1 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
3 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
4 double-word size when handling aggregate return values.
5 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
6 that homogeneous SFmode and DFmode aggregates are passed and returned
9 2020-02-21 Jakub Jelinek <jakub@redhat.com>
12 * opts.c (print_filtered_help): Translate help before appending
13 messages to it rather than after that.
15 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
17 PR rtl-optimization/PR92989
18 * lra-lives.c (process_bb_lives): Restore the original order
19 of the bb liveness update. Call make_hard_regno_dead for each
20 register clobbered at the start of an EH receiver.
22 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
25 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
26 self-recursively generated.
28 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
31 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
34 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
36 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
37 Document new target supports option.
39 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
41 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
42 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
43 * config/arm/iterators.md (MATMUL): New iterator.
44 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
45 (mmla_sfx): New attribute.
46 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
47 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
48 (UNSPEC_MATMUL_US): New.
50 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
52 * config/arm/arm.md: Prevent scalar shifts from being used when big
55 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
56 Richard Biener <rguenther@suse.de>
58 PR tree-optimization/93586
59 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
60 after mismatched array refs; do not sure type size information to
61 recover from unmatched referneces with !flag_strict_aliasing_p.
63 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
65 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
66 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
67 (scatter_store<mode>): Rename to ...
68 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
69 (scatter<mode>_exec): Delete. Move contents ...
70 (mask_scatter_store<mode>): ... here, and rename that to ...
71 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
72 Remove mode conversion.
73 (mask_gather_load<mode>): Rename to ...
74 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
75 Remove mode conversion.
76 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
78 2020-02-21 Martin Jambor <mjambor@suse.cz>
80 PR tree-optimization/93845
81 * tree-sra.c (verify_sra_access_forest): Only test access size of
84 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
86 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
87 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
88 (addv64di3_exec): Likewise.
89 (subv64di3): Likewise.
90 (subv64di3_exec): Likewise.
91 (addv64di3_zext): Likewise.
92 (addv64di3_zext_exec): Likewise.
93 (addv64di3_zext_dup): Likewise.
94 (addv64di3_zext_dup_exec): Likewise.
95 (addv64di3_zext_dup2): Likewise.
96 (addv64di3_zext_dup2_exec): Likewise.
97 (addv64di3_sext_dup2): Likewise.
98 (addv64di3_sext_dup2_exec): Likewise.
99 (<expander>v64di3): Likewise.
100 (<expander>v64di3_exec): Likewise.
101 (*<reduc_op>_dpp_shr_v64di): Likewise.
102 (*plus_carry_dpp_shr_v64di): Likewise.
103 * config/gcn/gcn.md (adddi3): Likewise.
104 (addptrdi3): Likewise.
105 (<expander>di3): Likewise.
107 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
109 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
111 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
113 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
114 support. Use aarch64_emit_mult instead of emitting multiplication
115 instructions directly.
116 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
117 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
119 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
121 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
122 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
123 instead of emitting multiplication instructions directly.
124 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
125 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
126 (@aarch64_frecps<mode>): New expanders.
128 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
130 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
131 on and produce uint64_ts rather than ints.
132 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
133 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
135 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
137 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
138 an unused xmsk register when handling approximate rsqrt.
140 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
142 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
143 flag_finite_math_only condition.
145 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
148 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
149 to destination operand for shufps alternative.
150 (*vec_extractv2si_1): Ditto.
152 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
155 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
158 2020-02-20 Martin Liska <mliska@suse.cz>
161 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
163 2020-02-20 Martin Liska <mliska@suse.cz>
166 * common/config/avr/avr-common.c: Remote trailing "|".
168 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
170 * collect2.c (maybe_run_lto_and_relink): Fix typo in
173 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
175 PR tree-optimization/93767
176 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
177 access-size bias from the offset calculations for negative strides.
179 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
181 * collect2.c (c_file, o_file): Make const again.
182 (ldout,lderrout, dump_ld_file): Remove.
183 (tool_cleanup): Avoid calling not signal-safe functions.
184 (maybe_run_lto_and_relink): Avoid possible signal handler
185 access to unintialzed memory (lto_o_files).
186 (main): Avoid leaking temp files in $TMPDIR.
187 Initialize c_file/o_file with concat, which avoids exposing
188 uninitialized memory to signal handler, which calls unlink(!).
189 Avoid calling maybe_unlink when the main function returns,
190 since the atexit handler is already doing this.
191 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
193 2020-02-19 Martin Jambor <mjambor@suse.cz>
195 PR tree-optimization/93776
196 * tree-sra.c (create_access): Do not create zero size accesses.
197 (get_access_for_expr): Do not search for zero sized accesses.
199 2020-02-19 Martin Jambor <mjambor@suse.cz>
201 PR tree-optimization/93667
202 * tree-sra.c (scalarizable_type_p): Return false if record fields
203 do not follow wach other.
205 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
207 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
208 rather than fmv.x.s/fmv.s.x.
210 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
212 * config/aarch64/aarch64-simd-builtins.def
213 (intrinsic_vec_smult_lo_): New.
214 (intrinsic_vec_umult_lo_): Likewise.
215 (vec_widen_smult_hi_): Likewise.
216 (vec_widen_umult_hi_): Likewise.
217 * config/aarch64/aarch64-simd.md
218 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
219 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
220 (vmull_high_s16): Likewise.
221 (vmull_high_s32): Likewise.
222 (vmull_high_u8): Likewise.
223 (vmull_high_u16): Likewise.
224 (vmull_high_u32): Likewise.
225 (vmull_s8): Likewise.
226 (vmull_s16): Likewise.
227 (vmull_s32): Likewise.
228 (vmull_u8): Likewise.
229 (vmull_u16): Likewise.
230 (vmull_u32): Likewise.
232 2020-02-18 Martin Liska <mliska@suse.cz>
234 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
235 bootstrap by missing removal of invalid sanity check.
237 2020-02-18 Martin Liska <mliska@suse.cz>
240 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
241 Always compare LHS of gimple_assign.
243 2020-02-18 Martin Liska <mliska@suse.cz>
246 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
247 and return type of functions.
248 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
249 Drop MALLOC attribute for void functions.
250 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
251 malloc_state for a new VOID clone.
253 2020-02-18 Martin Liska <mliska@suse.cz>
256 * common.opt: Add -fprofile-reproducibility.
257 * doc/invoke.texi: Document it.
258 * value-prof.c (dump_histogram_value):
259 Document and support behavior for counters[0]
260 being a negative value.
261 (get_nth_most_common_value): Handle negative
262 counters[0] in respect to flag_profile_reproducible.
264 2020-02-18 Jakub Jelinek <jakub@redhat.com>
267 * cgraph.c (verify_speculative_call): Use speculative_id instead of
268 speculative_uid in messages. Remove trailing whitespace from error
269 message. Use num_speculative_call_targets instead of
270 num_speculative_targets in a message.
271 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
272 edge messages and stmt instead of cal_stmt in reference message.
274 PR tree-optimization/93780
275 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
276 before calling build_vector_type.
277 (execute_update_addresses_taken): Likewise.
280 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
281 typo, functoin -> function.
282 * tree.c (free_lang_data_in_decl): Fix comment typo,
283 functoin -> function.
284 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
286 2020-02-17 David Malcolm <dmalcolm@redhat.com>
288 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
290 (print_option_information): Don't call get_option_url if URLs
293 2020-02-17 Alexandre Oliva <oliva@adacore.com>
295 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
296 handling of register_common-less targets.
298 2020-02-17 Martin Liska <mliska@suse.cz>
301 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
303 2020-02-17 Martin Liska <mliska@suse.cz>
306 * config/rs6000/rs6000.c (rs6000_option_override_internal):
309 2020-02-17 Martin Liska <mliska@suse.cz>
312 * config/rx/elf.opt: Fix typo.
314 2020-02-17 Richard Biener <rguenther@suse.de>
317 * opts-global.c (print_ignored_options): Use inform and
320 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
323 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
325 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
328 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
329 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
331 2020-02-15 Jason Merrill <jason@redhat.com>
333 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
335 2020-02-15 Jakub Jelinek <jakub@redhat.com>
337 PR tree-optimization/93744
338 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
339 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
340 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
341 sure @2 in the first and @1 in the other patterns has no side-effects.
343 2020-02-15 David Malcolm <dmalcolm@redhat.com>
344 Bernd Edlinger <bernd.edlinger@hotmail.de>
348 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
349 * configure.ac (--with-diagnostics-urls): New configuration
350 option, based on --with-diagnostics-color.
351 (DIAGNOSTICS_URLS_DEFAULT): New define.
352 * config.h: Regenerate.
353 * configure: Regenerate.
354 * diagnostic.c (diagnostic_urls_init): Handle -1 for
355 DIAGNOSTICS_URLS_DEFAULT from configure-time
356 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
357 and TERM_URLS environment variable.
358 * diagnostic-url.h (diagnostic_url_format): New enum type.
359 (diagnostic_urls_enabled_p): rename to...
360 (determine_url_format): ... this, and change return type.
361 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
362 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
363 the linux console, and mingw.
364 (diagnostic_urls_enabled_p): rename to...
365 (determine_url_format): ... this, and adjust.
366 * pretty-print.h (pretty_printer::show_urls): rename to...
367 (pretty_printer::url_format): ... this, and change to enum.
368 * pretty-print.c (pretty_printer::pretty_printer,
369 pp_begin_url, pp_end_url, test_urls): Adjust.
370 * doc/install.texi (--with-diagnostics-urls): Document the new
371 configuration option.
372 (--with-diagnostics-color): Document the existing interaction
373 with GCC_COLORS better.
374 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
375 vindex reference. Update description of defaults based on the above.
376 (-fdiagnostics-color): Update description of how -fdiagnostics-color
377 interacts with GCC_COLORS.
379 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
382 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
383 conjunction with TARGET_GNU_TLS in early return.
385 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
387 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
388 the mode is not wider than UNITS_PER_WORD.
390 2020-02-14 Martin Jambor <mjambor@suse.cz>
392 PR tree-optimization/93516
393 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
394 access of the same type as the parent.
395 (propagate_subaccesses_from_lhs): Likewise.
397 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
400 * config/i386/avx512vbmi2intrin.h
401 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
402 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
403 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
404 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
405 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
406 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
407 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
408 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
409 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
410 of lacking a closing parenthesis.
411 * config/i386/avx512vbmi2vlintrin.h
412 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
413 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
414 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
415 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
416 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
417 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
418 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
419 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
420 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
421 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
422 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
423 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
424 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
425 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
426 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
427 _mm_shldi_epi32, _mm_mask_shldi_epi32,
428 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
429 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
431 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
434 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
435 the target function entry.
437 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
439 * common/config/arc/arc-common.c (arc_option_optimization_table):
440 Disable if-conversion step when optimized for size.
442 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
444 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
445 R12-R15 are always in ARCOMPACT16_REGS register class.
446 * config/arc/arc.opt (mq-class): Deprecate.
447 * config/arc/constraint.md ("q"): Remove dependency on mq-class
449 * doc/invoke.texi (mq-class): Update text.
450 * common/config/arc/arc-common.c (arc_option_optimization_table):
453 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
455 * config/arc/arc.c (arc_insn_cost): New function.
456 (TARGET_INSN_COST): Define.
457 * config/arc/arc.md (cost): New attribute.
458 (add_n): Use arc_nonmemory_operand.
459 (ashlsi3_insn): Likewise, also update constraints.
460 (ashrsi3_insn): Likewise.
462 (add_shift): Likewise.
463 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
465 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
467 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
469 (umulsidi_600): Likewise.
471 2020-02-13 Jakub Jelinek <jakub@redhat.com>
474 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
475 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
476 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
477 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
478 pass __A to the builtin followed by __W instead of __A followed by
480 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
481 _mm512_mask_popcnt_epi64): Likewise.
482 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
483 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
484 _mm256_mask_popcnt_epi64): Likewise.
486 PR tree-optimization/93582
487 * fold-const.h (shift_bytes_in_array_left,
488 shift_bytes_in_array_right): Declare.
489 * fold-const.c (shift_bytes_in_array_left,
490 shift_bytes_in_array_right): New function, moved from
491 gimple-ssa-store-merging.c, no longer static.
492 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
493 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
494 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
495 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
496 shift_bytes_in_array.
497 (verify_shift_bytes_in_array): Rename to ...
498 (verify_shift_bytes_in_array_left): ... this. Use
499 shift_bytes_in_array_left instead of shift_bytes_in_array.
500 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
501 instead of verify_shift_bytes_in_array.
502 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
503 / native_interpret_expr where the store covers all needed bits,
504 punt on PDP-endian, otherwise allow all involved offsets and sizes
505 not to be byte-aligned.
508 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
509 use const_0_to_255_operand predicate instead of immediate_operand.
510 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
511 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
512 vgf2p8affineinvqb_<mode><mask_name>,
513 vgf2p8affineqb_<mode><mask_name>): Drop mode from
514 const_0_to_255_operand predicated operands.
516 2020-02-12 Jeff Law <law@redhat.com>
518 * config/h8300/h8300.md (comparison shortening peepholes): Use
519 a mode iterator to merge the HImode and SImode peepholes.
521 2020-02-12 Jakub Jelinek <jakub@redhat.com>
524 * real.c (is_even): Make static. Function comment fix.
525 (is_halfway_below): Make static, don't assert R is not inf/nan,
526 instead return false for those. Small formatting fixes.
528 2020-02-12 Martin Sebor <msebor@redhat.com>
531 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
532 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
533 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
534 (strlen_check_and_optimize_call): Adjust callee name.
536 2020-02-12 Jeff Law <law@redhat.com>
538 * config/h8300/h8300.md (comparison shortening peepholes): Drop
539 (and (xor)) variant. Combine other two into single peephole.
541 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
543 PR rtl-optimization/93565
544 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
546 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
548 * config/aarch64/aarch64-simd.md
549 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
550 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
551 generating separate ADDV and zero_extend patterns.
552 * config/aarch64/iterators.md (VDQV_E): New iterator.
554 2020-02-12 Jeff Law <law@redhat.com>
556 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
557 expanders, splits, etc.
558 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
559 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
560 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
561 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
562 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
563 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
565 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
567 2020-02-12 Jakub Jelinek <jakub@redhat.com>
570 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
571 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
572 TARGET_AVX512DQ from condition.
573 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
574 instead of <mask_mode512bit_condition> in condition. If
575 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
577 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
580 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
583 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
585 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
587 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
588 where strlen is more legible.
589 (rs6000_builtin_vectorized_libmass): Ditto.
590 (rs6000_print_options_internal): Ditto.
592 2020-02-11 Martin Sebor <msebor@redhat.com>
594 PR tree-optimization/93683
595 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
597 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
599 * config/rs6000/predicates.md (cint34_operand): Rename the
600 -mprefixed-addr option to be -mprefixed.
601 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
602 the -mprefixed-addr option to be -mprefixed.
603 (OTHER_FUTURE_MASKS): Likewise.
604 (POWERPC_MASKS): Likewise.
605 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
606 the -mprefixed-addr option to be -mprefixed. Change error
607 messages to refer to -mprefixed.
608 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
610 (rs6000_legitimate_offset_address_p): Likewise.
611 (rs6000_mode_dependent_address): Likewise.
612 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
613 "-mprefixed" for target attributes and pragmas.
614 (address_to_insn_form): Rename the -mprefixed-addr option to be
616 (rs6000_adjust_insn_length): Likewise.
617 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
618 -mprefixed-addr option to be -mprefixed.
619 (ASM_OUTPUT_OPCODE): Likewise.
620 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
621 -mprefixed-addr option to be -mprefixed.
622 * config/rs6000/rs6000.opt (-mprefixed): Rename the
623 -mprefixed-addr option to be prefixed. Change the option from
624 being undocumented to being documented.
625 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
626 -mprefixed option. Update the -mpcrel documentation to mention
629 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
631 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
632 including FIRST_PSEUDO_REGISTER - 1.
633 * ira-color.c (print_hard_reg_set): Ditto.
635 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
637 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
638 (USTERNOP_QUALIFIERS): New define.
639 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
640 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
641 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
642 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
643 * config/arm/arm_neon.h (vusdot_s32): New.
644 (vusdot_lane_s32): New.
645 (vusdotq_lane_s32): New.
646 (vsudot_lane_s32): New.
647 (vsudotq_lane_s32): New.
648 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
649 * config/arm/iterators.md (DOTPROD_I8MM): New.
650 (sup, opsuffix): Add <us/su>.
651 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
652 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
654 2020-02-11 Richard Biener <rguenther@suse.de>
656 PR tree-optimization/93661
657 PR tree-optimization/93662
658 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
660 * tree-sra.c (get_access_for_expr): Likewise.
662 2020-02-10 Jakub Jelinek <jakub@redhat.com>
665 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
666 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
667 Change condition from TARGET_AVX2 to TARGET_AVX.
669 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
672 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
675 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
677 Try to generate zero-based comparisons.
678 * config/cris/cris.c (cris_reduce_compare): New function.
679 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
680 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
681 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
683 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
686 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
687 in Thumb state and also as a destination in Arm state. Add T16
690 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
692 * md.texi (Define Subst): Match closing paren in example.
694 2020-02-10 Jakub Jelinek <jakub@redhat.com>
698 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
699 arguments of strncmp.
701 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
704 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
705 but different source value.
706 (adjust_callers_for_value_intersection): New function.
707 (gather_edges_for_value): Adjust order of callers to let a
708 non-self-recursive caller be the first element.
709 (self_recursive_pass_through_p): Add a new parameter "simple", and
710 check generalized self-recursive pass-through jump function.
711 (self_recursive_agg_pass_through_p): Likewise.
712 (find_more_scalar_values_for_callers_subset): Compute value from
713 pass-through jump function for self-recursive.
714 (intersect_with_plats): Cleanup previous implementation code for value
715 itersection with self-recursive call edge.
716 (intersect_with_agg_replacements): Likewise.
717 (intersect_aggregates_with_edge): Deduce value from pass-through jump
718 function for self-recursive call edge. Cleanup previous implementation
719 code for value intersection with self-recursive call edge.
720 (decide_whether_version_node): Remove dead callers and adjust order
721 to let a non-self-recursive caller be the first element.
723 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
725 * recog.c: Move pass_split_before_sched2 code in front of
726 pass_split_before_regstack.
727 (pass_data_split_before_sched2): Rename pass to split3 from split4.
728 (pass_data_split_before_regstack): Rename pass to split4 from split3.
729 (rest_of_handle_split_before_sched2): Remove.
730 (pass_split_before_sched2::execute): Unconditionally call
732 (enable_split_before_sched2): New function.
733 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
734 (pass_split_before_regstack::gate): Ditto.
735 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
736 Update name check for renamed split4 pass.
737 * config/sh/sh.c (register_sh_passes): Update pass insertion
738 point for renamed split4 pass.
740 2020-02-09 Jakub Jelinek <jakub@redhat.com>
742 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
743 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
744 copying them around between host and target.
746 2020-02-08 Andrew Pinski <apinski@marvell.com>
749 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
750 STRICT_ALIGNMENT also.
752 2020-02-08 Jim Wilson <jimw@sifive.com>
755 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
757 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
758 Jakub Jelinek <jakub@redhat.com>
761 * config/i386/i386.h (CALL_USED_REGISTERS): Make
762 xmm16-xmm31 call-used even in 64-bit ms-abi.
764 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
766 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
767 (simd_ummla, simd_usmmla): Likewise.
768 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
769 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
772 2020-02-07 Richard Biener <rguenther@suse.de>
775 * tree-inline.c (fold_marked_statements): Do a PRE walk,
776 skipping unreachable regions.
777 (optimize_inline_calls): Skip folding stmts when we didn't
780 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
783 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
784 Don't return aggregates with only SFmode and DFmode in SSE
786 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
788 2020-02-07 Jakub Jelinek <jakub@redhat.com>
791 * config/rs6000/rs6000-logue.c
792 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
793 if it fails, move rs into end_addr and retry. Add
794 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
795 the insn pattern doesn't describe well what exactly happens to
799 * config/i386/predicates.md (avx_identity_operand): Remove.
800 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
801 (avx_<castmode><avxsizesuffix>_<castmode>,
802 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
803 a VEC_CONCAT of the operand and UNSPEC_CAST.
804 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
805 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
809 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
810 recog_data.insn if distance_non_agu_define changed it.
812 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
815 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
816 we only had X-FORM (reg+reg) addressing for vectors. Also before
817 ISA 3.0, we only had X-FORM addressing for scalars in the
818 traditional Altivec registers.
820 2020-02-06 <zhongyunde@huawei.com>
821 Vladimir Makarov <vmakarov@redhat.com>
823 PR rtl-optimization/93561
824 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
827 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
829 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
832 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
834 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
835 where the low and the high 32 bits are equal to each other specially,
836 with an rldimi instruction.
838 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
840 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
842 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
844 * config/arm/arm-tables.opt: Regenerate.
846 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
849 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
850 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
851 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
853 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
855 PR rtl-optimization/87763
856 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
858 2020-02-06 Delia Burduv <delia.burduv@arm.com>
860 * config/aarch64/aarch64-simd-builtins.def
861 (bfmlaq): New built-in function.
862 (bfmlalb): New built-in function.
863 (bfmlalt): New built-in function.
864 (bfmlalb_lane): New built-in function.
865 (bfmlalt_lane): New built-in function.
866 * config/aarch64/aarch64-simd.md
867 (aarch64_bfmmlaqv4sf): New pattern.
868 (aarch64_bfmlal<bt>v4sf): New pattern.
869 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
870 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
871 (vbfmlalbq_f32): New intrinsic.
872 (vbfmlaltq_f32): New intrinsic.
873 (vbfmlalbq_lane_f32): New intrinsic.
874 (vbfmlaltq_lane_f32): New intrinsic.
875 (vbfmlalbq_laneq_f32): New intrinsic.
876 (vbfmlaltq_laneq_f32): New intrinsic.
877 * config/aarch64/iterators.md (BF_MLA): New int iterator.
878 (bt): New int attribute.
880 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
882 * config/i386/i386.md (*pushtf): Emit "#" instead of
883 calling gcc_unreachable in insn output.
886 (*pushsf_rex64): Ditto for alternatives other than 1.
887 (*pushsf): Ditto for alternatives other than 1.
889 2020-02-06 Martin Liska <mliska@suse.cz>
891 PR gcov-profile/91971
892 PR gcov-profile/93466
893 * coverage.c (coverage_init): Revert mangling of
894 path into filename. It can lead to huge filename length.
895 Creation of subfolders seem more natural.
897 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
900 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
901 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
902 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
904 2020-02-06 Jakub Jelinek <jakub@redhat.com>
907 * config/i386/predicates.md (avx_identity_operand): New predicate.
908 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
909 define_insn_and_split.
912 * omp-low.c (use_pointer_for_field): For nested constructs, also
913 look for map clauses on target construct.
914 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
915 taskreg_nesting_level.
918 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
919 shared clause, call omp_notice_variable on outer context if any.
921 2020-02-05 Jason Merrill <jason@redhat.com>
924 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
925 non-zero address even if weak and not yet defined.
927 2020-02-05 Martin Sebor <msebor@redhat.com>
929 PR tree-optimization/92765
930 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
931 * tree-ssa-strlen.c (compute_string_length): Remove.
932 (determine_min_objsize): Remove.
933 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
934 Avoid using type size as the upper bound on string length.
935 (handle_builtin_string_cmp): Add an argument. Adjust.
936 (strlen_check_and_optimize_call): Pass additional argument to
937 handle_builtin_string_cmp.
939 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
941 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
942 (*pushdi2_rex64 peephole2): Unconditionally split after
944 (*ashl<mode>3_doubleword): Ditto.
945 (*<shift_insn><mode>3_doubleword): Ditto.
947 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
950 * config/rs6000/rs6000.c (get_vector_offset): Fix
952 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
954 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
956 2020-02-05 David Malcolm <dmalcolm@redhat.com>
959 (Special Functions for Debugging the Analyzer): Update description
960 of __analyzer_dump_exploded_nodes.
962 2020-02-05 Jakub Jelinek <jakub@redhat.com>
965 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
966 include sets and not clobbers in the vzeroupper pattern.
967 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
968 the parallel has 17 (64-bit) or 9 (32-bit) elts.
969 (*avx_vzeroupper_1): New define_insn_and_split.
972 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
973 don't run when !optimize.
974 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
977 2020-02-05 Richard Biener <rguenther@suse.de>
980 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
981 checks before matching calls.
983 2020-02-05 Jakub Jelinek <jakub@redhat.com>
985 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
986 function comment typo.
989 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
990 simd_clone_create failed when i == 0, adjust clone->nargs by
993 2020-02-05 Martin Liska <mliska@suse.cz>
996 * doc/invoke.texi: Document that one should
997 not combine ASLR and -fpch.
999 2020-02-04 Richard Biener <rguenther@suse.de>
1001 PR tree-optimization/93538
1002 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1004 2020-02-04 Richard Biener <rguenther@suse.de>
1006 PR tree-optimization/91123
1007 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1008 (vn_walk_cb_data::last_vuse): New member.
1009 (vn_walk_cb_data::saved_operands): Likewsie.
1010 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1011 (vn_walk_cb_data::push_partial_def): Use finish.
1012 (vn_reference_lookup_2): Update last_vuse and use finish if
1013 we've saved operands.
1014 (vn_reference_lookup_3): Use finish and update calls to
1015 push_partial_defs everywhere. When translating through
1016 memcpy or aggregate copies save off operands and alias-set.
1017 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1018 operation for redundant store removal.
1020 2020-02-04 Richard Biener <rguenther@suse.de>
1022 PR tree-optimization/92819
1023 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1024 generating more stmts than before.
1026 2020-02-04 Martin Liska <mliska@suse.cz>
1028 * config/arm/arm.c (arm_gen_far_branch): Move the function
1029 outside of selftests.
1031 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1033 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1034 function to adjust PC-relative vector addresses.
1035 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1036 handle vectors with PC-relative addresses.
1038 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1040 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1042 (hard_reg_and_mode_to_addr_mask): Delete.
1043 (rs6000_adjust_vec_address): If the original vector address
1044 was REG+REG or REG+OFFSET and the element is not zero, do the add
1045 of the elements in the original address before adding the offset
1046 for the vector element. Use address_to_insn_form to validate the
1047 address using the register being loaded, rather than guessing
1048 whether the address is a DS-FORM or DQ-FORM address.
1050 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1052 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1053 to calculate the offset in memory from the start of a vector of a
1054 particular element. Add code to keep the element number in
1055 bounds if the element number is variable.
1056 (rs6000_adjust_vec_address): Move calculation of offset of the
1057 vector element to get_vector_offset.
1058 (rs6000_split_vec_extract_var): Do not do the initial AND of
1059 element here, move the code to get_vector_offset.
1061 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1063 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1066 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1068 * config/rs6000/constraints.md: Improve documentation.
1070 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1073 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1074 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1076 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1078 * config.gcc: Remove "carrizo" support.
1079 * config/gcn/gcn-opts.h (processor_type): Likewise.
1080 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1081 * config/gcn/gcn.opt (gpu_type): Likewise.
1082 * config/gcn/t-omp-device: Likewise.
1084 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1087 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1088 * config/arm/arm.c (arm_gen_far_branch): New function
1090 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1092 2020-02-03 Julian Brown <julian@codesourcery.com>
1093 Tobias Burnus <tobias@codesourcery.com>
1095 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1097 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1100 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1101 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1104 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1106 PR rtl-optimization/91333
1107 * ira-color.c (struct allocno_color_data): Add member
1109 (init_allocno_threads): Set the member up.
1110 (bucket_allocno_compare_func): Add compare hard reg
1113 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1115 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1117 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1118 * config.in: Regenerated.
1119 * configure: Regenerated.
1120 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1121 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1122 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1124 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1126 * configure: Regenerate.
1128 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1130 PR rtl-optimization/91333
1131 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1132 reg preferences comparison up.
1134 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1136 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1137 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1138 aarch64-sve-builtins-base.h.
1139 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1140 aarch64-sve-builtins-base.cc.
1141 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1142 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1144 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1145 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1146 (svcvtnt): New functions.
1147 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1148 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1149 (svcvtnt): New functions.
1150 (svcvt): Add a form that converts f32 to bf16.
1151 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1152 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1154 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1155 Treat B as bfloat16_t.
1156 (ternary_bfloat_lane_base): New class.
1157 (ternary_bfloat_def): Likewise.
1158 (ternary_bfloat): New shape.
1159 (ternary_bfloat_lane_def): New class.
1160 (ternary_bfloat_lane): New shape.
1161 (ternary_bfloat_lanex2_def): New class.
1162 (ternary_bfloat_lanex2): New shape.
1163 (ternary_bfloat_opt_n_def): New class.
1164 (ternary_bfloat_opt_n): New shape.
1165 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1166 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1167 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1168 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1169 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1170 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1171 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1172 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1173 the pattern off the narrow mode instead of the wider one.
1174 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1175 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1176 (sve_fp_op): Handle them.
1177 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1178 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1180 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1182 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1183 * config/aarch64/aarch64-modes.def (BF): Move definition before
1184 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1185 (SVE_MODES): Handle BF modes.
1186 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1188 (aarch64_full_sve_mode): Likewise.
1189 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1191 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1192 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1193 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1194 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1196 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1198 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1199 (TYPES_all_data): Add bf16.
1200 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1201 (register_tuple_type): Increase buffer size.
1202 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1203 (bf16): New type suffix.
1204 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1205 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1206 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1207 Change type from all_data to all_arith.
1208 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1211 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1212 Matthew Malcomson <matthew.malcomson@arm.com>
1213 Richard Sandiford <richard.sandiford@arm.com>
1215 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1216 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1217 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1218 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1219 __ARM_FEATURE_MATMUL_FP64.
1220 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1221 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1222 be disabled at the same time.
1223 (f32mm): New extension.
1224 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1225 (AARCH64_FL_F64MM): Bump to the next bit up.
1226 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1227 (TARGET_SVE_F64MM): New macros.
1228 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1229 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1230 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1231 (UNSPEC_ZIP2Q): New unspeccs.
1232 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1233 (optab, sur, perm_insn): Handle the new unspecs.
1234 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1235 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1236 TARGET_SVE_F64MM instead of separate tests.
1237 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1238 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1239 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1240 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1241 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1242 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1243 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1244 (TYPES_s_signed): New macro.
1245 (TYPES_s_integer): Use it.
1246 (TYPES_d_float): New macro.
1247 (TYPES_d_data): Use it.
1248 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1249 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1250 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1251 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1252 (svmmla): New shape.
1253 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1254 template parameters.
1255 (ternary_resize2_lane_base): Likewise.
1256 (ternary_resize2_base): New class.
1257 (ternary_qq_lane_base): Likewise.
1258 (ternary_intq_uintq_lane_def): Likewise.
1259 (ternary_intq_uintq_lane): New shape.
1260 (ternary_intq_uintq_opt_n_def): New class
1261 (ternary_intq_uintq_opt_n): New shape.
1262 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1263 (ternary_uintq_intq_def): New class.
1264 (ternary_uintq_intq): New shape.
1265 (ternary_uintq_intq_lane_def): New class.
1266 (ternary_uintq_intq_lane): New shape.
1267 (ternary_uintq_intq_opt_n_def): New class.
1268 (ternary_uintq_intq_opt_n): New shape.
1269 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1270 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1271 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1272 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1274 (svdotprod_lane_impl): ...this new class.
1275 (svmmla_impl, svusdot_impl): New classes.
1276 (svdot_lane): Update to use svdotprod_lane_impl.
1277 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1278 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1280 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1281 function, with no types defined.
1282 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1283 AARCH64_FL_I8MM functions.
1284 (svmmla): New AARCH64_FL_F32MM function.
1285 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1286 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1287 AARCH64_FL_F64MM function.
1288 (REQUIRED_EXTENSIONS):
1290 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1292 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1295 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1297 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1298 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1299 (*movti_internal): Do not check for
1300 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1301 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1302 just after check for TARGET_AVX.
1303 (*movdf_internal): Ditto.
1304 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1305 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1306 * config/i386/sse.md (mov<mode>_internal): Only check
1307 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1308 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1309 (<sse>_andnot<mode>3<mask_name>): Move check for
1310 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1311 (<code><mode>3<mask_name>): Ditto.
1312 (*andnot<mode>3): Ditto.
1313 (*andnottf3): Ditto.
1314 (*<code><mode>3): Ditto.
1315 (*<code>tf3): Ditto.
1316 (*andnot<VI:mode>3): Remove
1317 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1318 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1319 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1320 (sse4_1_blendv<ssemodesuffix>): Ditto.
1321 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1322 Explain that tune applies to 128bit instructions only.
1324 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1326 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1327 to definition of hsa_kernel_description. Parse assembly to find SGPR
1328 and VGPR count of kernel and store in hsa_kernel_description.
1330 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1332 PR rtl-optimization/91838
1333 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1334 to truncate if allowed or reject combination.
1336 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1338 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1339 (find_inv_vars_cb): Likewise.
1341 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1343 * calls.c (special_function_p): Split out the check for DECL_NAME
1344 being non-NULL and fndecl being extern at file scope into a
1345 new maybe_special_function_p and call it. Drop check for fndecl
1346 being non-NULL that was after a usage of DECL_NAME (fndecl).
1347 * tree.h (maybe_special_function_p): New inline function.
1349 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1351 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1352 (mask_gather_load<mode>): ... here, and zero-initialize the
1354 (maskload<mode>di): Zero-initialize the destination.
1357 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1360 * doc/analyzer.texi (Limitations): Note that constraints on
1361 floating-point values are currently ignored.
1363 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1366 * symtab.c (symtab_node::noninterposable_alias): If localalias
1367 already exists, but is not usable, append numbers after it until
1368 a unique name is found. Formatting fix.
1371 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1374 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1376 * config/gcn/gcn.c (print_operand): Handle LTGT.
1377 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1379 2020-01-30 Richard Biener <rguenther@suse.de>
1381 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1382 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1384 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1386 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1387 without a DECL in .data.rel.ro.local.
1389 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1392 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1396 * config/i386/sse.md
1397 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1398 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1399 any_extend code iterator instead of always zero_extend.
1400 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1401 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1402 Use any_extend code iterator instead of always zero_extend.
1403 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1404 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1405 Use any_extend code iterator instead of always zero_extend.
1406 (*sse2_pmovmskb_ext): New define_insn.
1407 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1410 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1411 (*popcountsi2_zext_falsedep): New define_insn.
1413 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1415 * config.in: Regenerated.
1416 * configure: Regenerated.
1418 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1421 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1422 LLVM's assembler changed the default in version 9.
1424 2020-01-24 Jeff Law <law@redhat.com>
1426 PR tree-optimization/89689
1427 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1429 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1433 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1435 PR rtl-optimization/87763
1436 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1437 simplification to handle subregs as well as bare regs.
1438 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1440 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1443 * ira.c (ira): Revert use of simplified LRA algorithm.
1445 2020-01-29 Martin Jambor <mjambor@suse.cz>
1447 PR tree-optimization/92706
1448 * tree-sra.c (struct access): Fields first_link, last_link,
1449 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1450 next_rhs_queued and grp_rhs_queued respectively, new fields
1451 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1452 (struct assign_link): Field next renamed to next_rhs, new field
1453 next_lhs. Updated comment.
1454 (work_queue_head): Renamed to rhs_work_queue_head.
1455 (lhs_work_queue_head): New variable.
1456 (add_link_to_lhs): New function.
1457 (relink_to_new_repr): Also relink LHS lists.
1458 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1459 (add_access_to_lhs_work_queue): New function.
1460 (pop_access_from_work_queue): Renamed to
1461 pop_access_from_rhs_work_queue.
1462 (pop_access_from_lhs_work_queue): New function.
1463 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1465 (child_would_conflict_in_lacc): Renamed to
1466 child_would_conflict_in_acc. Adjusted parameter names.
1467 (create_artificial_child_access): New parameter set_grp_read, use it.
1468 (subtree_mark_written_and_enqueue): Renamed to
1469 subtree_mark_written_and_rhs_enqueue.
1470 (propagate_subaccesses_across_link): Renamed to
1471 propagate_subaccesses_from_rhs.
1472 (propagate_subaccesses_from_lhs): New function.
1473 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1476 2020-01-29 Martin Jambor <mjambor@suse.cz>
1478 PR tree-optimization/92706
1479 * tree-sra.c (struct access): Adjust comment of
1480 grp_total_scalarization.
1481 (find_access_in_subtree): Look for single children spanning an entire
1483 (scalarizable_type_p): Allow register accesses, adjust callers.
1484 (completely_scalarize): Remove function.
1485 (scalarize_elem): Likewise.
1486 (create_total_scalarization_access): Likewise.
1487 (sort_and_splice_var_accesses): Do not track total scalarization
1489 (analyze_access_subtree): New parameter totally, adjust to new meaning
1490 of grp_total_scalarization.
1491 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1492 (can_totally_scalarize_forest_p): New function.
1493 (create_total_scalarization_access): Likewise.
1494 (create_total_access_and_reshape): Likewise.
1495 (total_should_skip_creating_access): Likewise.
1496 (totally_scalarize_subtree): Likewise.
1497 (analyze_all_variable_accesses): Perform total scalarization after
1498 subaccess propagation using the new functions above.
1499 (initialize_constant_pool_replacements): Output initializers by
1500 traversing the access tree.
1502 2020-01-29 Martin Jambor <mjambor@suse.cz>
1504 * tree-sra.c (verify_sra_access_forest): New function.
1505 (verify_all_sra_access_forests): Likewise.
1506 (create_artificial_child_access): Set parent.
1507 (analyze_all_variable_accesses): Call the verifier.
1509 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1511 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1512 if called on indirect edge.
1513 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1514 speculative call if needed.
1516 2020-01-29 Richard Biener <rguenther@suse.de>
1518 PR tree-optimization/93428
1519 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1520 permutation when the load node is created.
1521 (vect_analyze_slp_instance): Re-use it here.
1523 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1525 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1527 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1529 PR rtl-optimization/93272
1530 * ira-lives.c (process_out_of_region_eh_regs): New function.
1531 (process_bb_node_lives): Call it.
1533 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1535 * coverage.c (read_counts_file): Make error message lowercase.
1537 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1539 * profile-count.c (profile_quality_display_names): Fix ordering.
1541 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1544 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1545 hash only when edge is first within the sequence.
1546 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1547 (symbol_table::create_edge): Do not set target_prob.
1548 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1550 (cgraph_edge::make_speculative): Drop target_prob parameter.
1551 (cgraph_edge::speculative_call_info): Remove.
1552 (cgraph_edge::first_speculative_call_target): New member function.
1553 (update_call_stmt_hash_for_removing_direct_edge): New function.
1554 (cgraph_edge::resolve_speculation): Rewrite to new API.
1555 (cgraph_edge::speculative_call_for_target): New member function.
1556 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1557 multiple speculation targets.
1558 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1560 (verify_speculative_call): Verify that targets form an interval.
1561 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1562 (cgraph_edge::first_speculative_call_target): New member function.
1563 (cgraph_edge::next_speculative_call_target): New member function.
1564 (cgraph_edge::speculative_call_target_ref): New member function.
1565 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1566 (cgraph_edge): Remove target_prob.
1567 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1568 Fix handling of speculative calls.
1569 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1570 * ipa-fnsummary.c (analyze_function_body): Likewise.
1571 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1572 * ipa-profile.c (dump_histogram): Fix formating.
1573 (ipa_profile_generate_summary): Watch for overflows.
1574 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1575 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1576 (update_indirect_edges_after_inlining): Update to new API.
1577 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1579 * profile-count.h: (profile_probability::adjusted): New.
1580 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1581 updating of profile.
1582 * value-prof.c (gimple_ic_transform): Rename to ...
1583 (dump_ic_profile): ... this one; update dumping.
1584 (stream_in_histogram_value): Fix formating.
1585 (gimple_value_profile_transformations): Update.
1587 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1590 * config/i386/i386.md (*movoi_internal_avx): Remove
1591 TARGET_SSE_TYPELESS_STORES check.
1592 (*movti_internal): Prefer TARGET_AVX over
1593 TARGET_SSE_TYPELESS_STORES.
1594 (*movtf_internal): Likewise.
1595 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1596 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1597 from TARGET_SSE_TYPELESS_STORES.
1599 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1601 * diagnostic-core.h (warning_at): Rename overload to...
1602 (warning_meta): ...this.
1603 (emit_diagnostic_valist): Delete decl of overload taking
1604 diagnostic_metadata.
1605 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1606 (warning_at): Rename overload taking diagnostic_metadata to...
1607 (warning_meta): ...this.
1609 2020-01-28 Richard Biener <rguenther@suse.de>
1611 PR tree-optimization/93439
1612 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1613 * tree-cfg.c (move_sese_region_to_fn): ... here.
1614 (verify_types_in_gimple_reference): Verify used cliques are
1617 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1620 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1621 argument of a pointer to struct gcc_options and pass it to
1622 parse_mtune_ctrl_str.
1623 (ix86_function_specific_restore): Pass opts to
1624 set_ix86_tune_features.
1625 (ix86_option_override_internal): Likewise.
1626 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1627 gcc_options and use it for x_ix86_tune_ctrl_string.
1629 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1631 PR rtl-optimization/87763
1632 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1633 simplification to handle subregs as well as bare regs.
1634 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1636 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1638 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1639 for reduction chains that (now) include a call.
1641 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1643 PR tree-optimization/92822
1644 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1645 out the don't-care elements of a vector whose significant elements
1646 are duplicates, make the don't-care elements duplicates too.
1648 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1650 PR tree-optimization/93434
1651 * tree-predcom.c (split_data_refs_to_components): Record which
1652 components have had aliasing loads removed. Prevent store-store
1653 commoning for all such components.
1655 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1658 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1659 -1 or is_vshift is true, use new_vector with number of elts npatterns
1660 rather than new_unary_operation.
1662 PR tree-optimization/93454
1663 * gimple-fold.c (fold_array_ctor_reference): Perform
1664 elt_size.to_uhwi () just once, instead of calling it in every
1665 iteration. Punt if that value is above size of the temporary
1666 buffer. Decrease third native_encode_expr argument when
1667 bufoff + elt_sz is above size of buf.
1669 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1671 * config/mips/mips.c (mips_declare_object_name)
1672 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1674 2020-01-27 Martin Liska <mliska@suse.cz>
1676 PR gcov-profile/93403
1677 * tree-profile.c (gimple_init_gcov_profiler): Generate
1678 both __gcov_indirect_call_profiler_v4 and
1679 __gcov_indirect_call_profiler_v4_atomic.
1681 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1684 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1686 (@aarch64_split_simd_mov<mode>): Use it.
1687 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1688 Leave the vec_extract patterns to handle 2-element vectors.
1689 (aarch64_simd_mov_from_<mode>high): Likewise.
1690 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1691 (vec_extractv2dfv1df): Likewise.
1693 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1695 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1696 jump conditions for *compare_condjump<GPI:mode>.
1698 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1701 * digraph.cc (test_edge::test_edge): Specify template for base
1704 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1706 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1708 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1710 * config/arc/arc-protos.h (gen_mlo): Remove.
1711 (gen_mhi): Likewise.
1712 * config/arc/arc.c (AUX_MULHI): Define.
1713 (arc_must_save_reister): Special handling for r58/59.
1714 (arc_compute_frame_size): Consider mlo/mhi registers.
1715 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1717 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1718 mlo/mhi name selection.
1719 (arc_restore_callee_saves): Don't early restore blink when ISR.
1720 (arc_expand_prologue): Add mlo/mhi saving.
1721 (arc_expand_epilogue): Add mlo/mhi restoring.
1724 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1725 numbering when MUL64 option is used.
1726 (DWARF2_FRAME_REG_OUT): Define.
1727 * config/arc/arc.md (arc600_stall): New pattern.
1728 (VUNSPEC_ARC_ARC600_STALL): Define.
1729 (mulsi64): Use correct mlo/mhi registers.
1730 (mulsi_600): Clean it up.
1731 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1733 (mhi_operand): Likewise.
1735 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1736 Petro Karashchenko <petro.karashchenko@ring.com>
1738 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1739 attributes if needed.
1740 (prepare_move_operands): Generate special unspec instruction for
1742 (arc_isuncached_mem_p): Propagate uncached attribute to each
1744 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1745 (VUNSPEC_ARC_STDI): Likewise.
1746 (ALLI): New mode iterator.
1747 (mALLI): New mode attribute.
1748 (lddi): New instruction pattern.
1750 (stdidi_split): Split instruction for architectures which are not
1751 supporting ll64 option.
1752 (lddidi_split): Likewise.
1754 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1756 PR rtl-optimization/92989
1757 * lra-lives.c (process_bb_lives): Update the live-in set before
1758 processing additional clobbers.
1760 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1762 PR rtl-optimization/93170
1763 * cselib.c (cselib_invalidate_regno_val): New function, split out
1765 (cselib_invalidate_regno): ...here.
1766 (cselib_invalidated_by_call_p): New function.
1767 (cselib_process_insn): Iterate over all the hard-register entries in
1768 REG_VALUES and invalidate any that cross call-clobbered registers.
1770 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1772 * dojump.c (split_comparison): Use HONOR_NANS rather than
1773 HONOR_SNANS when splitting LTGT.
1775 2020-01-27 Martin Liska <mliska@suse.cz>
1778 * opts.c (print_filtered_help): Exclude language-specific
1779 options from --help=common unless enabled in all FEs.
1781 2020-01-27 Martin Liska <mliska@suse.cz>
1783 * opts.c (print_help): Exclude params from
1784 all except --help=param.
1786 2020-01-27 Martin Liska <mliska@suse.cz>
1789 * config/i386/i386-features.c (make_resolver_func):
1790 Align the code with ppc64 target implementation.
1791 Do not generate a unique name for resolver function.
1793 2020-01-27 Richard Biener <rguenther@suse.de>
1795 PR tree-optimization/93397
1796 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1797 converted reduction chain SLP graph adjustment.
1799 2020-01-26 Marek Polacek <polacek@redhat.com>
1802 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1805 2020-01-26 Jason Merrill <jason@redhat.com>
1808 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1811 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1813 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1814 (rx_setmem): Likewise.
1816 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1819 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1820 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1821 drop <di> from constraint of last operand.
1824 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1825 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1826 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1828 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1831 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1834 2020-01-24 Jeff Law <law@redhat.com>
1836 PR tree-optimization/92788
1837 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1840 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1843 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1844 *avx_vperm_broadcast_<mode>,
1845 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1846 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1847 Move before avx2_perm<mode>/avx512f_perm<mode>.
1850 * simplify-rtx.c (simplify_const_unary_operation,
1851 simplify_const_binary_operation): Punt for mode precision above
1852 MAX_BITSIZE_MODE_ANY_INT.
1854 2020-01-24 Andrew Pinski <apinski@marvell.com>
1856 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1859 2020-01-24 Jeff Law <law@redhat.com>
1862 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1863 for REGs. Call output_operand_lossage to get more reasonable
1866 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1868 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1869 gcn_fp_compare_operator.
1870 (vec_cmpu<mode>di): Use gcn_compare_operator.
1871 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1872 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1873 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1874 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1875 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1876 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1877 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1878 gcn_fp_compare_operator.
1879 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1880 gcn_fp_compare_operator.
1881 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1882 gcn_fp_compare_operator.
1883 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1884 gcn_fp_compare_operator.
1886 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1888 * doc/install.texi (Cross-Compiler-Specific Options): Document
1889 `--with-toolexeclibdir' option.
1891 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1893 * target.def (flags_regnum): Also mention effect on delay slot filling.
1894 * doc/tm.texi: Regenerate.
1896 2020-01-23 Jeff Law <law@redhat.com>
1898 PR translation/90162
1899 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1901 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1904 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1907 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1909 PR rtl-optimization/93402
1910 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1913 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1915 * config.in: Regenerated.
1916 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1917 for TARGET_LIBC_GNUSTACK.
1918 * configure: Regenerated.
1919 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1920 found to be 2.31 or greater.
1922 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1924 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1926 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1927 (mips_asm_file_end): New function. Delegate to
1928 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1929 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1931 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1934 * config/i386/i386-modes.def (POImode): New mode.
1935 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1936 * config/i386/i386.md (DPWI): New mode attribute.
1937 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1939 (QPWI): ... this. Use POI instead of OI for TImode.
1940 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1941 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1944 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1947 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1949 (speculation_tracker_rev): New pattern.
1950 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1951 Use speculation_tracker_rev to track the inverse condition.
1953 2020-01-23 Richard Biener <rguenther@suse.de>
1955 PR tree-optimization/93381
1956 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1957 alias-set of the def as argument and record the first one.
1958 (vn_walk_cb_data::first_set): New member.
1959 (vn_reference_lookup_3): Pass the alias-set of the current def
1960 to push_partial_def. Fix alias-set used in the aggregate copy
1962 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1963 * real.c (clear_significand_below): Fix out-of-bound access.
1965 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1968 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1969 New define_insn patterns.
1971 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1973 * doc/sourcebuild.texi (check-function-bodies): Add an
1974 optional target/xfail selector.
1976 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1978 PR rtl-optimization/93124
1979 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1980 bare USE and CLOBBER insns.
1982 2020-01-22 Andrew Pinski <apinski@marvell.com>
1984 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1986 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1989 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1990 diagnostic_manager into "ana" namespace.
1991 * selftest-run-tests.c (selftest::run_tests): Update for move of
1992 selftest::run_analyzer_selftests to
1993 ana::selftest::run_analyzer_selftests.
1995 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1997 * cfgexpand.c (union_stack_vars): Update the size.
1999 2020-01-22 Richard Biener <rguenther@suse.de>
2001 PR tree-optimization/93381
2002 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2003 throughout, handle all conversions the same.
2005 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2008 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2009 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2010 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2011 Call force_reg on high_in2 unconditionally.
2013 2020-01-22 Martin Liska <mliska@suse.cz>
2015 PR tree-optimization/92924
2016 * profile.c (compute_value_histograms): Divide
2019 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2022 * output.h (assemble_name_resolve): Declare.
2023 * varasm.c (assemble_name_resolve): New function.
2024 (assemble_name): Use it.
2025 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2027 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2029 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2030 update_web_docs_git instead of update_web_docs_svn.
2032 2020-01-21 Andrew Pinski <apinski@marvell.com>
2035 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2036 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2037 (*tlsgd_small_<mode>): Likewise.
2038 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2039 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2040 register. Convert that register back to dest using convert_mode.
2042 2020-01-21 Jim Wilson <jimw@sifive.com>
2044 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2047 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2048 Uros Bizjak <ubizjak@gmail.com>
2051 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2053 (legitimize_tls_address): Do GNU2 TLS address computation in
2054 ptr_mode and zero-extend result to Pmode.
2055 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2056 :P with :PTR and Pmode with ptr_mode.
2057 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2058 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2059 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2061 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2064 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2065 the last two operands are CONST_INT_P before using them as such.
2067 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2069 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2070 to get the integer element types.
2072 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2074 * config/aarch64/aarch64-sve-builtins.h
2075 (function_expander::convert_to_pmode): Declare.
2076 * config/aarch64/aarch64-sve-builtins.cc
2077 (function_expander::convert_to_pmode): New function.
2078 (function_expander::get_contiguous_base): Use it.
2079 (function_expander::prepare_gather_address_operands): Likewise.
2080 * config/aarch64/aarch64-sve-builtins-sve2.cc
2081 (svwhilerw_svwhilewr_impl::expand): Likewise.
2083 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2086 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2087 cfun->machine->label_is_assembled.
2088 (aarch64_print_patchable_function_entry): New.
2089 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2090 * config/aarch64/aarch64.h (struct machine_function): New field,
2093 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2096 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2099 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2102 * cgraph.c (cgraph_edge::resolve_speculation,
2103 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2104 call_stmt_site_hash.
2106 2020-01-21 Martin Liska <mliska@suse.cz>
2108 * config/rs6000/rs6000.c (common_mode_defined): Remove
2111 2020-01-21 Richard Biener <rguenther@suse.de>
2113 PR tree-optimization/92328
2114 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2115 type when value-numbering same-sized store by inserting a
2117 (eliminate_dom_walker::eliminate_stmt): When eliminating
2118 a redundant store handle bit-reinterpretation of the same value.
2120 2020-01-21 Andrew Pinski <apinski@marvel.com>
2123 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2125 (prepare_block_for_update): This. Use a worklist instead of
2128 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2130 * gcc/config/arm/arm.c (clear_operation_p):
2131 Initialise last_regno, skip first iteration
2132 based on the first_set value and use ints instead
2133 of the unnecessary HOST_WIDE_INTs.
2135 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2138 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2139 compare_mode other than SFmode or DFmode.
2141 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2144 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2145 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2146 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2148 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2150 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2152 2020-01-20 Andrew Pinski <apinski@marvell.com>
2155 * targhooks.c (default_print_patchable_function_entry): Use
2156 output_asm_insn to emit the nop instruction.
2158 2020-01-20 Fangrui Song <maskray@google.com>
2161 * targhooks.c (default_print_patchable_function_entry): Align to
2164 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2167 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2168 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2169 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2170 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2171 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2172 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2173 Remove the {q} suffix from lea.
2174 (*tls_dynamic_gnu2_call_64): Renamed to ...
2175 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2176 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2177 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2178 Pass Pmode to gen_tls_dynamic_gnu2_64.
2180 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2182 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2184 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2186 * config/aarch64/aarch64-sve-builtins-base.cc
2187 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2189 2020-01-20 Richard Biener <rguenther@suse.de>
2192 * dwarf2out.c (prune_unused_types): Unconditionally mark
2193 called function DIEs.
2195 2020-01-20 Martin Liska <mliska@suse.cz>
2197 PR tree-optimization/93199
2198 * tree-eh.c (struct leh_state): Add
2199 new field outer_non_cleanup.
2200 (cleanup_is_dead_in): Pass leh_state instead
2201 of eh_region. Add a checking that state->outer_non_cleanup
2202 points to outer non-clean up region.
2203 (lower_try_finally): Record outer_non_cleanup
2205 (lower_catch): Likewise.
2206 (lower_eh_filter): Likewise.
2207 (lower_eh_must_not_throw): Likewise.
2208 (lower_cleanup): Likewise.
2210 2020-01-20 Richard Biener <rguenther@suse.de>
2212 PR tree-optimization/93094
2213 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2214 (vect_transform_loop): Likewise.
2215 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2216 loop_vectorized_call to vect_transform_loop.
2217 * tree-vect-loop.c (vect_transform_loop): Pass down
2218 loop_vectorized_call to vect_loop_versioning.
2219 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2220 the earlier discovered loop_vectorized_call.
2222 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2224 * doc/contribute.texi: Update for SVN -> Git transition.
2225 * doc/install.texi: Likewise.
2227 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2230 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2231 speculative targets.
2232 (verify_speculative_call): New function
2233 (cgraph_node::verify_node): Use it.
2234 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2237 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2240 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2241 (cgraph_edge::make_direct): Remove all indirect targets.
2242 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2243 (cgraph_node::verify_node): Verify that only one call_stmt or
2244 lto_stmt_uid is set.
2245 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2247 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2248 (lto_output_ref): Simplify streaming of stmt.
2249 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2251 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2253 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2254 Mark parameter unused.
2256 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2258 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2260 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2262 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2264 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2266 * Makefile.in: Add coroutine-passes.o.
2267 * builtin-types.def (BT_CONST_SIZE): New.
2268 (BT_FN_BOOL_PTR): New.
2269 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2270 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2271 * coroutine-builtins.def: New file.
2272 * coroutine-passes.cc: New file.
2273 * function.h (struct GTY function): Add a bit to indicate that the
2274 function is a coroutine component.
2275 * internal-fn.c (expand_CO_FRAME): New.
2276 (expand_CO_YIELD): New.
2277 (expand_CO_SUSPN): New.
2278 (expand_CO_ACTOR): New.
2279 * internal-fn.def (CO_ACTOR): New.
2283 * passes.def: Add pass_coroutine_lower_builtins,
2284 pass_coroutine_early_expand_ifns.
2285 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2286 (make_pass_coroutine_early_expand_ifns): New.
2287 * doc/invoke.texi: Document the fcoroutines command line
2290 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2292 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2295 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2296 after checking the argument is a REG. Don't use REGNO (reg)
2297 again to set last_regno, reuse regno variable instead.
2299 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2301 * doc/analyzer.texi (Limitations): Add note about NaN.
2303 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2304 Sudakshina Das <sudi.das@arm.com>
2306 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2307 and valid immediate.
2308 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2309 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2310 * config/arm/constraints.md (Pg): New.
2311 * config/arm/predicates.md (long_shift_imm): New.
2312 (arm_reg_or_long_shift_imm): Likewise.
2313 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2314 (thumb2_lsll): Likewise.
2317 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2318 Sudakshina Das <sudi.das@arm.com>
2320 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2321 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2322 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2323 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2324 * config/arm/thumb2.md (thumb2_asrl): New.
2325 (thumb2_lsll): Likewise.
2327 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2329 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2332 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2334 * gdbinit.in (help-gcc-hooks): New command.
2335 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2336 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2339 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2341 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2342 correct target macro.
2344 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2346 * config/aarch64/aarch64-protos.h
2347 (aarch64_sve_ld1ro_operand_p): New.
2348 * config/aarch64/aarch64-sve-builtins-base.cc
2349 (class load_replicate): New.
2350 (class svld1ro_impl): New.
2351 (class svld1rq_impl): Change to inherit from load_replicate.
2352 (svld1ro): New sve intrinsic function base.
2353 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2354 New DEF_SVE_FUNCTION.
2355 * config/aarch64/aarch64-sve-builtins-base.h
2356 (svld1ro): New decl.
2357 * config/aarch64/aarch64-sve-builtins.cc
2358 (function_expander::add_mem_operand): Modify assert to allow
2360 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2362 * config/aarch64/aarch64.c
2363 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2364 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2365 (aarch64_sve_ld1ro_operand_p): New.
2366 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2367 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2368 * config/aarch64/predicates.md
2369 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2371 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2373 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2374 Introduce this ACLE specified predefined macro.
2375 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2376 (fp): Disabling this disables f64mm.
2377 (simd): Disabling this disables f64mm.
2378 (fp16): Disabling this disables f64mm.
2379 (sve): Disabling this disables f64mm.
2380 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2381 (AARCH64_ISA_F64MM): New.
2382 (TARGET_F64MM): New.
2383 * doc/invoke.texi (f64mm): Document new option.
2385 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2387 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2388 (neoversen1_tunings): Likewise.
2390 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2393 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2394 Add assert to ensure prolog has been emitted.
2395 (aarch64_split_atomic_op): Likewise.
2396 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2397 Use epilogue_completed rather than reload_completed.
2398 (aarch64_atomic_exchange<mode>): Likewise.
2399 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2400 (atomic_nand<mode>): Likewise.
2401 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2402 (atomic_fetch_nand<mode>): Likewise.
2403 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2404 (atomic_nand_fetch<mode>): Likewise.
2406 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2409 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2411 (REVERSE_CONDITION): Delete.
2412 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2413 (CCFP_CCFPE): Likewise.
2414 (e): New mode attribute.
2415 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2416 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2417 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2418 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2419 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2420 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2421 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2422 name of generator from gen_ccmpdi to gen_ccmpccdi.
2423 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2424 the previous comparison but aren't able to, use the new ccmp_rev
2427 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2429 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2430 than testing directly for INTEGER_CST.
2431 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2433 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2435 PR tree-optimization/93292
2436 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2437 get_vectype_for_scalar_type returns NULL.
2439 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2441 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2442 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2444 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2446 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2448 * params.opt: (max-predicted-iterations): Set bounds.
2449 * predict.c (real_almost_one, real_br_prob_base,
2450 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2451 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2452 probabilities; do not truncate to reg_br_prob_bases.
2453 (estimate_loops_at_level): Pass max_cyclic_prob.
2454 (estimate_loops): Compute max_cyclic_prob.
2455 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2457 * profile-count.c (profile_probability::to_sreal): New.
2458 * profile-count.h (class sreal): Move up in file.
2459 (profile_probability::to_sreal): Declare.
2461 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2464 (arm_invalid_conversion): New function for target hook.
2465 (arm_invalid_unary_op): New function for target hook.
2466 (arm_invalid_binary_op): New function for target hook.
2468 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2470 * config.gcc: Add arm_bf16.h.
2471 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2472 (arm_simd_builtin_std_type): Add BFmode.
2473 (arm_init_simd_builtin_types): Define element types for vector types.
2474 (arm_init_bf16_types): New function.
2475 (arm_init_builtins): Add arm_init_bf16_types function call.
2476 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2477 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2478 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2479 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2480 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2481 (arm_mangle_type): Add __bf16.
2482 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2483 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2484 arm_bf16_ptr_type_node.
2485 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2486 define_split between ARM registers.
2487 * config/arm/arm_bf16.h: New file.
2488 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2489 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2491 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2492 * config/arm/vfp.md: Add BFmode to movhf patterns.
2494 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2495 Andre Vieira <andre.simoesdiasvieira@arm.com>
2497 * config/arm/arm-cpus.in (mve, mve_float): New features.
2498 (dsp, mve, mve.fp): New options.
2499 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2500 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2501 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2503 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2504 Thomas Preud'homme <thomas.preudhomme@arm.com>
2506 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2508 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2509 error for using -mcmse when targeting Armv8.1-M Mainline.
2511 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2512 Thomas Preud'homme <thomas.preudhomme@arm.com>
2514 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2515 address in r4 when targeting Armv8.1-M Mainline.
2516 (nonsecure_call_value_internal): Likewise.
2517 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2518 a register match_operand again. Emit BLXNS when targeting
2520 (nonsecure_call_value_reg_thumb2): Likewise.
2522 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2523 Thomas Preud'homme <thomas.preudhomme@arm.com>
2525 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2526 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2527 variable as true when floating-point ABI is not hard. Replace
2528 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2529 Generate VLSTM and VLLDM instruction respectively before and
2530 after a function call to cmse_nonsecure_call function.
2531 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2532 (VUNSPEC_VLLDM): Likewise.
2533 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2534 (lazy_load_multiple_insn): Likewise.
2536 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2537 Thomas Preud'homme <thomas.preudhomme@arm.com>
2539 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2540 (arm_emit_vfp_multi_reg_pop): Likewise.
2541 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2542 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2543 restore callee-saved VFP registers.
2545 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2546 Thomas Preud'homme <thomas.preudhomme@arm.com>
2548 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2549 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2550 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2551 callee-saved GPRs as well as clear ip register before doing a nonsecure
2552 call then restore callee-saved GPRs after it when targeting
2554 (arm_reorg): Adapt to function rename.
2556 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2557 Thomas Preud'homme <thomas.preudhomme@arm.com>
2559 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2560 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2561 clear_vfp_multiple pattern based on a new vfp parameter.
2562 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2563 targeting Armv8.1-M Mainline.
2564 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2565 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2566 whether VFP registers are available before looking call_used_regs for a
2568 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2569 of prototype of clear_operation_p.
2570 (clear_vfp_multiple_operation): New predicate.
2571 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2572 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2574 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2575 Thomas Preud'homme <thomas.preudhomme@arm.com>
2577 * config/arm/arm-protos.h (clear_operation_p): Declare.
2578 * config/arm/arm.c (clear_operation_p): New function.
2579 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2580 targeting Armv8.1-M Mainline or successor.
2581 (output_return_instruction): Only output APSR register clearing if
2582 Armv8.1-M Mainline instructions not available.
2583 (thumb_exit): Likewise.
2584 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2585 * config/arm/thumb2.md (clear_apsr): New define_insn.
2586 (clear_multiple): Likewise.
2587 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2589 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2590 Thomas Preud'homme <thomas.preudhomme@arm.com>
2592 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2593 (use_return_insn): Also return false for Armv8.1-M Mainline.
2594 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2595 Mainline instructions are available.
2596 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2597 when targeting Armv8.1-M Mainline Security Extensions.
2598 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2599 Mainline entry function.
2600 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2601 targeting Armv8.1-M Mainline or successor.
2602 (arm_expand_epilogue): Fix indentation of caller-saved register
2603 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2605 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2606 (FP_SYSREGS): Likewise.
2607 (enum vfp_sysregs_encoding): Define enum.
2608 (fp_sysreg_names): Declare.
2609 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2610 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2611 (pop_fpsysreg_insn): Likewise.
2613 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2614 Thomas Preud'homme <thomas.preudhomme@arm.com>
2616 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2617 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2618 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2619 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2620 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2621 (ARMv8_1m_main): New feature group.
2622 (armv8.1-m.main): New architecture.
2623 * config/arm/arm-tables.opt: Regenerate.
2624 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2625 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2626 (arm_options_perform_arch_sanity_checks): Error out when targeting
2627 Armv8.1-M Mainline Security Extensions.
2628 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2630 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2632 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2633 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2634 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2635 aarch64_bfdot_laneq): New.
2636 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2637 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2638 vbfdotq_laneq_f32): New.
2639 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2640 VBFMLA_W, VBF): New.
2641 (isquadop): Add V4BF, V8BF.
2643 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2645 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2646 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2647 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2648 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2649 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2650 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2651 usdot_laneq, sudot_lane,sudot_laneq): New.
2652 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2653 (aarch64_<sur>dot_lane): New.
2654 * config/aarch64/arm_neon.h (vusdot_s32): New.
2656 (vusdot_lane_s32): New.
2657 (vsudot_lane_s32): New.
2658 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2659 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2661 2020-01-16 Martin Liska <mliska@suse.cz>
2663 * value-prof.c (dump_histogram_value): Fix
2664 obvious spacing issue.
2666 2020-01-16 Andrew Pinski <apinski@marvell.com>
2668 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2669 !storage_order_barrier_p.
2671 2020-01-16 Andrew Pinski <apinski@marvell.com>
2673 * sched-int.h (_dep): Add unused bit-field field for the padding.
2674 * sched-deps.c (init_dep_1): Init unused field.
2676 2020-01-16 Andrew Pinski <apinski@marvell.com>
2678 * optabs.h (create_expand_operand): Initialize target field also.
2680 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2682 PR tree-optimization/92429
2683 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2684 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2686 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2689 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2691 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2692 aarch64_sve_int_mode to each mode.
2694 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2696 * doc/analyzer.texi (Overview): Add note about
2697 -fdump-ipa-analyzer.
2699 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2701 PR tree-optimization/93231
2702 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2703 input_type is unsigned. Use tree_to_shwi for shift constant.
2704 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2705 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2706 inputs more efficiently.
2708 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2710 * config/i386/i386.md (*movsf_internal): Do not require
2711 SSE2 ISA for alternatives 14 and 15.
2713 2020-01-15 Richard Biener <rguenther@suse.de>
2716 * tree-eh.c (sink_clobbers): If we already visited the destination
2717 block do not defer insertion.
2718 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2719 the purpose of defered insertion.
2721 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2723 * BASE-VER: Bump to 10.0.1.
2725 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2727 PR tree-optimization/93247
2728 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2729 type of the stmt that we're going to vectorize.
2731 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2733 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2734 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2737 2020-01-15 Martin Liska <mliska@suse.cz>
2739 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2740 2 calls of streamer_read_hwi in a function call.
2742 2020-01-15 Richard Biener <rguenther@suse.de>
2744 * alias.c (record_alias_subset): Avoid redundant work when
2745 subset is already recorded.
2747 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2749 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2750 the analyzer options provide CWE identifiers.
2752 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2754 * tree-diagnostic-path.cc (path_summary::event_range::print):
2755 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2756 using get_pure_location.
2758 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2760 PR tree-optimization/93262
2761 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2762 perform head trimming only if the last argument is constant,
2763 either all ones, or larger or equal to head trim, in the latter
2764 case decrease the last argument by head_trim.
2766 PR tree-optimization/93249
2767 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2768 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2769 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2770 perform head trim unless we can prove there are no '\0' chars
2771 from the source among the first head_trim chars.
2773 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2775 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2777 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2780 * config/i386/sse.md
2781 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2782 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2783 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2784 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2785 just a single alternative instead of two, make operands 1 and 2
2788 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2791 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2794 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2796 * Makefile.in (lang_opt_files): Add analyzer.opt.
2797 (ANALYZER_OBJS): New.
2798 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2799 tristate.o and ANALYZER_OBJS.
2800 (TEXI_GCCINT_FILES): Add analyzer.texi.
2801 * common.opt (-fanalyzer): New driver option.
2802 * config.in: Regenerate.
2803 * configure: Regenerate.
2804 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2805 (gccdepdir): Also create depdir for "analyzer" subdir.
2806 * digraph.cc: New file.
2807 * digraph.h: New file.
2808 * doc/analyzer.texi: New file.
2809 * doc/gccint.texi ("Static Analyzer") New menu item.
2810 (analyzer.texi): Include it.
2811 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2812 ("Warning Options"): Add static analysis warnings to the list.
2813 (-Wno-analyzer-double-fclose): New option.
2814 (-Wno-analyzer-double-free): New option.
2815 (-Wno-analyzer-exposure-through-output-file): New option.
2816 (-Wno-analyzer-file-leak): New option.
2817 (-Wno-analyzer-free-of-non-heap): New option.
2818 (-Wno-analyzer-malloc-leak): New option.
2819 (-Wno-analyzer-possible-null-argument): New option.
2820 (-Wno-analyzer-possible-null-dereference): New option.
2821 (-Wno-analyzer-null-argument): New option.
2822 (-Wno-analyzer-null-dereference): New option.
2823 (-Wno-analyzer-stale-setjmp-buffer): New option.
2824 (-Wno-analyzer-tainted-array-index): New option.
2825 (-Wno-analyzer-use-after-free): New option.
2826 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2827 (-Wno-analyzer-use-of-uninitialized-value): New option.
2828 (-Wanalyzer-too-complex): New option.
2829 (-fanalyzer-call-summaries): New warning.
2830 (-fanalyzer-checker=): New warning.
2831 (-fanalyzer-fine-grained): New warning.
2832 (-fno-analyzer-state-merge): New warning.
2833 (-fno-analyzer-state-purge): New warning.
2834 (-fanalyzer-transitivity): New warning.
2835 (-fanalyzer-verbose-edges): New warning.
2836 (-fanalyzer-verbose-state-changes): New warning.
2837 (-fanalyzer-verbosity=): New warning.
2838 (-fdump-analyzer): New warning.
2839 (-fdump-analyzer-callgraph): New warning.
2840 (-fdump-analyzer-exploded-graph): New warning.
2841 (-fdump-analyzer-exploded-nodes): New warning.
2842 (-fdump-analyzer-exploded-nodes-2): New warning.
2843 (-fdump-analyzer-exploded-nodes-3): New warning.
2844 (-fdump-analyzer-supergraph): New warning.
2845 * doc/sourcebuild.texi (dg-require-dot): New.
2846 (dg-check-dot): New.
2847 * gdbinit.in (break-on-saved-diagnostic): New command.
2848 * graphviz.cc: New file.
2849 * graphviz.h: New file.
2850 * ordered-hash-map-tests.cc: New file.
2851 * ordered-hash-map.h: New file.
2852 * passes.def (pass_analyzer): Add before
2853 pass_ipa_whole_program_visibility.
2854 * selftest-run-tests.c (selftest::run_tests): Call
2855 selftest::ordered_hash_map_tests_cc_tests.
2856 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2858 * shortest-paths.h: New file.
2859 * timevar.def (TV_ANALYZER): New timevar.
2860 (TV_ANALYZER_SUPERGRAPH): Likewise.
2861 (TV_ANALYZER_STATE_PURGE): Likewise.
2862 (TV_ANALYZER_PLAN): Likewise.
2863 (TV_ANALYZER_SCC): Likewise.
2864 (TV_ANALYZER_WORKLIST): Likewise.
2865 (TV_ANALYZER_DUMP): Likewise.
2866 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2867 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2868 * tree-pass.h (make_pass_analyzer): New decl.
2869 * tristate.cc: New file.
2870 * tristate.h: New file.
2872 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2875 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2876 alternatives 9 and 10.
2878 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2880 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2881 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2882 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2883 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2884 (selftest::hash_map_tests_c_tests): Call it.
2885 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2886 New static constant, using the value of = H::empty_zero_p.
2887 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2888 from default_hash_traits <Value>.
2889 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2891 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2892 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2893 calls to mark_empty with !Descriptor::empty_zero_p.
2894 (hash_table::empty_slow): Conditionalize the memset call with a
2895 check that Descriptor::empty_zero_p; otherwise, loop through the
2896 entries calling mark_empty on them.
2897 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2898 (pointer_hash::empty_zero_p): Likewise.
2899 (pair_hash::empty_zero_p): Likewise.
2900 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2902 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2903 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2904 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2905 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2906 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2907 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2908 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2909 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2911 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2914 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2916 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2917 fix typo on return value.
2919 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2922 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2924 (cgraph_edge::make_speculative): Add param for setting speculative_id
2926 (cgraph_edge::speculative_call_info): Update comments and find reference
2927 by speculative_id for multiple indirect targets.
2928 (cgraph_edge::resolve_speculation): Decrease the speculations
2929 for indirect edge, drop it's speculative if not direct target
2930 left. Update comments.
2931 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2932 (cgraph_node::dump): Print num_speculative_call_targets.
2933 (cgraph_node::verify_node): Don't report error if speculative
2934 edge not include statement.
2935 (cgraph_edge::num_speculative_call_targets_p): New function.
2936 * cgraph.h (int common_target_id): Remove.
2937 (int common_target_probability): Remove.
2938 (num_speculative_call_targets): New variable.
2939 (make_speculative): Add param for setting speculative_id.
2940 (cgraph_edge::num_speculative_call_targets_p): New declare.
2941 (target_prob): New variable.
2942 (speculative_id): New variable.
2943 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2944 call summaries for multiple speculative call targets.
2945 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2946 * ipa-profile.c (struct speculative_call_target): New struct.
2947 (class speculative_call_summary): New class.
2948 (class speculative_call_summaries): New class.
2949 (call_sums): New variable.
2950 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2951 (ipa_profile_write_edge_summary): New function.
2952 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2953 (ipa_profile_dump_all_summaries): New function.
2954 (ipa_profile_read_edge_summary): New function.
2955 (ipa_profile_read_summary_section): New function.
2956 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2957 (ipa_profile): Generate num_speculative_call_targets from
2959 * ipa-ref.h (speculative_id): New variable.
2960 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2961 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2962 common_target_probability. Stream out speculative_id and
2963 num_speculative_call_targets.
2964 (input_edge): Likewise.
2965 * predict.c (dump_prediction): Remove edges count assert to be
2967 * symtab.c (symtab_node::create_reference): Init speculative_id.
2968 (symtab_node::clone_references): Clone speculative_id.
2969 (symtab_node::clone_referring): Clone speculative_id.
2970 (symtab_node::clone_reference): Clone speculative_id.
2971 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2972 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2973 if indirect call contains multiple speculative targets.
2974 * value-prof.h (check_ic_target): Remove.
2975 * value-prof.c (gimple_value_profile_transformations):
2976 Use void function gimple_ic_transform.
2977 * value-prof.c (gimple_ic_transform): Handle topn case.
2978 Fix comment typos. Change it to a void function.
2980 2020-01-13 Andrew Pinski <apinski@marvell.com>
2982 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2983 (octeontx2t98): New define.
2984 (octeontx2t96): New define.
2985 (octeontx2t93): New define.
2986 (octeontx2f95): New define.
2987 (octeontx2f95n): New define.
2988 (octeontx2f95mm): New define.
2989 * config/aarch64/aarch64-tune.md: Regenerate.
2990 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2992 2020-01-13 Jason Merrill <jason@redhat.com>
2994 PR c++/33799 - destroy return value if local cleanup throws.
2995 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2997 2020-01-13 Martin Liska <mliska@suse.cz>
2999 * ipa-cp.c (get_max_overall_size): Use newly
3000 renamed param param_ipa_cp_unit_growth.
3001 * params.opt: Remove legacy param name.
3003 2020-01-13 Martin Sebor <msebor@redhat.com>
3005 PR tree-optimization/93213
3006 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3007 stores to be eliminated.
3009 2020-01-13 Martin Liska <mliska@suse.cz>
3011 * opts.c (print_help): Do not print CL_PARAM
3012 and CL_WARNING for CL_OPTIMIZATION.
3014 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3017 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3018 depending on optimization settings.
3020 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3022 PR tree-optimization/90838
3023 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3024 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3025 argument rather than to initialize temporary for targets that
3026 don't use the mode argument at all. Initialize ctzval to avoid
3029 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3031 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3032 * tree-core.h: Document it.
3033 * gimplify.c (gimplify_omp_workshare): Set it.
3034 * omp-low.c (lower_omp_target): Use it.
3035 * tree-pretty-print.c (dump_omp_clause): Print it.
3037 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3038 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3040 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3042 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3043 * common.opt (fdiagnostics-path-format=): New option.
3044 (diagnostic_path_format): New enum.
3045 (fdiagnostics-show-path-depths): New option.
3046 * coretypes.h (diagnostic_event_id_t): New forward decl.
3047 * diagnostic-color.c (color_dict): Add "path".
3048 * diagnostic-event-id.h: New file.
3049 * diagnostic-format-json.cc (json_from_expanded_location): Make
3051 (json_end_diagnostic): Call context->make_json_for_path if it
3052 exists and the diagnostic has a path.
3053 (diagnostic_output_format_init): Clear context->print_path.
3054 * diagnostic-path.h: New file.
3055 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3056 when printing a run of events in a diagnostic_path so that they
3057 all get the same color.
3058 (layout::m_diagnostic_path_p): New field.
3059 (layout::layout): Initialize it.
3060 (layout::print_any_labels): Don't colorize the label text for an
3061 event in a diagnostic_path.
3062 (gcc_rich_location::add_location_if_nearby): Add
3063 "restrict_to_current_line_spans" and "label" params. Pass the
3064 former to layout.maybe_add_location_range; pass the latter
3065 when calling add_range.
3066 * diagnostic.c: Include "diagnostic-path.h".
3067 (diagnostic_initialize): Initialize context->path_format and
3068 context->show_path_depths.
3069 (diagnostic_show_any_path): New function.
3070 (diagnostic_path::interprocedural_p): New function.
3071 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3072 (simple_diagnostic_path::num_events): New function.
3073 (simple_diagnostic_path::get_event): New function.
3074 (simple_diagnostic_path::add_event): New function.
3075 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3076 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3077 (debug): New overload taking a diagnostic_path *.
3078 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3079 * diagnostic.h (enum diagnostic_path_format): New enum.
3080 (json::value): New forward decl.
3081 (diagnostic_context::path_format): New field.
3082 (diagnostic_context::show_path_depths): New field.
3083 (diagnostic_context::print_path): New callback field.
3084 (diagnostic_context::make_json_for_path): New callback field.
3085 (diagnostic_show_any_path): New decl.
3086 (json_from_expanded_location): New decl.
3087 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3088 (-fdiagnostics-show-path-depths): New option.
3089 (-fdiagnostics-color): Add "path" to description of default
3090 GCC_COLORS; describe it.
3091 (-fdiagnostics-format=json): Document how diagnostic paths are
3092 represented in the JSON output format.
3093 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3094 Add optional params "restrict_to_current_line_spans" and "label".
3095 * opts.c (common_handle_option): Handle
3096 OPT_fdiagnostics_path_format_ and
3097 OPT_fdiagnostics_show_path_depths.
3098 * pretty-print.c: Include "diagnostic-event-id.h".
3099 (pp_format): Implement "%@" format code for printing
3100 diagnostic_event_id_t *.
3101 (selftest::test_pp_format): Add tests for "%@".
3102 * selftest-run-tests.c (selftest::run_tests): Call
3103 selftest::tree_diagnostic_path_cc_tests.
3104 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3105 * toplev.c (general_init): Initialize global_dc->path_format and
3106 global_dc->show_path_depths.
3107 * tree-diagnostic-path.cc: New file.
3108 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3109 non-static. Drop "diagnostic" param in favor of storing the
3110 original value of "where" and re-using it.
3111 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3112 maybe_unwind_expanded_macro_loc.
3113 (tree_diagnostics_defaults): Initialize context->print_path and
3114 context->make_json_for_path.
3115 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3117 (default_tree_make_json_for_path): New decl.
3118 (maybe_unwind_expanded_macro_loc): New decl.
3120 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3122 PR tree-optimization/93210
3123 * fold-const.h (native_encode_initializer,
3124 can_native_interpret_type_p): Declare.
3125 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3127 (native_encode_initializer): New function, moved from dwarf2out.c.
3128 Adjust to native_encode_expr compatible arguments, including dry-run
3129 and partial extraction modes. Don't handle STRING_CST.
3130 (can_native_interpret_type_p): No longer static.
3131 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3132 offset / BITS_PER_UNIT fits into int and don't call it if
3133 can_native_interpret_type_p fails. If suboff is NULL and for
3134 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3135 native_encode_initializer.
3136 (fold_const_aggregate_ref_1): Formatting fix.
3137 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3138 (tree_add_const_value_attribute): Adjust caller.
3140 PR tree-optimization/90838
3141 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3142 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3143 CTZ_DEFINED_VALUE_AT_ZERO.
3145 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3148 * lra-constraints.c (match_reload): Permit input operands have the
3149 same mode as output while other input operands have a different
3152 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3154 PR tree-optimization/90838
3155 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3156 (check_ctz_string): Likewise.
3157 (optimize_count_trailing_zeroes): Likewise.
3158 (simplify_count_trailing_zeroes): Likewise.
3159 (pass_forwprop::execute): Try ctz simplification.
3160 * match.pd: Add matching for ctz idioms.
3162 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3164 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3166 (aarch64_invalid_unary_op): New function for target hook.
3167 (aarch64_invalid_binary_op): New function for target hook.
3169 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3171 * config.gcc: Add arm_bf16.h.
3172 * config/aarch64/aarch64-builtins.c
3173 (aarch64_simd_builtin_std_type): Add BFmode.
3174 (aarch64_init_simd_builtin_types): Define element types for vector
3176 (aarch64_init_bf16_types): New function.
3177 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3178 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3180 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3181 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3183 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3184 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3185 * config/aarch64/aarch64.c
3186 (aarch64_classify_vector_mode): Add support for BF types.
3187 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3188 (aarch64_vq_mode): Add support for BF types.
3189 (aarch64_simd_container_mode): Add support for BF types.
3190 (aarch64_mangle_type): Add support for BF scalar type.
3191 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3192 * config/aarch64/arm_bf16.h: New file.
3193 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3194 * config/aarch64/iterators.md: Add BF types to mode attributes.
3195 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3197 2020-01-10 Jason Merrill <jason@redhat.com>
3199 PR c++/93173 - incorrect tree sharing.
3200 * gimplify.c (copy_if_shared): No longer static.
3201 * gimplify.h: Declare it.
3203 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3205 * doc/invoke.texi (-msve-vector-bits=): Document that
3206 -msve-vector-bits=128 now generates VL-specific code for
3207 little-endian targets.
3208 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3209 build_vector_type_for_mode to construct the data vector types.
3210 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3211 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3212 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3213 for 128-bit vectors.
3215 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3217 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3220 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3222 * config/aarch64/aarch64-builtins.c
3223 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3224 rather than checking the number of elements and the element mode.
3226 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3228 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3229 get_related_vectype_for_scalar_type rather than build_vector_type
3230 to create the index type for a conditional reduction.
3232 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3234 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3235 for any type of gather or scatter, including strided accesses.
3237 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3239 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3242 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3244 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3246 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3247 parameter and its use to reset DR_OFFSET's.
3248 (vect_transform_loop): Remove orig_drs_init argument.
3249 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3250 member of dr_vec_info rather than the offset of the associated
3251 data_reference's innermost_loop_behavior.
3252 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3253 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3254 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3255 get_dr_vinfo_offset.
3256 (vectorizable_store): Likewise.
3257 (vectorizable_load): Likewise.
3259 2020-01-10 Richard Biener <rguenther@suse.de>
3261 * gimple-ssa-store-merging
3262 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3264 2020-01-10 Martin Liska <mliska@suse.cz>
3267 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3268 encapsulation that was there before r280040.
3270 2020-01-10 Richard Biener <rguenther@suse.de>
3273 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3274 sequences to avoid walking them again for secondary opportunities.
3275 (pass_lower_eh_dispatch::execute): Instead actually insert
3278 2020-01-10 Richard Biener <rguenther@suse.de>
3281 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3282 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3283 avoid quadraticness.
3285 2020-01-10 Martin Jambor <mjambor@suse.cz>
3287 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3288 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3289 to get param_ipa_sra_max_replacements.
3290 (param_splitting_across_edge): Pass the caller to
3291 pull_accesses_from_callee.
3293 2020-01-10 Martin Jambor <mjambor@suse.cz>
3295 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3296 * ipa-cp.c (max_new_size): Removed.
3297 (orig_overall_size): New variable.
3298 (get_max_overall_size): New function.
3299 (estimate_local_effects): Use it. Adjust dump.
3300 (decide_about_value): Likewise.
3301 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3302 orig_overall_size. Adjust dump.
3303 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3305 2020-01-10 Martin Jambor <mjambor@suse.cz>
3307 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3308 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3309 instead of param_ipa_max_agg_items.
3310 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3311 optimization info for the callee.
3313 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3315 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3316 markers if debug_inline_points is false.
3318 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3320 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3322 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3323 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3324 aarch64-sve-builtins-sve2.h.
3325 (aarch64-sve-builtins-sve2.o): New rule.
3326 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3327 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3328 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3329 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3330 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3331 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3333 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3334 instructions that are handled here.
3335 (@cond_asrd<mode>): Generalize to...
3336 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3337 (*cond_asrd<mode>_2): Generalize to...
3338 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3339 (*cond_asrd<mode>_z): Generalize to...
3340 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3341 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3342 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3343 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3344 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3346 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3347 (@aarch64_scatter_stnt<mode>): Likewise.
3348 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3349 (@aarch64_mul_lane_<mode>): Likewise.
3350 (@aarch64_sve_suqadd<mode>_const): Likewise.
3351 (*<sur>h<addsub><mode>): Generalize to...
3352 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3354 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3355 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3356 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3357 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3358 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3359 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3360 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3361 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3362 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3363 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3364 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3365 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3366 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3367 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3368 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3369 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3370 (@aarch64_sve2_xar<mode>): Likewise.
3371 (@aarch64_sve2_bcax<mode>): Likewise.
3372 (*aarch64_sve2_eor3<mode>): Rename to...
3373 (@aarch64_sve2_eor3<mode>): ...this.
3374 (@aarch64_sve2_bsl<mode>): New expander.
3375 (@aarch64_sve2_nbsl<mode>): Likewise.
3376 (@aarch64_sve2_bsl1n<mode>): Likewise.
3377 (@aarch64_sve2_bsl2n<mode>): Likewise.
3378 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3379 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3380 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3381 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3382 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3383 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3384 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3385 (<su>mull<bt><Vwide>): Generalize to...
3386 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3388 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3389 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3390 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3391 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3392 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3393 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3394 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3395 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3396 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3397 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3398 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3399 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3400 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3401 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3402 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3403 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3404 (<SHRNB:r>shrnb<mode>): Generalize to...
3405 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3407 (<SHRNT:r>shrnt<mode>): Generalize to...
3408 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3410 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3411 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3412 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3413 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3414 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3415 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3416 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3417 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3418 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3419 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3420 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3421 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3422 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3423 (@aarch64_sve2_cvtnt<mode>): Likewise.
3424 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3425 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3426 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3427 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3428 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3429 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3430 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3431 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3432 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3433 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3434 (@aarch64_sve2_pmul<mode>): Likewise.
3435 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3436 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3437 (@aarch64_sve2_tbl2<mode>): Likewise.
3438 (@aarch64_sve2_tbx<mode>): Likewise.
3439 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3440 (@aarch64_sve2_histcnt<mode>): Likewise.
3441 (@aarch64_sve2_histseg<mode>): Likewise.
3442 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3443 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3444 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3445 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3446 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3447 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3448 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3449 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3450 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3451 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3452 (SVE2_PMULL_PAIR_I): New mode iterators.
3453 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3454 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3455 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3456 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3457 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3458 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3459 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3460 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3461 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3462 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3463 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3464 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3465 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3466 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3467 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3468 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3469 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3470 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3471 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3472 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3473 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3474 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3475 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3476 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3477 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3478 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3479 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3480 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3481 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3482 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3483 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3485 (VNARROW, Ventype): New mode attributes.
3486 (Vewtype): Handle VNx2DI. Fix typo in comment.
3487 (VDOUBLE): New mode attribute.
3488 (sve_lane_con): Handle VNx8HI.
3489 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3490 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3491 (sve_int_op, sve_int_op_rev): Handle the above codes.
3492 (sve_pred_int_rhs2_operand): Likewise.
3493 (MULLBT, SHRNB, SHRNT): Delete.
3494 (SVE_INT_SHIFT_IMM): New int iterator.
3495 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3496 and UNSPEC_WHILEHS for TARGET_SVE2.
3497 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3498 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3499 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3500 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3501 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3502 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3503 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3504 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3505 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3506 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3507 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3508 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3509 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3510 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3511 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3512 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3513 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3514 (optab): Handle the new unspecs.
3515 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3517 (lr): Handle the new unspecs.
3519 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3520 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3521 (sve_int_qsub_op): New int attributes.
3522 (sve_fp_op, rot): Handle the new unspecs.
3523 * config/aarch64/aarch64-sve-builtins.h
3524 (function_resolver::require_matching_pointer_type): Declare.
3525 (function_resolver::resolve_unary): Add an optional boolean argument.
3526 (function_resolver::finish_opt_n_resolution): Add an optional
3527 type_suffix_index argument.
3528 (gimple_folder::redirect_call): Declare.
3529 (gimple_expander::prepare_gather_address_operands): Add an optional
3531 * config/aarch64/aarch64-sve-builtins.cc: Include
3532 aarch64-sve-builtins-sve2.h.
3533 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3534 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3535 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3536 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3537 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3538 (TYPES_s_unsigned): Likewise.
3539 (TYPES_s_integer): Use TYPES_s_unsigned.
3540 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3541 (TYPES_sd_integer): Use them.
3542 (TYPES_d_unsigned): New macro.
3543 (TYPES_d_integer): Use it.
3544 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3545 (TYPES_cvt_narrow): Likewise.
3546 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3547 (preds_mx): New variable.
3548 (function_builder::add_overloaded_function): Allow the new feature
3549 set to be more restrictive than the original one.
3550 (function_resolver::infer_pointer_type): Remove qualifiers from
3551 the pointer type before printing it.
3552 (function_resolver::require_matching_pointer_type): New function.
3553 (function_resolver::resolve_sv_displacement): Handle functions
3554 that don't support 32-bit vector indices or svint32_t vector offsets.
3555 (function_resolver::finish_opt_n_resolution): Take the inferred type
3556 as a separate argument.
3557 (function_resolver::resolve_unary): Optionally treat all forms in
3558 the same way as normal merging functions.
3559 (gimple_folder::redirect_call): New function.
3560 (function_expander::prepare_gather_address_operands): Add an argument
3561 that says whether scaled forms are available. If they aren't,
3562 handle scaling of vector indices and don't add the extension and
3564 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3565 fall back to using cond_* instead.
3566 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3567 Split out the member variables into...
3568 (rtx_code_function_base): ...this new base class.
3569 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3570 (unspec_based_function): Split out the member variables into...
3571 (unspec_based_function_base): ...this new base class.
3572 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3573 (unspec_based_function_exact_insn): New class.
3574 (unspec_based_add_function, unspec_based_add_lane_function)
3575 (unspec_based_lane_function, unspec_based_pred_function)
3576 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3577 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3578 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3580 (unspec_based_fused_function): New class.
3581 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3582 (unspec_based_fused_lane_function): New class.
3583 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3585 (CODE_FOR_MODE1): New macro.
3586 (fixed_insn_function): New class.
3587 (while_comparison): Likewise.
3588 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3589 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3590 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3591 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3592 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3593 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3594 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3595 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3596 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3597 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3598 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3599 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3600 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3601 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3602 Also add an initial argument for unary_convert_narrowt, regardless
3603 of the predication type.
3604 (build_32_64): Allow loads and stores to specify MODE_none.
3605 (build_sv_index64, build_sv_uint_offset): New functions.
3606 (long_type_suffix): New function.
3607 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3608 (binary_imm_long_base, load_gather_sv_base): Likewise.
3609 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3610 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3611 (unary_narrowb_base, unary_narrowt_base): Likewise.
3612 (binary_long_lane_def, binary_long_lane): New shape.
3613 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3614 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3615 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3616 (binary_to_uint_def, binary_to_uint): Likewise.
3617 (binary_wide_def, binary_wide): Likewise.
3618 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3619 (compare_def, compare): Likewise.
3620 (compare_ptr_def, compare_ptr): Likewise.
3621 (load_ext_gather_index_restricted_def,
3622 load_ext_gather_index_restricted): Likewise.
3623 (load_ext_gather_offset_restricted_def,
3624 load_ext_gather_offset_restricted): Likewise.
3625 (load_gather_sv_def): Inherit from load_gather_sv_base.
3626 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3627 (shift_left_imm_def, shift_left_imm): Likewise.
3628 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3629 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3630 (store_scatter_index_restricted_def,
3631 store_scatter_index_restricted): Likewise.
3632 (store_scatter_offset_restricted_def,
3633 store_scatter_offset_restricted): Likewise.
3634 (tbl_tuple_def, tbl_tuple): Likewise.
3635 (ternary_long_lane_def, ternary_long_lane): Likewise.
3636 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3637 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3638 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3639 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3640 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3641 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3642 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3643 (ternary_uint_def, ternary_uint): Likewise.
3644 (unary_convert): Fix typo in comment.
3645 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3646 (unary_long_def, unary_long): Likewise.
3647 (unary_narrowb_def, unary_narrowb): Likewise.
3648 (unary_narrowt_def, unary_narrowt): Likewise.
3649 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3650 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3651 (unary_to_int_def, unary_to_int): Likewise.
3652 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3653 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3654 (svasrd_impl): Delete.
3655 (svcadd_impl::expand): Handle integer operations too.
3656 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3657 new functions to derive the unspec numbers.
3658 (svmla_svmls_lane_impl): Replace with...
3659 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3660 integer operations too.
3661 (svwhile_impl): Rename to...
3662 (svwhilelx_impl): ...this and inherit from while_comparison.
3663 (svasrd): Use unspec_based_function.
3664 (svmla_lane): Use svmla_lane_impl.
3665 (svmls_lane): Use svmls_lane_impl.
3666 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3667 (svwhilele, svwhilelt): Use svwhilelx_impl.
3668 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3669 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3670 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3671 * config/aarch64/aarch64-sve-builtins.def: Include
3672 aarch64-sve-builtins-sve2.def.
3674 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3676 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3677 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3678 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3679 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3680 immediates as well as vector ones.
3681 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3682 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3683 (aarch64_sve_qsub_immediate): Update calls accordingly.
3685 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3687 * config/aarch64/aarch64-sve2.md: Add banner comments.
3688 (<su>mulh<r>s<mode>3): Move further up file.
3689 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3690 (*aarch64_sve2_sra<mode>): Move further down file.
3691 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3693 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3695 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3697 (while_optab_cmp): Handle them.
3698 * config/aarch64/aarch64-sve.md
3699 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3700 and add a "@" marker.
3701 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3702 instead of gen_aarch64_sve2_while_ptest.
3703 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3705 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3707 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3708 (UNSPEC_WHILELE): ...this.
3709 (UNSPEC_WHILE_LO): Rename to...
3710 (UNSPEC_WHILELO): ...this.
3711 (UNSPEC_WHILE_LS): Rename to...
3712 (UNSPEC_WHILELS): ...this.
3713 (UNSPEC_WHILE_LT): Rename to...
3714 (UNSPEC_WHILELT): ...this.
3715 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3716 (cmp_op, while_optab_cmp): Likewise.
3717 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3718 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3719 (svwhilelt): Likewise.
3721 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3723 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3724 (unary_to_uint): Define.
3725 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3726 (unary_count): Rename to...
3727 (unary_to_uint_def, unary_to_uint): ...this.
3728 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3730 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3732 * config/aarch64/aarch64-sve-builtins-functions.h
3733 (code_for_mode_function): New class.
3734 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3735 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3736 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3737 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3738 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3740 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3742 * config/aarch64/iterators.md (addsub): New code attribute.
3743 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3745 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3746 in the asm string and attributes. Fix indentation.
3747 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3749 (@aarch64_sve_<optab><mode>): ...this.
3750 * config/aarch64/aarch64-sve-builtins.h
3751 (function_expander::expand_signed_unpred_op): Delete.
3752 * config/aarch64/aarch64-sve-builtins.cc
3753 (function_expander::expand_signed_unpred_op): Likewise.
3754 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3755 try using code_for_aarch64_sve instead.
3756 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3757 (svqsub_impl): Likewise.
3758 (svqadd, svqsub): Use rtx_code_function instead.
3760 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3762 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3763 (HADDSUB, sur, addsub): Remove them.
3765 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3767 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3768 functions the same way, rather than singling out those that
3769 aren't mapped directly to optabs.
3771 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3773 * target.def (compatible_vector_types_p): New target hook.
3774 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3775 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3776 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3777 * doc/tm.texi: Regenerate.
3778 * gimple-expr.c: Include target.h.
3779 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3780 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3782 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3783 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3784 Use the original predicate if it already has a suitable type.
3786 2020-01-09 Martin Jambor <mjambor@suse.cz>
3788 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3789 resolve_speculation and redirect_call_stmt_to_callee static. Change
3790 return type of set_call_stmt to cgraph_edge *.
3791 * auto-profile.c (afdo_indirect_call): Adjust call to
3792 redirect_call_stmt_to_callee.
3793 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3794 make the this pointer explicit, adjust self-recursive calls and the
3795 call top make_direct. Return the resulting edge.
3796 (cgraph_edge::remove): Make this pointer explicit.
3797 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3798 (cgraph_edge::make_direct): Likewise, adjust call to
3799 resolve_speculation.
3800 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3801 call to set_call_stmt.
3802 (cgraph_update_edges_for_call_stmt_node): Update call to
3803 set_call_stmt and remove.
3804 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3805 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3806 (cgraph_node::create_edge_including_clones): Moved "first" definition
3807 of edge to the block where it was used. Adjusted calls to
3809 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3810 cgraph_edge::remove.
3811 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3812 make_direct and redirect_call_stmt_to_callee.
3813 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3814 resolve_speculation and make_direct.
3815 * ipa-inline-transform.c (inline_transform): Adjust call to
3816 redirect_call_stmt_to_callee.
3817 (check_speculations_1):: Adjust call to resolve_speculation.
3818 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3819 resolve-speculation.
3820 (inline_small_functions): Adjust call to resolve_speculation.
3821 (ipa_inline): Likewise.
3822 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3824 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3825 safe with regards to edge removal, adjust calls to
3826 redirect_call_stmt_to_callee.
3827 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3828 and redirect_call_stmt_to_callee.
3829 * multiple_target.c (create_dispatcher_calls): Adjust call to
3830 redirect_call_stmt_to_callee
3831 (redirect_to_specific_clone): Likewise.
3832 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3833 Adjust calls to cgraph_edge::remove.
3834 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3835 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3836 (expand_call_inline): Adjust call to cgraph_edge::remove.
3838 2020-01-09 Martin Liska <mliska@suse.cz>
3840 * params.opt: Set Optimization for
3841 param_max_speculative_devirt_maydefs.
3843 2020-01-09 Martin Sebor <msebor@redhat.com>
3847 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3849 2020-01-09 Martin Liska <mliska@suse.cz>
3851 * auto-profile.c (auto_profile): Use opt_for_fn
3853 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3854 (propagate_vals_across_arith_jfunc): Likewise.
3855 (hint_time_bonus): Likewise.
3856 (incorporate_penalties): Likewise.
3857 (good_cloning_opportunity_p): Likewise.
3858 (perform_estimation_of_a_value): Likewise.
3859 (estimate_local_effects): Likewise.
3860 (ipcp_propagate_stage): Likewise.
3861 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3862 (set_switch_stmt_execution_predicate): Likewise.
3863 (analyze_function_body): Likewise.
3864 * ipa-inline-analysis.c (offline_size): Likewise.
3865 * ipa-inline.c (early_inliner): Likewise.
3866 * ipa-prop.c (ipa_analyze_node): Likewise.
3867 (ipcp_transform_function): Likewise.
3868 * ipa-sra.c (process_scan_results): Likewise.
3869 (ipa_sra_summarize_function): Likewise.
3870 * params.opt: Rename ipcp-unit-growth to
3871 ipa-cp-unit-growth. Add Optimization for various
3872 IPA-related parameters.
3874 2020-01-09 Richard Biener <rguenther@suse.de>
3877 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3879 2020-01-09 Richard Biener <rguenther@suse.de>
3881 PR tree-optimization/93040
3882 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3884 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3886 * common/config/avr/avr-common.c (avr_option_optimization_table)
3887 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3889 2020-01-09 Martin Liska <mliska@suse.cz>
3891 * cgraphclones.c (symbol_table::materialize_all_clones):
3892 Use cgraph_node::dump_name.
3894 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3897 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3898 output_operand_lossage instead of gcc_unreachable.
3899 * doc/md.texi (riscv f constraint): Fix typo.
3902 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3903 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3904 CONST_SCALAR_INT_P instead of CONST_INT_P.
3905 (*subv<mode>4_1): Rename to ...
3906 (subv<mode>4_1): ... this.
3907 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3908 define_insn_and_split patterns.
3909 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3912 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3914 * vec.c (class selftest::count_dtor): New class.
3915 (selftest::test_auto_delete_vec): New test.
3916 (selftest::vec_c_tests): Call it.
3917 * vec.h (class auto_delete_vec): New class template.
3918 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3920 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3922 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3924 2020-01-08 Jim Wilson <jimw@sifive.com>
3926 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3927 use of TLS_MODEL_LOCAL_EXEC when not pic.
3929 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3931 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3934 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3937 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3938 *stack_protect_set_3 peephole2): Also check that the second
3939 insns source is general_operand.
3942 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3943 predicate for output operand instead of register_operand.
3944 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3945 memory destination and non-memory operands[2].
3947 2020-01-08 Martin Liska <mliska@suse.cz>
3949 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3950 ::dump_asm_name instead of (::name or ::asm_name).
3951 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3952 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3953 (analyze_functions): Likewise.
3954 (expand_all_functions): Likewise.
3955 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3956 (propagate_bits_across_jump_function): Likewise.
3957 (dump_profile_updates): Likewise.
3958 (ipcp_store_bits_results): Likewise.
3959 (ipcp_store_vr_results): Likewise.
3960 * ipa-devirt.c (dump_targets): Likewise.
3961 * ipa-fnsummary.c (analyze_function_body): Likewise.
3962 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3963 (process_hsa_functions): Likewise.
3964 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3965 (set_alias_uids): Likewise.
3966 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3967 * ipa-inline.c (recursive_inlining): Likewise.
3968 (inline_to_all_callers_1): Likewise.
3969 (ipa_inline): Likewise.
3970 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3971 (ipa_propagate_frequency): Likewise.
3972 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3973 (remove_described_reference): Likewise.
3974 * ipa-pure-const.c (worse_state): Likewise.
3975 (check_retval_uses): Likewise.
3976 (analyze_function): Likewise.
3977 (propagate_pure_const): Likewise.
3978 (propagate_nothrow): Likewise.
3979 (dump_malloc_lattice): Likewise.
3980 (propagate_malloc): Likewise.
3981 (pass_local_pure_const::execute): Likewise.
3982 * ipa-visibility.c (optimize_weakref): Likewise.
3983 (function_and_variable_visibility): Likewise.
3984 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3985 (ipa_discover_variable_flags): Likewise.
3986 * lto-streamer-out.c (output_function): Likewise.
3987 (output_constructor): Likewise.
3988 * tree-inline.c (copy_bb): Likewise.
3989 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3990 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3992 2020-01-08 Richard Biener <rguenther@suse.de>
3995 * tree-eh.c (sink_clobbers): Update virtual operands for
3996 the first and last stmt only. Add a dry-run capability.
3997 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3998 after CFG manipulations and in RPO order to catch all
3999 secondary opportunities reliably.
4001 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4004 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4006 2019-01-08 Richard Biener <rguenther@suse.de>
4009 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4010 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4011 virtual operand, also updating SSA use.
4012 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4013 Update stmt after resetting virtual operand.
4014 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4015 * gimple-iterator.c (gsi_remove): When not removing the stmt
4016 permanently do not delink immediate uses or mark the stmt modified.
4018 2020-01-08 Martin Liska <mliska@suse.cz>
4020 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4021 (ipa_call_context::estimate_size_and_time): Likewise.
4022 (inline_analyze_function): Likewise.
4024 2020-01-08 Martin Liska <mliska@suse.cz>
4026 * cgraph.c (cgraph_node::dump): Use systematically
4029 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4031 Add -nodevicespecs option for avr.
4034 * config/avr/avr.opt (-nodevicespecs): New driver option.
4035 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4036 "-specs=device-specs/..." if that option is not set.
4037 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4039 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4041 Implement 64-bit double functions for avr.
4044 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4045 --with-double-comparison.
4046 * doc/install.texi: Document them.
4047 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4048 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4049 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4050 * doc/invoke.texi (AVR Built-in Macros): Document them.
4051 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4052 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4053 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4055 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4058 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4059 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4060 when only building rm-profile multilibs.
4062 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4065 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4066 lattice for a value to check.
4067 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4068 finite propagation in self-recursive scc.
4070 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4072 * ipa-inline.c (caller_growth_limits): Restore the AND.
4074 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4076 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4077 (VEC_ALLREG_ALT): New iterator.
4078 (VEC_ALLREG_INT_MODE): New iterator.
4079 (VCMP_MODE): New iterator.
4080 (VCMP_MODE_INT): New iterator.
4081 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4082 (vec_cmp<u>v64qidi): New define_expand.
4083 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4084 (vec_cmpu<mode>di_exec): New define_expand.
4085 (vec_cmp<u>v64qidi_exec): New define_expand.
4086 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4087 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4088 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4089 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4090 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4091 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4092 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4093 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4094 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4095 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4097 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4098 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4100 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4102 * config/gcn/constraints.md (DA): Update description and match.
4104 (Db): New constraint.
4105 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4107 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4108 Implement 'Db' mixed immediate type.
4109 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4110 (addcv64si3_dup<exec_vcc>): Delete.
4111 (subcv64si3<exec_vcc>): Rework constraints.
4112 (addv64di3): Rework constraints.
4113 (addv64di3_exec): Rework constraints.
4114 (subv64di3): Rework constraints.
4115 (addv64di3_dup): Delete.
4116 (addv64di3_dup_exec): Delete.
4117 (addv64di3_zext): Rework constraints.
4118 (addv64di3_zext_exec): Rework constraints.
4119 (addv64di3_zext_dup): Rework constraints.
4120 (addv64di3_zext_dup_exec): Rework constraints.
4121 (addv64di3_zext_dup2): Rework constraints.
4122 (addv64di3_zext_dup2_exec): Rework constraints.
4123 (addv64di3_sext_dup2): Rework constraints.
4124 (addv64di3_sext_dup2_exec): Rework constraints.
4126 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4128 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4129 existing target checks.
4131 2020-01-07 Richard Biener <rguenther@suse.de>
4133 * doc/install.texi: Bump minimal supported MPC version.
4135 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4137 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4138 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4139 * langhooks.c: Include stor-layout.h.
4140 (lhd_simulate_enum_decl): New function.
4141 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4142 handle_arm_sve_h for the LTO frontend.
4143 (register_vector_type): Cope with null returns from pushdecl.
4145 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4147 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4148 (aarch64_sve::nvectors_if_data_type): Replace with...
4149 (aarch64_sve::builtin_type_p): ...this.
4150 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4151 (find_vector_type): Delete.
4152 (add_sve_type_attribute): New function.
4153 (lookup_sve_type_attribute): Likewise.
4154 (register_builtin_types): Add an "SVE type" attribute to each type.
4155 (register_tuple_type): Likewise.
4156 (svbool_type_p, nvectors_if_data_type): Delete.
4157 (mangle_builtin_type): Use lookup_sve_type_attribute.
4158 (builtin_type_p): Likewise. Add an overload that returns the
4159 number of constituent vector and predicate registers.
4160 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4161 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4162 instead of aarch64_sve_argument_p.
4163 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4164 (aarch64_pass_by_reference): Likewise.
4165 (aarch64_function_value_1): Likewise.
4166 (aarch64_return_in_memory): Likewise.
4167 (aarch64_layout_arg): Likewise.
4169 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4171 PR tree-optimization/93156
4172 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4173 least significant bit is always clear.
4175 PR tree-optimization/93118
4176 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4177 simplifier with two intermediate conversions.
4179 2020-01-07 Martin Liska <mliska@suse.cz>
4181 * params.opt: Add Optimization for various parameters.
4183 2020-01-07 Martin Liska <mliska@suse.cz>
4186 * doc/extend.texi: Explain cloning for target_clone
4189 2020-01-07 Martin Liska <mliska@suse.cz>
4191 PR tree-optimization/92860
4192 * common.opt: Make in Optimization option
4193 as it is affected by -O0, which is an Optimization
4195 * tree-inline.c (tree_inlinable_function_p):
4196 Use opt_for_fn for warn_inline.
4197 (expand_call_inline): Likewise.
4199 2020-01-07 Martin Liska <mliska@suse.cz>
4201 PR tree-optimization/92860
4202 * common.opt: Make flag_ree as optimization
4205 2020-01-07 Martin Liska <mliska@suse.cz>
4207 PR optimization/92860
4208 * params.opt: Mark param_min_crossjump_insns with Optimization
4211 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4213 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4214 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4216 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4218 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4219 helper function to return the valid addressing formats for a given
4220 hard register and mode.
4221 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4223 * config/rs6000/constraints.md (Q constraint): Update
4225 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4228 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4229 Use 'Q' for doing vector extract from memory.
4230 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4232 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4233 doing vector extract from memory.
4234 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4235 extract from memory.
4237 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4238 for the offset being 34-bits when -mcpu=future is used.
4240 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4242 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4243 instead of cmpib_comparison_operator in cmpib patterns.
4244 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4245 of cmpib_comparison_operator. Revise comment.
4247 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4249 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4250 in an IFN_DIV_POW2 node to be equal.
4252 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4254 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4255 (vect_check_scalar_mask): ...this.
4256 (vectorizable_store, vectorizable_load): Update call accordingly.
4257 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4258 argument in calls to conditional internal functions.
4260 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4262 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4263 '0' matching inputs.
4264 (subv64di3_exec): Likewise.
4266 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4268 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4269 * doc/md.texi (movstr): Likewise.
4271 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4273 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4276 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4278 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4280 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4281 to a temporary file and use move-if-change to update the real
4282 file where necessary.
4284 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4286 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4287 rather than Upa for CPY /M.
4289 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4291 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4294 2020-01-06 Martin Liska <mliska@suse.cz>
4296 PR tree-optimization/92860
4297 * params.opt: Mark param_max_combine_insns with Optimization
4300 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4303 * config/i386/i386.md (SWIDWI): New mode iterator.
4304 (DWI, dwi): Add TImode variants.
4305 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4306 <general_hilo_operand> instead of <general_operand>. Use
4307 CONST_SCALAR_INT_P instead of CONST_INT_P.
4308 (*addv<mode>4_1): Rename to ...
4309 (addv<mode>4_1): ... this.
4310 (QWI): New mode attribute.
4311 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4312 define_insn_and_split patterns.
4313 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4315 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4316 <general_hilo_operand> instead of <general_operand>.
4317 (*addcarry<mode>_1): New define_insn.
4318 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4320 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4322 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4323 Use "call" instead of "set".
4325 2020-01-03 Martin Jambor <mjambor@suse.cz>
4328 * ipa-cp.c (print_all_lattices): Skip functions without info.
4330 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4333 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4334 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4335 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4336 for 'e' simd clones.
4339 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4341 (mprefer-vector-width=): Add Save.
4342 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4343 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4344 (ix86_debug_options, ix86_function_specific_print): Adjust
4345 ix86_target_string callers.
4346 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4347 (ix86_valid_target_attribute_tree): Likewise.
4348 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4349 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4350 ix86_target_string caller.
4353 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4354 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4355 instead of gen_int_shift_amount + convert_modes.
4357 PR rtl-optimization/93088
4358 * loop-iv.c (find_single_def_src): Punt after looking through
4359 128 reg copies for regs with single definitions. Move definitions
4362 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4364 * config/arm/arm-c.c (arm_cpu_builtins): Define
4365 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4366 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4367 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4368 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4369 * config/arm/arm-tables.opt: Regenerated.
4370 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4371 arm_arch_i8mm and arm_arch_bf16 when enabled.
4372 * config/arm/arm.h (TARGET_I8MM): New macro.
4373 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4374 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4375 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4376 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4377 (v8_6_a_simd_variants): New.
4378 (v8_*_a_simd_variants): Add i8mm and bf16.
4379 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4381 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4384 * predict.c (compute_function_frequency): Don't call
4385 warn_function_cold on functions that already have cold attribute.
4387 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4390 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4391 COMDAT group function labels in .data.rel.ro.local section.
4392 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4395 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4396 comparison_operator in B and S integer comparisons. Likewise, use
4397 ordered_comparison_operator instead of cmpib_comparison_operator in
4399 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4401 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4403 Update copyright years.
4405 * gcc.c (process_command): Update copyright notice dates.
4406 * gcov-dump.c (print_version): Ditto.
4407 * gcov.c (print_version): Ditto.
4408 * gcov-tool.c (print_version): Ditto.
4409 * gengtype.c (create_file): Ditto.
4410 * doc/cpp.texi: Bump @copying's copyright year.
4411 * doc/cppinternals.texi: Ditto.
4412 * doc/gcc.texi: Ditto.
4413 * doc/gccint.texi: Ditto.
4414 * doc/gcov.texi: Ditto.
4415 * doc/install.texi: Ditto.
4416 * doc/invoke.texi: Ditto.
4418 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4420 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4423 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4425 PR tree-optimization/93098
4426 * match.pd (popcount): For shift amounts, use integer_onep
4427 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4428 tests. Make sure that precision is power of two larger than or equal
4429 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4430 instead of ULL suffixed constants. Formatting fixes.
4432 Copyright (C) 2020 Free Software Foundation, Inc.
4434 Copying and distribution of this file, with or without modification,
4435 are permitted in any medium without royalty provided the copyright
4436 notice and this notice are preserved.