760 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2761 input_type is unsigned. Use tree_to_shwi for shift constant.
2762 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2763 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2764 inputs more efficiently.
2766 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2768 * config/i386/i386.md (*movsf_internal): Do not require
2769 SSE2 ISA for alternatives 14 and 15.
2771 2020-01-15 Richard Biener <rguenther@suse.de>
2774 * tree-eh.c (sink_clobbers): If we already visited the destination
2775 block do not defer insertion.
2776 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2777 the purpose of defered insertion.
2779 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2781 * BASE-VER: Bump to 10.0.1.
2783 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2785 PR tree-optimization/93247
2786 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2787 type of the stmt that we're going to vectorize.
2789 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2791 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2792 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2795 2020-01-15 Martin Liska <mliska@suse.cz>
2797 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2798 2 calls of streamer_read_hwi in a function call.
2800 2020-01-15 Richard Biener <rguenther@suse.de>
2802 * alias.c (record_alias_subset): Avoid redundant work when
2803 subset is already recorded.
2805 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2807 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2808 the analyzer options provide CWE identifiers.
2810 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2812 * tree-diagnostic-path.cc (path_summary::event_range::print):
2813 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2814 using get_pure_location.
2816 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2818 PR tree-optimization/93262
2819 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2820 perform head trimming only if the last argument is constant,
2821 either all ones, or larger or equal to head trim, in the latter
2822 case decrease the last argument by head_trim.
2824 PR tree-optimization/93249
2825 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2826 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2827 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2828 perform head trim unless we can prove there are no '\0' chars
2829 from the source among the first head_trim chars.
2831 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2833 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2835 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2838 * config/i386/sse.md
2839 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2840 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2841 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2842 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2843 just a single alternative instead of two, make operands 1 and 2
2846 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2849 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2852 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2854 * Makefile.in (lang_opt_files): Add analyzer.opt.
2855 (ANALYZER_OBJS): New.
2856 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2857 tristate.o and ANALYZER_OBJS.
2858 (TEXI_GCCINT_FILES): Add analyzer.texi.
2859 * common.opt (-fanalyzer): New driver option.
2860 * config.in: Regenerate.
2861 * configure: Regenerate.
2862 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2863 (gccdepdir): Also create depdir for "analyzer" subdir.
2864 * digraph.cc: New file.
2865 * digraph.h: New file.
2866 * doc/analyzer.texi: New file.
2867 * doc/gccint.texi ("Static Analyzer") New menu item.
2868 (analyzer.texi): Include it.
2869 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2870 ("Warning Options"): Add static analysis warnings to the list.
2871 (-Wno-analyzer-double-fclose): New option.
2872 (-Wno-analyzer-double-free): New option.
2873 (-Wno-analyzer-exposure-through-output-file): New option.
2874 (-Wno-analyzer-file-leak): New option.
2875 (-Wno-analyzer-free-of-non-heap): New option.
2876 (-Wno-analyzer-malloc-leak): New option.
2877 (-Wno-analyzer-possible-null-argument): New option.
2878 (-Wno-analyzer-possible-null-dereference): New option.
2879 (-Wno-analyzer-null-argument): New option.
2880 (-Wno-analyzer-null-dereference): New option.
2881 (-Wno-analyzer-stale-setjmp-buffer): New option.
2882 (-Wno-analyzer-tainted-array-index): New option.
2883 (-Wno-analyzer-use-after-free): New option.
2884 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2885 (-Wno-analyzer-use-of-uninitialized-value): New option.
2886 (-Wanalyzer-too-complex): New option.
2887 (-fanalyzer-call-summaries): New warning.
2888 (-fanalyzer-checker=): New warning.
2889 (-fanalyzer-fine-grained): New warning.
2890 (-fno-analyzer-state-merge): New warning.
2891 (-fno-analyzer-state-purge): New warning.
2892 (-fanalyzer-transitivity): New warning.
2893 (-fanalyzer-verbose-edges): New warning.
2894 (-fanalyzer-verbose-state-changes): New warning.
2895 (-fanalyzer-verbosity=): New warning.
2896 (-fdump-analyzer): New warning.
2897 (-fdump-analyzer-callgraph): New warning.
2898 (-fdump-analyzer-exploded-graph): New warning.
2899 (-fdump-analyzer-exploded-nodes): New warning.
2900 (-fdump-analyzer-exploded-nodes-2): New warning.
2901 (-fdump-analyzer-exploded-nodes-3): New warning.
2902 (-fdump-analyzer-supergraph): New warning.
2903 * doc/sourcebuild.texi (dg-require-dot): New.
2904 (dg-check-dot): New.
2905 * gdbinit.in (break-on-saved-diagnostic): New command.
2906 * graphviz.cc: New file.
2907 * graphviz.h: New file.
2908 * ordered-hash-map-tests.cc: New file.
2909 * ordered-hash-map.h: New file.
2910 * passes.def (pass_analyzer): Add before
2911 pass_ipa_whole_program_visibility.
2912 * selftest-run-tests.c (selftest::run_tests): Call
2913 selftest::ordered_hash_map_tests_cc_tests.
2914 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2916 * shortest-paths.h: New file.
2917 * timevar.def (TV_ANALYZER): New timevar.
2918 (TV_ANALYZER_SUPERGRAPH): Likewise.
2919 (TV_ANALYZER_STATE_PURGE): Likewise.
2920 (TV_ANALYZER_PLAN): Likewise.
2921 (TV_ANALYZER_SCC): Likewise.
2922 (TV_ANALYZER_WORKLIST): Likewise.
2923 (TV_ANALYZER_DUMP): Likewise.
2924 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2925 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2926 * tree-pass.h (make_pass_analyzer): New decl.
2927 * tristate.cc: New file.
2928 * tristate.h: New file.
2930 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2933 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2934 alternatives 9 and 10.
2936 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2938 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2939 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2940 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2941 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2942 (selftest::hash_map_tests_c_tests): Call it.
2943 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2944 New static constant, using the value of = H::empty_zero_p.
2945 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2946 from default_hash_traits <Value>.
2947 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2949 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2950 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2951 calls to mark_empty with !Descriptor::empty_zero_p.
2952 (hash_table::empty_slow): Conditionalize the memset call with a
2953 check that Descriptor::empty_zero_p; otherwise, loop through the
2954 entries calling mark_empty on them.
2955 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2956 (pointer_hash::empty_zero_p): Likewise.
2957 (pair_hash::empty_zero_p): Likewise.
2958 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2960 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2961 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2962 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2963 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2964 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2965 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2966 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2967 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2969 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2972 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2974 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2975 fix typo on return value.
2977 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2980 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2982 (cgraph_edge::make_speculative): Add param for setting speculative_id
2984 (cgraph_edge::speculative_call_info): Update comments and find reference
2985 by speculative_id for multiple indirect targets.
2986 (cgraph_edge::resolve_speculation): Decrease the speculations
2987 for indirect edge, drop it's speculative if not direct target
2988 left. Update comments.
2989 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2990 (cgraph_node::dump): Print num_speculative_call_targets.
2991 (cgraph_node::verify_node): Don't report error if speculative
2992 edge not include statement.
2993 (cgraph_edge::num_speculative_call_targets_p): New function.
2994 * cgraph.h (int common_target_id): Remove.
2995 (int common_target_probability): Remove.
2996 (num_speculative_call_targets): New variable.
2997 (make_speculative): Add param for setting speculative_id.
2998 (cgraph_edge::num_speculative_call_targets_p): New declare.
2999 (target_prob): New variable.
3000 (speculative_id): New variable.
3001 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3002 call summaries for multiple speculative call targets.
3003 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3004 * ipa-profile.c (struct speculative_call_target): New struct.
3005 (class speculative_call_summary): New class.
3006 (class speculative_call_summaries): New class.
3007 (call_sums): New variable.
3008 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3009 (ipa_profile_write_edge_summary): New function.
3010 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3011 (ipa_profile_dump_all_summaries): New function.
3012 (ipa_profile_read_edge_summary): New function.
3013 (ipa_profile_read_summary_section): New function.
3014 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3015 (ipa_profile): Generate num_speculative_call_targets from
3017 * ipa-ref.h (speculative_id): New variable.
3018 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3019 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3020 common_target_probability. Stream out speculative_id and
3021 num_speculative_call_targets.
3022 (input_edge): Likewise.
3023 * predict.c (dump_prediction): Remove edges count assert to be
3025 * symtab.c (symtab_node::create_reference): Init speculative_id.
3026 (symtab_node::clone_references): Clone speculative_id.
3027 (symtab_node::clone_referring): Clone speculative_id.
3028 (symtab_node::clone_reference): Clone speculative_id.
3029 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3030 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3031 if indirect call contains multiple speculative targets.
3032 * value-prof.h (check_ic_target): Remove.
3033 * value-prof.c (gimple_value_profile_transformations):
3034 Use void function gimple_ic_transform.
3035 * value-prof.c (gimple_ic_transform): Handle topn case.
3036 Fix comment typos. Change it to a void function.
3038 2020-01-13 Andrew Pinski <apinski@marvell.com>
3040 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3041 (octeontx2t98): New define.
3042 (octeontx2t96): New define.
3043 (octeontx2t93): New define.
3044 (octeontx2f95): New define.
3045 (octeontx2f95n): New define.
3046 (octeontx2f95mm): New define.
3047 * config/aarch64/aarch64-tune.md: Regenerate.
3048 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3050 2020-01-13 Jason Merrill <jason@redhat.com>
3052 PR c++/33799 - destroy return value if local cleanup throws.
3053 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3055 2020-01-13 Martin Liska <mliska@suse.cz>
3057 * ipa-cp.c (get_max_overall_size): Use newly
3058 renamed param param_ipa_cp_unit_growth.
3059 * params.opt: Remove legacy param name.
3061 2020-01-13 Martin Sebor <msebor@redhat.com>
3063 PR tree-optimization/93213
3064 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3065 stores to be eliminated.
3067 2020-01-13 Martin Liska <mliska@suse.cz>
3069 * opts.c (print_help): Do not print CL_PARAM
3070 and CL_WARNING for CL_OPTIMIZATION.
3072 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3075 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3076 depending on optimization settings.
3078 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3080 PR tree-optimization/90838
3081 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3082 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3083 argument rather than to initialize temporary for targets that
3084 don't use the mode argument at all. Initialize ctzval to avoid
3087 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3089 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3090 * tree-core.h: Document it.
3091 * gimplify.c (gimplify_omp_workshare): Set it.
3092 * omp-low.c (lower_omp_target): Use it.
3093 * tree-pretty-print.c (dump_omp_clause): Print it.
3095 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3096 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3098 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3100 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3101 * common.opt (fdiagnostics-path-format=): New option.
3102 (diagnostic_path_format): New enum.
3103 (fdiagnostics-show-path-depths): New option.
3104 * coretypes.h (diagnostic_event_id_t): New forward decl.
3105 * diagnostic-color.c (color_dict): Add "path".
3106 * diagnostic-event-id.h: New file.
3107 * diagnostic-format-json.cc (json_from_expanded_location): Make
3109 (json_end_diagnostic): Call context->make_json_for_path if it
3110 exists and the diagnostic has a path.
3111 (diagnostic_output_format_init): Clear context->print_path.
3112 * diagnostic-path.h: New file.
3113 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3114 when printing a run of events in a diagnostic_path so that they
3115 all get the same color.
3116 (layout::m_diagnostic_path_p): New field.
3117 (layout::layout): Initialize it.
3118 (layout::print_any_labels): Don't colorize the label text for an
3119 event in a diagnostic_path.
3120 (gcc_rich_location::add_location_if_nearby): Add
3121 "restrict_to_current_line_spans" and "label" params. Pass the
3122 former to layout.maybe_add_location_range; pass the latter
3123 when calling add_range.
3124 * diagnostic.c: Include "diagnostic-path.h".
3125 (diagnostic_initialize): Initialize context->path_format and
3126 context->show_path_depths.
3127 (diagnostic_show_any_path): New function.
3128 (diagnostic_path::interprocedural_p): New function.
3129 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3130 (simple_diagnostic_path::num_events): New function.
3131 (simple_diagnostic_path::get_event): New function.
3132 (simple_diagnostic_path::add_event): New function.
3133 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3134 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3135 (debug): New overload taking a diagnostic_path *.
3136 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3137 * diagnostic.h (enum diagnostic_path_format): New enum.
3138 (json::value): New forward decl.
3139 (diagnostic_context::path_format): New field.
3140 (diagnostic_context::show_path_depths): New field.
3141 (diagnostic_context::print_path): New callback field.
3142 (diagnostic_context::make_json_for_path): New callback field.
3143 (diagnostic_show_any_path): New decl.
3144 (json_from_expanded_location): New decl.
3145 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3146 (-fdiagnostics-show-path-depths): New option.
3147 (-fdiagnostics-color): Add "path" to description of default
3148 GCC_COLORS; describe it.
3149 (-fdiagnostics-format=json): Document how diagnostic paths are
3150 represented in the JSON output format.
3151 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3152 Add optional params "restrict_to_current_line_spans" and "label".
3153 * opts.c (common_handle_option): Handle
3154 OPT_fdiagnostics_path_format_ and
3155 OPT_fdiagnostics_show_path_depths.
3156 * pretty-print.c: Include "diagnostic-event-id.h".
3157 (pp_format): Implement "%@" format code for printing
3158 diagnostic_event_id_t *.
3159 (selftest::test_pp_format): Add tests for "%@".
3160 * selftest-run-tests.c (selftest::run_tests): Call
3161 selftest::tree_diagnostic_path_cc_tests.
3162 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3163 * toplev.c (general_init): Initialize global_dc->path_format and
3164 global_dc->show_path_depths.
3165 * tree-diagnostic-path.cc: New file.
3166 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3167 non-static. Drop "diagnostic" param in favor of storing the
3168 original value of "where" and re-using it.
3169 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3170 maybe_unwind_expanded_macro_loc.
3171 (tree_diagnostics_defaults): Initialize context->print_path and
3172 context->make_json_for_path.
3173 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3175 (default_tree_make_json_for_path): New decl.
3176 (maybe_unwind_expanded_macro_loc): New decl.
3178 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3180 PR tree-optimization/93210
3181 * fold-const.h (native_encode_initializer,
3182 can_native_interpret_type_p): Declare.
3183 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3185 (native_encode_initializer): New function, moved from dwarf2out.c.
3186 Adjust to native_encode_expr compatible arguments, including dry-run
3187 and partial extraction modes. Don't handle STRING_CST.
3188 (can_native_interpret_type_p): No longer static.
3189 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3190 offset / BITS_PER_UNIT fits into int and don't call it if
3191 can_native_interpret_type_p fails. If suboff is NULL and for
3192 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3193 native_encode_initializer.
3194 (fold_const_aggregate_ref_1): Formatting fix.
3195 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3196 (tree_add_const_value_attribute): Adjust caller.
3198 PR tree-optimization/90838
3199 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3200 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3201 CTZ_DEFINED_VALUE_AT_ZERO.
3203 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3206 * lra-constraints.c (match_reload): Permit input operands have the
3207 same mode as output while other input operands have a different
3210 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3212 PR tree-optimization/90838
3213 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3214 (check_ctz_string): Likewise.
3215 (optimize_count_trailing_zeroes): Likewise.
3216 (simplify_count_trailing_zeroes): Likewise.
3217 (pass_forwprop::execute): Try ctz simplification.
3218 * match.pd: Add matching for ctz idioms.
3220 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3222 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3224 (aarch64_invalid_unary_op): New function for target hook.
3225 (aarch64_invalid_binary_op): New function for target hook.
3227 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3229 * config.gcc: Add arm_bf16.h.
3230 * config/aarch64/aarch64-builtins.c
3231 (aarch64_simd_builtin_std_type): Add BFmode.
3232 (aarch64_init_simd_builtin_types): Define element types for vector
3234 (aarch64_init_bf16_types): New function.
3235 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3236 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3238 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3239 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3241 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3242 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3243 * config/aarch64/aarch64.c
3244 (aarch64_classify_vector_mode): Add support for BF types.
3245 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3246 (aarch64_vq_mode): Add support for BF types.
3247 (aarch64_simd_container_mode): Add support for BF types.
3248 (aarch64_mangle_type): Add support for BF scalar type.
3249 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3250 * config/aarch64/arm_bf16.h: New file.
3251 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3252 * config/aarch64/iterators.md: Add BF types to mode attributes.
3253 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3255 2020-01-10 Jason Merrill <jason@redhat.com>
3257 PR c++/93173 - incorrect tree sharing.
3258 * gimplify.c (copy_if_shared): No longer static.
3259 * gimplify.h: Declare it.
3261 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3263 * doc/invoke.texi (-msve-vector-bits=): Document that
3264 -msve-vector-bits=128 now generates VL-specific code for
3265 little-endian targets.
3266 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3267 build_vector_type_for_mode to construct the data vector types.
3268 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3269 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3270 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3271 for 128-bit vectors.
3273 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3275 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3278 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3280 * config/aarch64/aarch64-builtins.c
3281 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3282 rather than checking the number of elements and the element mode.
3284 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3286 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3287 get_related_vectype_for_scalar_type rather than build_vector_type
3288 to create the index type for a conditional reduction.
3290 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3292 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3293 for any type of gather or scatter, including strided accesses.
3295 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3297 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3300 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3302 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3304 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3305 parameter and its use to reset DR_OFFSET's.
3306 (vect_transform_loop): Remove orig_drs_init argument.
3307 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3308 member of dr_vec_info rather than the offset of the associated
3309 data_reference's innermost_loop_behavior.
3310 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3311 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3312 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3313 get_dr_vinfo_offset.
3314 (vectorizable_store): Likewise.
3315 (vectorizable_load): Likewise.
3317 2020-01-10 Richard Biener <rguenther@suse.de>
3319 * gimple-ssa-store-merging
3320 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3322 2020-01-10 Martin Liska <mliska@suse.cz>
3325 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3326 encapsulation that was there before r280040.
3328 2020-01-10 Richard Biener <rguenther@suse.de>
3331 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3332 sequences to avoid walking them again for secondary opportunities.
3333 (pass_lower_eh_dispatch::execute): Instead actually insert
3336 2020-01-10 Richard Biener <rguenther@suse.de>
3339 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3340 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3341 avoid quadraticness.
3343 2020-01-10 Martin Jambor <mjambor@suse.cz>
3345 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3346 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3347 to get param_ipa_sra_max_replacements.
3348 (param_splitting_across_edge): Pass the caller to
3349 pull_accesses_from_callee.
3351 2020-01-10 Martin Jambor <mjambor@suse.cz>
3353 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3354 * ipa-cp.c (max_new_size): Removed.
3355 (orig_overall_size): New variable.
3356 (get_max_overall_size): New function.
3357 (estimate_local_effects): Use it. Adjust dump.
3358 (decide_about_value): Likewise.
3359 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3360 orig_overall_size. Adjust dump.
3361 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3363 2020-01-10 Martin Jambor <mjambor@suse.cz>
3365 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3366 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3367 instead of param_ipa_max_agg_items.
3368 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3369 optimization info for the callee.
3371 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3373 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3374 markers if debug_inline_points is false.
3376 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3378 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3380 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3381 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3382 aarch64-sve-builtins-sve2.h.
3383 (aarch64-sve-builtins-sve2.o): New rule.
3384 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3385 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3386 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3387 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3388 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3389 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3391 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3392 instructions that are handled here.
3393 (@cond_asrd<mode>): Generalize to...
3394 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3395 (*cond_asrd<mode>_2): Generalize to...
3396 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3397 (*cond_asrd<mode>_z): Generalize to...
3398 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3399 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3400 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3401 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3402 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3404 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3405 (@aarch64_scatter_stnt<mode>): Likewise.
3406 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3407 (@aarch64_mul_lane_<mode>): Likewise.
3408 (@aarch64_sve_suqadd<mode>_const): Likewise.
3409 (*<sur>h<addsub><mode>): Generalize to...
3410 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3412 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3413 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3414 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3415 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3416 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3417 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3418 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3419 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3420 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3421 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3422 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3423 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3424 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3425 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3426 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3427 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3428 (@aarch64_sve2_xar<mode>): Likewise.
3429 (@aarch64_sve2_bcax<mode>): Likewise.
3430 (*aarch64_sve2_eor3<mode>): Rename to...
3431 (@aarch64_sve2_eor3<mode>): ...this.
3432 (@aarch64_sve2_bsl<mode>): New expander.
3433 (@aarch64_sve2_nbsl<mode>): Likewise.
3434 (@aarch64_sve2_bsl1n<mode>): Likewise.
3435 (@aarch64_sve2_bsl2n<mode>): Likewise.
3436 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3437 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3438 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3439 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3440 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3441 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3442 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3443 (<su>mull<bt><Vwide>): Generalize to...
3444 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3446 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3447 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3448 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3449 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3450 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3451 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3452 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3453 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3454 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3455 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3456 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3457 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3458 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3459 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3460 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3461 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3462 (<SHRNB:r>shrnb<mode>): Generalize to...
3463 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3465 (<SHRNT:r>shrnt<mode>): Generalize to...
3466 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3468 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3469 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3470 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3471 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3472 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3473 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3474 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3475 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3476 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3477 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3478 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3479 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3480 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3481 (@aarch64_sve2_cvtnt<mode>): Likewise.
3482 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3483 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3484 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3485 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3486 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3487 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3488 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3489 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3490 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3491 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3492 (@aarch64_sve2_pmul<mode>): Likewise.
3493 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3494 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3495 (@aarch64_sve2_tbl2<mode>): Likewise.
3496 (@aarch64_sve2_tbx<mode>): Likewise.
3497 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3498 (@aarch64_sve2_histcnt<mode>): Likewise.
3499 (@aarch64_sve2_histseg<mode>): Likewise.
3500 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3501 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3502 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3503 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3504 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3505 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3506 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3507 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3508 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3509 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3510 (SVE2_PMULL_PAIR_I): New mode iterators.
3511 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3512 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3513 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3514 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3515 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3516 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3517 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3518 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3519 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3520 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3521 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3522 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3523 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3524 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3525 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3526 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3527 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3528 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3529 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3530 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3531 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3532 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3533 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3534 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3535 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3536 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3537 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3538 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3539 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3540 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3541 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3543 (VNARROW, Ventype): New mode attributes.
3544 (Vewtype): Handle VNx2DI. Fix typo in comment.
3545 (VDOUBLE): New mode attribute.
3546 (sve_lane_con): Handle VNx8HI.
3547 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3548 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3549 (sve_int_op, sve_int_op_rev): Handle the above codes.
3550 (sve_pred_int_rhs2_operand): Likewise.
3551 (MULLBT, SHRNB, SHRNT): Delete.
3552 (SVE_INT_SHIFT_IMM): New int iterator.
3553 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3554 and UNSPEC_WHILEHS for TARGET_SVE2.
3555 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3556 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3557 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3558 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3559 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3560 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3561 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3562 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3563 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3564 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3565 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3566 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3567 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3568 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3569 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3570 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3571 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3572 (optab): Handle the new unspecs.
3573 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3575 (lr): Handle the new unspecs.
3577 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3578 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3579 (sve_int_qsub_op): New int attributes.
3580 (sve_fp_op, rot): Handle the new unspecs.
3581 * config/aarch64/aarch64-sve-builtins.h
3582 (function_resolver::require_matching_pointer_type): Declare.
3583 (function_resolver::resolve_unary): Add an optional boolean argument.
3584 (function_resolver::finish_opt_n_resolution): Add an optional
3585 type_suffix_index argument.
3586 (gimple_folder::redirect_call): Declare.
3587 (gimple_expander::prepare_gather_address_operands): Add an optional
3589 * config/aarch64/aarch64-sve-builtins.cc: Include
3590 aarch64-sve-builtins-sve2.h.
3591 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3592 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3593 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3594 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3595 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3596 (TYPES_s_unsigned): Likewise.
3597 (TYPES_s_integer): Use TYPES_s_unsigned.
3598 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3599 (TYPES_sd_integer): Use them.
3600 (TYPES_d_unsigned): New macro.
3601 (TYPES_d_integer): Use it.
3602 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3603 (TYPES_cvt_narrow): Likewise.
3604 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3605 (preds_mx): New variable.
3606 (function_builder::add_overloaded_function): Allow the new feature
3607 set to be more restrictive than the original one.
3608 (function_resolver::infer_pointer_type): Remove qualifiers from
3609 the pointer type before printing it.
3610 (function_resolver::require_matching_pointer_type): New function.
3611 (function_resolver::resolve_sv_displacement): Handle functions
3612 that don't support 32-bit vector indices or svint32_t vector offsets.
3613 (function_resolver::finish_opt_n_resolution): Take the inferred type
3614 as a separate argument.
3615 (function_resolver::resolve_unary): Optionally treat all forms in
3616 the same way as normal merging functions.
3617 (gimple_folder::redirect_call): New function.
3618 (function_expander::prepare_gather_address_operands): Add an argument
3619 that says whether scaled forms are available. If they aren't,
3620 handle scaling of vector indices and don't add the extension and
3622 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3623 fall back to using cond_* instead.
3624 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3625 Split out the member variables into...
3626 (rtx_code_function_base): ...this new base class.
3627 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3628 (unspec_based_function): Split out the member variables into...
3629 (unspec_based_function_base): ...this new base class.
3630 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3631 (unspec_based_function_exact_insn): New class.
3632 (unspec_based_add_function, unspec_based_add_lane_function)
3633 (unspec_based_lane_function, unspec_based_pred_function)
3634 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3635 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3636 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3638 (unspec_based_fused_function): New class.
3639 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3640 (unspec_based_fused_lane_function): New class.
3641 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3643 (CODE_FOR_MODE1): New macro.
3644 (fixed_insn_function): New class.
3645 (while_comparison): Likewise.
3646 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3647 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3648 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3649 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3650 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3651 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3652 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3653 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3654 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3655 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3656 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3657 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3658 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3659 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3660 Also add an initial argument for unary_convert_narrowt, regardless
3661 of the predication type.
3662 (build_32_64): Allow loads and stores to specify MODE_none.
3663 (build_sv_index64, build_sv_uint_offset): New functions.
3664 (long_type_suffix): New function.
3665 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3666 (binary_imm_long_base, load_gather_sv_base): Likewise.
3667 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3668 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3669 (unary_narrowb_base, unary_narrowt_base): Likewise.
3670 (binary_long_lane_def, binary_long_lane): New shape.
3671 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3672 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3673 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3674 (binary_to_uint_def, binary_to_uint): Likewise.
3675 (binary_wide_def, binary_wide): Likewise.
3676 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3677 (compare_def, compare): Likewise.
3678 (compare_ptr_def, compare_ptr): Likewise.
3679 (load_ext_gather_index_restricted_def,
3680 load_ext_gather_index_restricted): Likewise.
3681 (load_ext_gather_offset_restricted_def,
3682 load_ext_gather_offset_restricted): Likewise.
3683 (load_gather_sv_def): Inherit from load_gather_sv_base.
3684 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3685 (shift_left_imm_def, shift_left_imm): Likewise.
3686 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3687 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3688 (store_scatter_index_restricted_def,
3689 store_scatter_index_restricted): Likewise.
3690 (store_scatter_offset_restricted_def,
3691 store_scatter_offset_restricted): Likewise.
3692 (tbl_tuple_def, tbl_tuple): Likewise.
3693 (ternary_long_lane_def, ternary_long_lane): Likewise.
3694 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3695 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3696 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3697 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3698 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3699 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3700 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3701 (ternary_uint_def, ternary_uint): Likewise.
3702 (unary_convert): Fix typo in comment.
3703 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3704 (unary_long_def, unary_long): Likewise.
3705 (unary_narrowb_def, unary_narrowb): Likewise.
3706 (unary_narrowt_def, unary_narrowt): Likewise.
3707 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3708 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3709 (unary_to_int_def, unary_to_int): Likewise.
3710 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3711 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3712 (svasrd_impl): Delete.
3713 (svcadd_impl::expand): Handle integer operations too.
3714 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3715 new functions to derive the unspec numbers.
3716 (svmla_svmls_lane_impl): Replace with...
3717 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3718 integer operations too.
3719 (svwhile_impl): Rename to...
3720 (svwhilelx_impl): ...this and inherit from while_comparison.
3721 (svasrd): Use unspec_based_function.
3722 (svmla_lane): Use svmla_lane_impl.
3723 (svmls_lane): Use svmls_lane_impl.
3724 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3725 (svwhilele, svwhilelt): Use svwhilelx_impl.
3726 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3727 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3728 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3729 * config/aarch64/aarch64-sve-builtins.def: Include
3730 aarch64-sve-builtins-sve2.def.
3732 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3734 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3735 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3736 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3737 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3738 immediates as well as vector ones.
3739 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3740 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3741 (aarch64_sve_qsub_immediate): Update calls accordingly.
3743 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3745 * config/aarch64/aarch64-sve2.md: Add banner comments.
3746 (<su>mulh<r>s<mode>3): Move further up file.
3747 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3748 (*aarch64_sve2_sra<mode>): Move further down file.
3749 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3751 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3753 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3755 (while_optab_cmp): Handle them.
3756 * config/aarch64/aarch64-sve.md
3757 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3758 and add a "@" marker.
3759 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3760 instead of gen_aarch64_sve2_while_ptest.
3761 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3763 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3765 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3766 (UNSPEC_WHILELE): ...this.
3767 (UNSPEC_WHILE_LO): Rename to...
3768 (UNSPEC_WHILELO): ...this.
3769 (UNSPEC_WHILE_LS): Rename to...
3770 (UNSPEC_WHILELS): ...this.
3771 (UNSPEC_WHILE_LT): Rename to...
3772 (UNSPEC_WHILELT): ...this.
3773 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3774 (cmp_op, while_optab_cmp): Likewise.
3775 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3776 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3777 (svwhilelt): Likewise.
3779 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3781 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3782 (unary_to_uint): Define.
3783 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3784 (unary_count): Rename to...
3785 (unary_to_uint_def, unary_to_uint): ...this.
3786 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3788 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3790 * config/aarch64/aarch64-sve-builtins-functions.h
3791 (code_for_mode_function): New class.
3792 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3793 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3794 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3795 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3796 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3798 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3800 * config/aarch64/iterators.md (addsub): New code attribute.
3801 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3803 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3804 in the asm string and attributes. Fix indentation.
3805 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3807 (@aarch64_sve_<optab><mode>): ...this.
3808 * config/aarch64/aarch64-sve-builtins.h
3809 (function_expander::expand_signed_unpred_op): Delete.
3810 * config/aarch64/aarch64-sve-builtins.cc
3811 (function_expander::expand_signed_unpred_op): Likewise.
3812 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3813 try using code_for_aarch64_sve instead.
3814 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3815 (svqsub_impl): Likewise.
3816 (svqadd, svqsub): Use rtx_code_function instead.
3818 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3820 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3821 (HADDSUB, sur, addsub): Remove them.
3823 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3825 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3826 functions the same way, rather than singling out those that
3827 aren't mapped directly to optabs.
3829 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3831 * target.def (compatible_vector_types_p): New target hook.
3832 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3833 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3834 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3835 * doc/tm.texi: Regenerate.
3836 * gimple-expr.c: Include target.h.
3837 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3838 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3840 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3841 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3842 Use the original predicate if it already has a suitable type.
3844 2020-01-09 Martin Jambor <mjambor@suse.cz>
3846 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3847 resolve_speculation and redirect_call_stmt_to_callee static. Change
3848 return type of set_call_stmt to cgraph_edge *.
3849 * auto-profile.c (afdo_indirect_call): Adjust call to
3850 redirect_call_stmt_to_callee.
3851 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3852 make the this pointer explicit, adjust self-recursive calls and the
3853 call top make_direct. Return the resulting edge.
3854 (cgraph_edge::remove): Make this pointer explicit.
3855 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3856 (cgraph_edge::make_direct): Likewise, adjust call to
3857 resolve_speculation.
3858 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3859 call to set_call_stmt.
3860 (cgraph_update_edges_for_call_stmt_node): Update call to
3861 set_call_stmt and remove.
3862 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3863 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3864 (cgraph_node::create_edge_including_clones): Moved "first" definition
3865 of edge to the block where it was used. Adjusted calls to
3867 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3868 cgraph_edge::remove.
3869 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3870 make_direct and redirect_call_stmt_to_callee.
3871 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3872 resolve_speculation and make_direct.
3873 * ipa-inline-transform.c (inline_transform): Adjust call to
3874 redirect_call_stmt_to_callee.
3875 (check_speculations_1):: Adjust call to resolve_speculation.
3876 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3877 resolve-speculation.
3878 (inline_small_functions): Adjust call to resolve_speculation.
3879 (ipa_inline): Likewise.
3880 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3882 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3883 safe with regards to edge removal, adjust calls to
3884 redirect_call_stmt_to_callee.
3885 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3886 and redirect_call_stmt_to_callee.
3887 * multiple_target.c (create_dispatcher_calls): Adjust call to
3888 redirect_call_stmt_to_callee
3889 (redirect_to_specific_clone): Likewise.
3890 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3891 Adjust calls to cgraph_edge::remove.
3892 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3893 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3894 (expand_call_inline): Adjust call to cgraph_edge::remove.
3896 2020-01-09 Martin Liska <mliska@suse.cz>
3898 * params.opt: Set Optimization for
3899 param_max_speculative_devirt_maydefs.
3901 2020-01-09 Martin Sebor <msebor@redhat.com>
3905 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3907 2020-01-09 Martin Liska <mliska@suse.cz>
3909 * auto-profile.c (auto_profile): Use opt_for_fn
3911 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3912 (propagate_vals_across_arith_jfunc): Likewise.
3913 (hint_time_bonus): Likewise.
3914 (incorporate_penalties): Likewise.
3915 (good_cloning_opportunity_p): Likewise.
3916 (perform_estimation_of_a_value): Likewise.
3917 (estimate_local_effects): Likewise.
3918 (ipcp_propagate_stage): Likewise.
3919 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3920 (set_switch_stmt_execution_predicate): Likewise.
3921 (analyze_function_body): Likewise.
3922 * ipa-inline-analysis.c (offline_size): Likewise.
3923 * ipa-inline.c (early_inliner): Likewise.
3924 * ipa-prop.c (ipa_analyze_node): Likewise.
3925 (ipcp_transform_function): Likewise.
3926 * ipa-sra.c (process_scan_results): Likewise.
3927 (ipa_sra_summarize_function): Likewise.
3928 * params.opt: Rename ipcp-unit-growth to
3929 ipa-cp-unit-growth. Add Optimization for various
3930 IPA-related parameters.
3932 2020-01-09 Richard Biener <rguenther@suse.de>
3935 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3937 2020-01-09 Richard Biener <rguenther@suse.de>
3939 PR tree-optimization/93040
3940 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3942 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3944 * common/config/avr/avr-common.c (avr_option_optimization_table)
3945 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3947 2020-01-09 Martin Liska <mliska@suse.cz>
3949 * cgraphclones.c (symbol_table::materialize_all_clones):
3950 Use cgraph_node::dump_name.
3952 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3955 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3956 output_operand_lossage instead of gcc_unreachable.
3957 * doc/md.texi (riscv f constraint): Fix typo.
3960 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3961 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3962 CONST_SCALAR_INT_P instead of CONST_INT_P.
3963 (*subv<mode>4_1): Rename to ...
3964 (subv<mode>4_1): ... this.
3965 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3966 define_insn_and_split patterns.
3967 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3970 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3972 * vec.c (class selftest::count_dtor): New class.
3973 (selftest::test_auto_delete_vec): New test.
3974 (selftest::vec_c_tests): Call it.
3975 * vec.h (class auto_delete_vec): New class template.
3976 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3978 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3980 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3982 2020-01-08 Jim Wilson <jimw@sifive.com>
3984 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3985 use of TLS_MODEL_LOCAL_EXEC when not pic.
3987 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3989 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3992 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3995 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3996 *stack_protect_set_3 peephole2): Also check that the second
3997 insns source is general_operand.
4000 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4001 predicate for output operand instead of register_operand.
4002 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4003 memory destination and non-memory operands[2].
4005 2020-01-08 Martin Liska <mliska@suse.cz>
4007 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4008 ::dump_asm_name instead of (::name or ::asm_name).
4009 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4010 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4011 (analyze_functions): Likewise.
4012 (expand_all_functions): Likewise.
4013 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4014 (propagate_bits_across_jump_function): Likewise.
4015 (dump_profile_updates): Likewise.
4016 (ipcp_store_bits_results): Likewise.
4017 (ipcp_store_vr_results): Likewise.
4018 * ipa-devirt.c (dump_targets): Likewise.
4019 * ipa-fnsummary.c (analyze_function_body): Likewise.
4020 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4021 (process_hsa_functions): Likewise.
4022 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4023 (set_alias_uids): Likewise.
4024 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4025 * ipa-inline.c (recursive_inlining): Likewise.
4026 (inline_to_all_callers_1): Likewise.
4027 (ipa_inline): Likewise.
4028 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4029 (ipa_propagate_frequency): Likewise.
4030 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4031 (remove_described_reference): Likewise.
4032 * ipa-pure-const.c (worse_state): Likewise.
4033 (check_retval_uses): Likewise.
4034 (analyze_function): Likewise.
4035 (propagate_pure_const): Likewise.
4036 (propagate_nothrow): Likewise.
4037 (dump_malloc_lattice): Likewise.
4038 (propagate_malloc): Likewise.
4039 (pass_local_pure_const::execute): Likewise.
4040 * ipa-visibility.c (optimize_weakref): Likewise.
4041 (function_and_variable_visibility): Likewise.
4042 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4043 (ipa_discover_variable_flags): Likewise.
4044 * lto-streamer-out.c (output_function): Likewise.
4045 (output_constructor): Likewise.
4046 * tree-inline.c (copy_bb): Likewise.
4047 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4048 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4050 2020-01-08 Richard Biener <rguenther@suse.de>
4053 * tree-eh.c (sink_clobbers): Update virtual operands for
4054 the first and last stmt only. Add a dry-run capability.
4055 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4056 after CFG manipulations and in RPO order to catch all
4057 secondary opportunities reliably.
4059 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4062 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4064 2019-01-08 Richard Biener <rguenther@suse.de>
4067 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4068 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4069 virtual operand, also updating SSA use.
4070 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4071 Update stmt after resetting virtual operand.
4072 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4073 * gimple-iterator.c (gsi_remove): When not removing the stmt
4074 permanently do not delink immediate uses or mark the stmt modified.
4076 2020-01-08 Martin Liska <mliska@suse.cz>
4078 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4079 (ipa_call_context::estimate_size_and_time): Likewise.
4080 (inline_analyze_function): Likewise.
4082 2020-01-08 Martin Liska <mliska@suse.cz>
4084 * cgraph.c (cgraph_node::dump): Use systematically
4087 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4089 Add -nodevicespecs option for avr.
4092 * config/avr/avr.opt (-nodevicespecs): New driver option.
4093 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4094 "-specs=device-specs/..." if that option is not set.
4095 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4097 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4099 Implement 64-bit double functions for avr.
4102 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4103 --with-double-comparison.
4104 * doc/install.texi: Document them.
4105 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4106 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4107 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4108 * doc/invoke.texi (AVR Built-in Macros): Document them.
4109 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4110 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4111 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4113 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4116 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4117 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4118 when only building rm-profile multilibs.
4120 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4123 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4124 lattice for a value to check.
4125 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4126 finite propagation in self-recursive scc.
4128 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4130 * ipa-inline.c (caller_growth_limits): Restore the AND.
4132 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4134 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4135 (VEC_ALLREG_ALT): New iterator.
4136 (VEC_ALLREG_INT_MODE): New iterator.
4137 (VCMP_MODE): New iterator.
4138 (VCMP_MODE_INT): New iterator.
4139 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4140 (vec_cmp<u>v64qidi): New define_expand.
4141 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4142 (vec_cmpu<mode>di_exec): New define_expand.
4143 (vec_cmp<u>v64qidi_exec): New define_expand.
4144 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4145 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4146 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4147 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4148 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4149 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4150 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4151 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4152 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4153 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4155 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4156 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4158 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4160 * config/gcn/constraints.md (DA): Update description and match.
4162 (Db): New constraint.
4163 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4165 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4166 Implement 'Db' mixed immediate type.
4167 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4168 (addcv64si3_dup<exec_vcc>): Delete.
4169 (subcv64si3<exec_vcc>): Rework constraints.
4170 (addv64di3): Rework constraints.
4171 (addv64di3_exec): Rework constraints.
4172 (subv64di3): Rework constraints.
4173 (addv64di3_dup): Delete.
4174 (addv64di3_dup_exec): Delete.
4175 (addv64di3_zext): Rework constraints.
4176 (addv64di3_zext_exec): Rework constraints.
4177 (addv64di3_zext_dup): Rework constraints.
4178 (addv64di3_zext_dup_exec): Rework constraints.
4179 (addv64di3_zext_dup2): Rework constraints.
4180 (addv64di3_zext_dup2_exec): Rework constraints.
4181 (addv64di3_sext_dup2): Rework constraints.
4182 (addv64di3_sext_dup2_exec): Rework constraints.
4184 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4186 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4187 existing target checks.
4189 2020-01-07 Richard Biener <rguenther@suse.de>
4191 * doc/install.texi: Bump minimal supported MPC version.
4193 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4195 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4196 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4197 * langhooks.c: Include stor-layout.h.
4198 (lhd_simulate_enum_decl): New function.
4199 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4200 handle_arm_sve_h for the LTO frontend.
4201 (register_vector_type): Cope with null returns from pushdecl.
4203 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4205 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4206 (aarch64_sve::nvectors_if_data_type): Replace with...
4207 (aarch64_sve::builtin_type_p): ...this.
4208 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4209 (find_vector_type): Delete.
4210 (add_sve_type_attribute): New function.
4211 (lookup_sve_type_attribute): Likewise.
4212 (register_builtin_types): Add an "SVE type" attribute to each type.
4213 (register_tuple_type): Likewise.
4214 (svbool_type_p, nvectors_if_data_type): Delete.
4215 (mangle_builtin_type): Use lookup_sve_type_attribute.
4216 (builtin_type_p): Likewise. Add an overload that returns the
4217 number of constituent vector and predicate registers.
4218 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4219 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4220 instead of aarch64_sve_argument_p.
4221 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4222 (aarch64_pass_by_reference): Likewise.
4223 (aarch64_function_value_1): Likewise.
4224 (aarch64_return_in_memory): Likewise.
4225 (aarch64_layout_arg): Likewise.
4227 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4229 PR tree-optimization/93156
4230 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4231 least significant bit is always clear.
4233 PR tree-optimization/93118
4234 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4235 simplifier with two intermediate conversions.
4237 2020-01-07 Martin Liska <mliska@suse.cz>
4239 * params.opt: Add Optimization for various parameters.
4241 2020-01-07 Martin Liska <mliska@suse.cz>
4244 * doc/extend.texi: Explain cloning for target_clone
4247 2020-01-07 Martin Liska <mliska@suse.cz>
4249 PR tree-optimization/92860
4250 * common.opt: Make in Optimization option
4251 as it is affected by -O0, which is an Optimization
4253 * tree-inline.c (tree_inlinable_function_p):
4254 Use opt_for_fn for warn_inline.
4255 (expand_call_inline): Likewise.
4257 2020-01-07 Martin Liska <mliska@suse.cz>
4259 PR tree-optimization/92860
4260 * common.opt: Make flag_ree as optimization
4263 2020-01-07 Martin Liska <mliska@suse.cz>
4265 PR optimization/92860
4266 * params.opt: Mark param_min_crossjump_insns with Optimization
4269 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4271 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4272 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4274 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4276 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4277 helper function to return the valid addressing formats for a given
4278 hard register and mode.
4279 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4281 * config/rs6000/constraints.md (Q constraint): Update
4283 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4286 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4287 Use 'Q' for doing vector extract from memory.
4288 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4290 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4291 doing vector extract from memory.
4292 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4293 extract from memory.
4295 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4296 for the offset being 34-bits when -mcpu=future is used.
4298 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4300 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4301 instead of cmpib_comparison_operator in cmpib patterns.
4302 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4303 of cmpib_comparison_operator. Revise comment.
4305 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4307 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4308 in an IFN_DIV_POW2 node to be equal.
4310 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4312 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4313 (vect_check_scalar_mask): ...this.
4314 (vectorizable_store, vectorizable_load): Update call accordingly.
4315 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4316 argument in calls to conditional internal functions.
4318 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4320 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4321 '0' matching inputs.
4322 (subv64di3_exec): Likewise.
4324 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4326 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4327 * doc/md.texi (movstr): Likewise.
4329 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4331 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4334 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4336 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4338 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4339 to a temporary file and use move-if-change to update the real
4340 file where necessary.
4342 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4344 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4345 rather than Upa for CPY /M.
4347 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4349 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4352 2020-01-06 Martin Liska <mliska@suse.cz>
4354 PR tree-optimization/92860
4355 * params.opt: Mark param_max_combine_insns with Optimization
4358 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4361 * config/i386/i386.md (SWIDWI): New mode iterator.
4362 (DWI, dwi): Add TImode variants.
4363 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4364 <general_hilo_operand> instead of <general_operand>. Use
4365 CONST_SCALAR_INT_P instead of CONST_INT_P.
4366 (*addv<mode>4_1): Rename to ...
4367 (addv<mode>4_1): ... this.
4368 (QWI): New mode attribute.
4369 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4370 define_insn_and_split patterns.
4371 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4373 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4374 <general_hilo_operand> instead of <general_operand>.
4375 (*addcarry<mode>_1): New define_insn.
4376 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4378 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4380 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4381 Use "call" instead of "set".
4383 2020-01-03 Martin Jambor <mjambor@suse.cz>
4386 * ipa-cp.c (print_all_lattices): Skip functions without info.
4388 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4391 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4392 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4393 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4394 for 'e' simd clones.
4397 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4399 (mprefer-vector-width=): Add Save.
4400 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4401 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4402 (ix86_debug_options, ix86_function_specific_print): Adjust
4403 ix86_target_string callers.
4404 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4405 (ix86_valid_target_attribute_tree): Likewise.
4406 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4407 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4408 ix86_target_string caller.
4411 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4412 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4413 instead of gen_int_shift_amount + convert_modes.
4415 PR rtl-optimization/93088
4416 * loop-iv.c (find_single_def_src): Punt after looking through
4417 128 reg copies for regs with single definitions. Move definitions
4420 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4422 * config/arm/arm-c.c (arm_cpu_builtins): Define
4423 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4424 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4425 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4426 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4427 * config/arm/arm-tables.opt: Regenerated.
4428 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4429 arm_arch_i8mm and arm_arch_bf16 when enabled.
4430 * config/arm/arm.h (TARGET_I8MM): New macro.
4431 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4432 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4433 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4434 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4435 (v8_6_a_simd_variants): New.
4436 (v8_*_a_simd_variants): Add i8mm and bf16.
4437 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4439 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4442 * predict.c (compute_function_frequency): Don't call
4443 warn_function_cold on functions that already have cold attribute.
4445 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4448 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4449 COMDAT group function labels in .data.rel.ro.local section.
4450 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4453 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4454 comparison_operator in B and S integer comparisons. Likewise, use
4455 ordered_comparison_operator instead of cmpib_comparison_operator in
4457 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4459 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4461 Update copyright years.
4463 * gcc.c (process_command): Update copyright notice dates.
4464 * gcov-dump.c (print_version): Ditto.
4465 * gcov.c (print_version): Ditto.
4466 * gcov-tool.c (print_version): Ditto.
4467 * gengtype.c (create_file): Ditto.
4468 * doc/cpp.texi: Bump @copying's copyright year.
4469 * doc/cppinternals.texi: Ditto.
4470 * doc/gcc.texi: Ditto.
4471 * doc/gccint.texi: Ditto.
4472 * doc/gcov.texi: Ditto.
4473 * doc/install.texi: Ditto.
4474 * doc/invoke.texi: Ditto.
4476 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4478 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4481 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4483 PR tree-optimization/93098
4484 * match.pd (popcount): For shift amounts, use integer_onep
4485 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4486 tests. Make sure that precision is power of two larger than or equal
4487 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4488 instead of ULL suffixed constants. Formatting fixes.
4490 Copyright (C) 2020 Free Software Foundation, Inc.
4492 Copying and distribution of this file, with or without modification,
4493 are permitted in any medium without royalty provided the copyright
4494 notice and this notice are preserved.