1 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
4 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
7 2020-02-20 Martin Liska <mliska@suse.cz>
10 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
12 2020-02-20 Martin Liska <mliska@suse.cz>
15 * common/config/avr/avr-common.c: Remote trailing "|".
17 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
19 * collect2.c (maybe_run_lto_and_relink): Fix typo in
22 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
24 PR tree-optimization/93767
25 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
26 access-size bias from the offset calculations for negative strides.
28 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
30 * collect2.c (c_file, o_file): Make const again.
31 (ldout,lderrout, dump_ld_file): Remove.
32 (tool_cleanup): Avoid calling not signal-safe functions.
33 (maybe_run_lto_and_relink): Avoid possible signal handler
34 access to unintialzed memory (lto_o_files).
35 (main): Avoid leaking temp files in $TMPDIR.
36 Initialize c_file/o_file with concat, which avoids exposing
37 uninitialized memory to signal handler, which calls unlink(!).
38 Avoid calling maybe_unlink when the main function returns,
39 since the atexit handler is already doing this.
40 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
42 2020-02-19 Martin Jambor <mjambor@suse.cz>
44 PR tree-optimization/93776
45 * tree-sra.c (create_access): Do not create zero size accesses.
46 (get_access_for_expr): Do not search for zero sized accesses.
48 2020-02-19 Martin Jambor <mjambor@suse.cz>
50 PR tree-optimization/93667
51 * tree-sra.c (scalarizable_type_p): Return false if record fields
52 do not follow wach other.
54 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
56 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
57 rather than fmv.x.s/fmv.s.x.
59 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
61 * config/aarch64/aarch64-simd-builtins.def
62 (intrinsic_vec_smult_lo_): New.
63 (intrinsic_vec_umult_lo_): Likewise.
64 (vec_widen_smult_hi_): Likewise.
65 (vec_widen_umult_hi_): Likewise.
66 * config/aarch64/aarch64-simd.md
67 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
68 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
69 (vmull_high_s16): Likewise.
70 (vmull_high_s32): Likewise.
71 (vmull_high_u8): Likewise.
72 (vmull_high_u16): Likewise.
73 (vmull_high_u32): Likewise.
75 (vmull_s16): Likewise.
76 (vmull_s32): Likewise.
78 (vmull_u16): Likewise.
79 (vmull_u32): Likewise.
81 2020-02-18 Martin Liska <mliska@suse.cz>
83 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
84 bootstrap by missing removal of invalid sanity check.
86 2020-02-18 Martin Liska <mliska@suse.cz>
89 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
90 Always compare LHS of gimple_assign.
92 2020-02-18 Martin Liska <mliska@suse.cz>
95 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
96 and return type of functions.
97 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
98 Drop MALLOC attribute for void functions.
99 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
100 malloc_state for a new VOID clone.
102 2020-02-18 Martin Liska <mliska@suse.cz>
105 * common.opt: Add -fprofile-reproducibility.
106 * doc/invoke.texi: Document it.
107 * value-prof.c (dump_histogram_value):
108 Document and support behavior for counters[0]
109 being a negative value.
110 (get_nth_most_common_value): Handle negative
111 counters[0] in respect to flag_profile_reproducible.
113 2020-02-18 Jakub Jelinek <jakub@redhat.com>
116 * cgraph.c (verify_speculative_call): Use speculative_id instead of
117 speculative_uid in messages. Remove trailing whitespace from error
118 message. Use num_speculative_call_targets instead of
119 num_speculative_targets in a message.
120 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
121 edge messages and stmt instead of cal_stmt in reference message.
123 PR tree-optimization/93780
124 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
125 before calling build_vector_type.
126 (execute_update_addresses_taken): Likewise.
129 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
130 typo, functoin -> function.
131 * tree.c (free_lang_data_in_decl): Fix comment typo,
132 functoin -> function.
133 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
135 2020-02-17 David Malcolm <dmalcolm@redhat.com>
137 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
139 (print_option_information): Don't call get_option_url if URLs
142 2020-02-17 Alexandre Oliva <oliva@adacore.com>
144 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
145 handling of register_common-less targets.
147 2020-02-17 Martin Liska <mliska@suse.cz>
150 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
152 2020-02-17 Martin Liska <mliska@suse.cz>
155 * config/rs6000/rs6000.c (rs6000_option_override_internal):
158 2020-02-17 Martin Liska <mliska@suse.cz>
161 * config/rx/elf.opt: Fix typo.
163 2020-02-17 Richard Biener <rguenther@suse.de>
166 * opts-global.c (print_ignored_options): Use inform and
169 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
172 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
174 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
177 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
178 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
180 2020-02-15 Jason Merrill <jason@redhat.com>
182 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
184 2020-02-15 Jakub Jelinek <jakub@redhat.com>
186 PR tree-optimization/93744
187 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
188 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
189 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
190 sure @2 in the first and @1 in the other patterns has no side-effects.
192 2020-02-15 David Malcolm <dmalcolm@redhat.com>
193 Bernd Edlinger <bernd.edlinger@hotmail.de>
197 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
198 * configure.ac (--with-diagnostics-urls): New configuration
199 option, based on --with-diagnostics-color.
200 (DIAGNOSTICS_URLS_DEFAULT): New define.
201 * config.h: Regenerate.
202 * configure: Regenerate.
203 * diagnostic.c (diagnostic_urls_init): Handle -1 for
204 DIAGNOSTICS_URLS_DEFAULT from configure-time
205 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
206 and TERM_URLS environment variable.
207 * diagnostic-url.h (diagnostic_url_format): New enum type.
208 (diagnostic_urls_enabled_p): rename to...
209 (determine_url_format): ... this, and change return type.
210 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
211 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
212 the linux console, and mingw.
213 (diagnostic_urls_enabled_p): rename to...
214 (determine_url_format): ... this, and adjust.
215 * pretty-print.h (pretty_printer::show_urls): rename to...
216 (pretty_printer::url_format): ... this, and change to enum.
217 * pretty-print.c (pretty_printer::pretty_printer,
218 pp_begin_url, pp_end_url, test_urls): Adjust.
219 * doc/install.texi (--with-diagnostics-urls): Document the new
220 configuration option.
221 (--with-diagnostics-color): Document the existing interaction
222 with GCC_COLORS better.
223 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
224 vindex reference. Update description of defaults based on the above.
225 (-fdiagnostics-color): Update description of how -fdiagnostics-color
226 interacts with GCC_COLORS.
228 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
231 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
232 conjunction with TARGET_GNU_TLS in early return.
234 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
236 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
237 the mode is not wider than UNITS_PER_WORD.
239 2020-02-14 Martin Jambor <mjambor@suse.cz>
241 PR tree-optimization/93516
242 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
243 access of the same type as the parent.
244 (propagate_subaccesses_from_lhs): Likewise.
246 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
249 * config/i386/avx512vbmi2intrin.h
250 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
251 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
252 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
253 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
254 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
255 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
256 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
257 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
258 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
259 of lacking a closing parenthesis.
260 * config/i386/avx512vbmi2vlintrin.h
261 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
262 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
263 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
264 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
265 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
266 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
267 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
268 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
269 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
270 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
271 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
272 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
273 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
274 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
275 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
276 _mm_shldi_epi32, _mm_mask_shldi_epi32,
277 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
278 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
280 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
283 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
284 the target function entry.
286 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
288 * common/config/arc/arc-common.c (arc_option_optimization_table):
289 Disable if-conversion step when optimized for size.
291 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
293 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
294 R12-R15 are always in ARCOMPACT16_REGS register class.
295 * config/arc/arc.opt (mq-class): Deprecate.
296 * config/arc/constraint.md ("q"): Remove dependency on mq-class
298 * doc/invoke.texi (mq-class): Update text.
299 * common/config/arc/arc-common.c (arc_option_optimization_table):
302 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
304 * config/arc/arc.c (arc_insn_cost): New function.
305 (TARGET_INSN_COST): Define.
306 * config/arc/arc.md (cost): New attribute.
307 (add_n): Use arc_nonmemory_operand.
308 (ashlsi3_insn): Likewise, also update constraints.
309 (ashrsi3_insn): Likewise.
311 (add_shift): Likewise.
312 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
314 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
316 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
318 (umulsidi_600): Likewise.
320 2020-02-13 Jakub Jelinek <jakub@redhat.com>
323 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
324 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
325 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
326 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
327 pass __A to the builtin followed by __W instead of __A followed by
329 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
330 _mm512_mask_popcnt_epi64): Likewise.
331 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
332 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
333 _mm256_mask_popcnt_epi64): Likewise.
335 PR tree-optimization/93582
336 * fold-const.h (shift_bytes_in_array_left,
337 shift_bytes_in_array_right): Declare.
338 * fold-const.c (shift_bytes_in_array_left,
339 shift_bytes_in_array_right): New function, moved from
340 gimple-ssa-store-merging.c, no longer static.
341 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
342 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
343 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
344 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
345 shift_bytes_in_array.
346 (verify_shift_bytes_in_array): Rename to ...
347 (verify_shift_bytes_in_array_left): ... this. Use
348 shift_bytes_in_array_left instead of shift_bytes_in_array.
349 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
350 instead of verify_shift_bytes_in_array.
351 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
352 / native_interpret_expr where the store covers all needed bits,
353 punt on PDP-endian, otherwise allow all involved offsets and sizes
354 not to be byte-aligned.
357 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
358 use const_0_to_255_operand predicate instead of immediate_operand.
359 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
360 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
361 vgf2p8affineinvqb_<mode><mask_name>,
362 vgf2p8affineqb_<mode><mask_name>): Drop mode from
363 const_0_to_255_operand predicated operands.
365 2020-02-12 Jeff Law <law@redhat.com>
367 * config/h8300/h8300.md (comparison shortening peepholes): Use
368 a mode iterator to merge the HImode and SImode peepholes.
370 2020-02-12 Jakub Jelinek <jakub@redhat.com>
373 * real.c (is_even): Make static. Function comment fix.
374 (is_halfway_below): Make static, don't assert R is not inf/nan,
375 instead return false for those. Small formatting fixes.
377 2020-02-12 Martin Sebor <msebor@redhat.com>
380 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
381 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
382 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
383 (strlen_check_and_optimize_call): Adjust callee name.
385 2020-02-12 Jeff Law <law@redhat.com>
387 * config/h8300/h8300.md (comparison shortening peepholes): Drop
388 (and (xor)) variant. Combine other two into single peephole.
390 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
392 PR rtl-optimization/93565
393 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
395 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
397 * config/aarch64/aarch64-simd.md
398 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
399 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
400 generating separate ADDV and zero_extend patterns.
401 * config/aarch64/iterators.md (VDQV_E): New iterator.
403 2020-02-12 Jeff Law <law@redhat.com>
405 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
406 expanders, splits, etc.
407 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
408 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
409 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
410 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
411 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
412 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
414 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
416 2020-02-12 Jakub Jelinek <jakub@redhat.com>
419 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
420 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
421 TARGET_AVX512DQ from condition.
422 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
423 instead of <mask_mode512bit_condition> in condition. If
424 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
426 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
429 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
432 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
434 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
436 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
437 where strlen is more legible.
438 (rs6000_builtin_vectorized_libmass): Ditto.
439 (rs6000_print_options_internal): Ditto.
441 2020-02-11 Martin Sebor <msebor@redhat.com>
443 PR tree-optimization/93683
444 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
446 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
448 * config/rs6000/predicates.md (cint34_operand): Rename the
449 -mprefixed-addr option to be -mprefixed.
450 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
451 the -mprefixed-addr option to be -mprefixed.
452 (OTHER_FUTURE_MASKS): Likewise.
453 (POWERPC_MASKS): Likewise.
454 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
455 the -mprefixed-addr option to be -mprefixed. Change error
456 messages to refer to -mprefixed.
457 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
459 (rs6000_legitimate_offset_address_p): Likewise.
460 (rs6000_mode_dependent_address): Likewise.
461 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
462 "-mprefixed" for target attributes and pragmas.
463 (address_to_insn_form): Rename the -mprefixed-addr option to be
465 (rs6000_adjust_insn_length): Likewise.
466 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
467 -mprefixed-addr option to be -mprefixed.
468 (ASM_OUTPUT_OPCODE): Likewise.
469 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
470 -mprefixed-addr option to be -mprefixed.
471 * config/rs6000/rs6000.opt (-mprefixed): Rename the
472 -mprefixed-addr option to be prefixed. Change the option from
473 being undocumented to being documented.
474 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
475 -mprefixed option. Update the -mpcrel documentation to mention
478 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
480 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
481 including FIRST_PSEUDO_REGISTER - 1.
482 * ira-color.c (print_hard_reg_set): Ditto.
484 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
486 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
487 (USTERNOP_QUALIFIERS): New define.
488 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
489 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
490 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
491 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
492 * config/arm/arm_neon.h (vusdot_s32): New.
493 (vusdot_lane_s32): New.
494 (vusdotq_lane_s32): New.
495 (vsudot_lane_s32): New.
496 (vsudotq_lane_s32): New.
497 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
498 * config/arm/iterators.md (DOTPROD_I8MM): New.
499 (sup, opsuffix): Add <us/su>.
500 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
501 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
503 2020-02-11 Richard Biener <rguenther@suse.de>
505 PR tree-optimization/93661
506 PR tree-optimization/93662
507 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
509 * tree-sra.c (get_access_for_expr): Likewise.
511 2020-02-10 Jakub Jelinek <jakub@redhat.com>
514 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
515 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
516 Change condition from TARGET_AVX2 to TARGET_AVX.
518 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
521 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
524 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
526 Try to generate zero-based comparisons.
527 * config/cris/cris.c (cris_reduce_compare): New function.
528 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
529 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
530 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
532 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
535 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
536 in Thumb state and also as a destination in Arm state. Add T16
539 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
541 * md.texi (Define Subst): Match closing paren in example.
543 2020-02-10 Jakub Jelinek <jakub@redhat.com>
547 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
548 arguments of strncmp.
550 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
553 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
554 but different source value.
555 (adjust_callers_for_value_intersection): New function.
556 (gather_edges_for_value): Adjust order of callers to let a
557 non-self-recursive caller be the first element.
558 (self_recursive_pass_through_p): Add a new parameter "simple", and
559 check generalized self-recursive pass-through jump function.
560 (self_recursive_agg_pass_through_p): Likewise.
561 (find_more_scalar_values_for_callers_subset): Compute value from
562 pass-through jump function for self-recursive.
563 (intersect_with_plats): Cleanup previous implementation code for value
564 itersection with self-recursive call edge.
565 (intersect_with_agg_replacements): Likewise.
566 (intersect_aggregates_with_edge): Deduce value from pass-through jump
567 function for self-recursive call edge. Cleanup previous implementation
568 code for value intersection with self-recursive call edge.
569 (decide_whether_version_node): Remove dead callers and adjust order
570 to let a non-self-recursive caller be the first element.
572 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
574 * recog.c: Move pass_split_before_sched2 code in front of
575 pass_split_before_regstack.
576 (pass_data_split_before_sched2): Rename pass to split3 from split4.
577 (pass_data_split_before_regstack): Rename pass to split4 from split3.
578 (rest_of_handle_split_before_sched2): Remove.
579 (pass_split_before_sched2::execute): Unconditionally call
581 (enable_split_before_sched2): New function.
582 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
583 (pass_split_before_regstack::gate): Ditto.
584 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
585 Update name check for renamed split4 pass.
586 * config/sh/sh.c (register_sh_passes): Update pass insertion
587 point for renamed split4 pass.
589 2020-02-09 Jakub Jelinek <jakub@redhat.com>
591 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
592 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
593 copying them around between host and target.
595 2020-02-08 Andrew Pinski <apinski@marvell.com>
598 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
599 STRICT_ALIGNMENT also.
601 2020-02-08 Jim Wilson <jimw@sifive.com>
604 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
606 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
607 Jakub Jelinek <jakub@redhat.com>
610 * config/i386/i386.h (CALL_USED_REGISTERS): Make
611 xmm16-xmm31 call-used even in 64-bit ms-abi.
613 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
615 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
616 (simd_ummla, simd_usmmla): Likewise.
617 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
618 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
621 2020-02-07 Richard Biener <rguenther@suse.de>
624 * tree-inline.c (fold_marked_statements): Do a PRE walk,
625 skipping unreachable regions.
626 (optimize_inline_calls): Skip folding stmts when we didn't
629 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
632 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
633 Don't return aggregates with only SFmode and DFmode in SSE
635 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
637 2020-02-07 Jakub Jelinek <jakub@redhat.com>
640 * config/rs6000/rs6000-logue.c
641 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
642 if it fails, move rs into end_addr and retry. Add
643 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
644 the insn pattern doesn't describe well what exactly happens to
648 * config/i386/predicates.md (avx_identity_operand): Remove.
649 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
650 (avx_<castmode><avxsizesuffix>_<castmode>,
651 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
652 a VEC_CONCAT of the operand and UNSPEC_CAST.
653 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
654 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
658 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
659 recog_data.insn if distance_non_agu_define changed it.
661 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
664 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
665 we only had X-FORM (reg+reg) addressing for vectors. Also before
666 ISA 3.0, we only had X-FORM addressing for scalars in the
667 traditional Altivec registers.
669 2020-02-06 <zhongyunde@huawei.com>
670 Vladimir Makarov <vmakarov@redhat.com>
672 PR rtl-optimization/93561
673 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
676 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
678 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
681 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
683 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
684 where the low and the high 32 bits are equal to each other specially,
685 with an rldimi instruction.
687 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
689 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
691 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
693 * config/arm/arm-tables.opt: Regenerate.
695 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
698 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
699 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
700 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
702 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
704 PR rtl-optimization/87763
705 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
707 2020-02-06 Delia Burduv <delia.burduv@arm.com>
709 * config/aarch64/aarch64-simd-builtins.def
710 (bfmlaq): New built-in function.
711 (bfmlalb): New built-in function.
712 (bfmlalt): New built-in function.
713 (bfmlalb_lane): New built-in function.
714 (bfmlalt_lane): New built-in function.
715 * config/aarch64/aarch64-simd.md
716 (aarch64_bfmmlaqv4sf): New pattern.
717 (aarch64_bfmlal<bt>v4sf): New pattern.
718 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
719 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
720 (vbfmlalbq_f32): New intrinsic.
721 (vbfmlaltq_f32): New intrinsic.
722 (vbfmlalbq_lane_f32): New intrinsic.
723 (vbfmlaltq_lane_f32): New intrinsic.
724 (vbfmlalbq_laneq_f32): New intrinsic.
725 (vbfmlaltq_laneq_f32): New intrinsic.
726 * config/aarch64/iterators.md (BF_MLA): New int iterator.
727 (bt): New int attribute.
729 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
731 * config/i386/i386.md (*pushtf): Emit "#" instead of
732 calling gcc_unreachable in insn output.
735 (*pushsf_rex64): Ditto for alternatives other than 1.
736 (*pushsf): Ditto for alternatives other than 1.
738 2020-02-06 Martin Liska <mliska@suse.cz>
740 PR gcov-profile/91971
741 PR gcov-profile/93466
742 * coverage.c (coverage_init): Revert mangling of
743 path into filename. It can lead to huge filename length.
744 Creation of subfolders seem more natural.
746 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
749 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
750 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
751 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
753 2020-02-06 Jakub Jelinek <jakub@redhat.com>
756 * config/i386/predicates.md (avx_identity_operand): New predicate.
757 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
758 define_insn_and_split.
761 * omp-low.c (use_pointer_for_field): For nested constructs, also
762 look for map clauses on target construct.
763 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
764 taskreg_nesting_level.
767 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
768 shared clause, call omp_notice_variable on outer context if any.
770 2020-02-05 Jason Merrill <jason@redhat.com>
773 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
774 non-zero address even if weak and not yet defined.
776 2020-02-05 Martin Sebor <msebor@redhat.com>
778 PR tree-optimization/92765
779 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
780 * tree-ssa-strlen.c (compute_string_length): Remove.
781 (determine_min_objsize): Remove.
782 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
783 Avoid using type size as the upper bound on string length.
784 (handle_builtin_string_cmp): Add an argument. Adjust.
785 (strlen_check_and_optimize_call): Pass additional argument to
786 handle_builtin_string_cmp.
788 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
790 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
791 (*pushdi2_rex64 peephole2): Unconditionally split after
793 (*ashl<mode>3_doubleword): Ditto.
794 (*<shift_insn><mode>3_doubleword): Ditto.
796 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
799 * config/rs6000/rs6000.c (get_vector_offset): Fix
801 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
803 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
805 2020-02-05 David Malcolm <dmalcolm@redhat.com>
808 (Special Functions for Debugging the Analyzer): Update description
809 of __analyzer_dump_exploded_nodes.
811 2020-02-05 Jakub Jelinek <jakub@redhat.com>
814 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
815 include sets and not clobbers in the vzeroupper pattern.
816 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
817 the parallel has 17 (64-bit) or 9 (32-bit) elts.
818 (*avx_vzeroupper_1): New define_insn_and_split.
821 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
822 don't run when !optimize.
823 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
826 2020-02-05 Richard Biener <rguenther@suse.de>
829 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
830 checks before matching calls.
832 2020-02-05 Jakub Jelinek <jakub@redhat.com>
834 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
835 function comment typo.
838 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
839 simd_clone_create failed when i == 0, adjust clone->nargs by
842 2020-02-05 Martin Liska <mliska@suse.cz>
845 * doc/invoke.texi: Document that one should
846 not combine ASLR and -fpch.
848 2020-02-04 Richard Biener <rguenther@suse.de>
850 PR tree-optimization/93538
851 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
853 2020-02-04 Richard Biener <rguenther@suse.de>
855 PR tree-optimization/91123
856 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
857 (vn_walk_cb_data::last_vuse): New member.
858 (vn_walk_cb_data::saved_operands): Likewsie.
859 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
860 (vn_walk_cb_data::push_partial_def): Use finish.
861 (vn_reference_lookup_2): Update last_vuse and use finish if
862 we've saved operands.
863 (vn_reference_lookup_3): Use finish and update calls to
864 push_partial_defs everywhere. When translating through
865 memcpy or aggregate copies save off operands and alias-set.
866 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
867 operation for redundant store removal.
869 2020-02-04 Richard Biener <rguenther@suse.de>
871 PR tree-optimization/92819
872 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
873 generating more stmts than before.
875 2020-02-04 Martin Liska <mliska@suse.cz>
877 * config/arm/arm.c (arm_gen_far_branch): Move the function
878 outside of selftests.
880 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
882 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
883 function to adjust PC-relative vector addresses.
884 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
885 handle vectors with PC-relative addresses.
887 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
889 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
891 (hard_reg_and_mode_to_addr_mask): Delete.
892 (rs6000_adjust_vec_address): If the original vector address
893 was REG+REG or REG+OFFSET and the element is not zero, do the add
894 of the elements in the original address before adding the offset
895 for the vector element. Use address_to_insn_form to validate the
896 address using the register being loaded, rather than guessing
897 whether the address is a DS-FORM or DQ-FORM address.
899 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
901 * config/rs6000/rs6000.c (get_vector_offset): New helper function
902 to calculate the offset in memory from the start of a vector of a
903 particular element. Add code to keep the element number in
904 bounds if the element number is variable.
905 (rs6000_adjust_vec_address): Move calculation of offset of the
906 vector element to get_vector_offset.
907 (rs6000_split_vec_extract_var): Do not do the initial AND of
908 element here, move the code to get_vector_offset.
910 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
912 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
915 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
917 * config/rs6000/constraints.md: Improve documentation.
919 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
922 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
923 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
925 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
927 * config.gcc: Remove "carrizo" support.
928 * config/gcn/gcn-opts.h (processor_type): Likewise.
929 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
930 * config/gcn/gcn.opt (gpu_type): Likewise.
931 * config/gcn/t-omp-device: Likewise.
933 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
936 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
937 * config/arm/arm.c (arm_gen_far_branch): New function
939 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
941 2020-02-03 Julian Brown <julian@codesourcery.com>
942 Tobias Burnus <tobias@codesourcery.com>
944 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
946 2020-02-03 Jakub Jelinek <jakub@redhat.com>
949 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
950 valid RTL to sum up the lowest and second lowest bytes of the popcnt
953 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
955 PR rtl-optimization/91333
956 * ira-color.c (struct allocno_color_data): Add member
958 (init_allocno_threads): Set the member up.
959 (bucket_allocno_compare_func): Add compare hard reg
962 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
964 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
966 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
967 * config.in: Regenerated.
968 * configure: Regenerated.
969 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
970 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
971 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
973 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
975 * configure: Regenerate.
977 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
979 PR rtl-optimization/91333
980 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
981 reg preferences comparison up.
983 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
985 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
986 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
987 aarch64-sve-builtins-base.h.
988 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
989 aarch64-sve-builtins-base.cc.
990 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
991 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
993 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
994 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
995 (svcvtnt): New functions.
996 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
997 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
998 (svcvtnt): New functions.
999 (svcvt): Add a form that converts f32 to bf16.
1000 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1001 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1003 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1004 Treat B as bfloat16_t.
1005 (ternary_bfloat_lane_base): New class.
1006 (ternary_bfloat_def): Likewise.
1007 (ternary_bfloat): New shape.
1008 (ternary_bfloat_lane_def): New class.
1009 (ternary_bfloat_lane): New shape.
1010 (ternary_bfloat_lanex2_def): New class.
1011 (ternary_bfloat_lanex2): New shape.
1012 (ternary_bfloat_opt_n_def): New class.
1013 (ternary_bfloat_opt_n): New shape.
1014 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1015 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1016 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1017 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1018 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1019 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1020 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1021 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1022 the pattern off the narrow mode instead of the wider one.
1023 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1024 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1025 (sve_fp_op): Handle them.
1026 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1027 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1029 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1031 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1032 * config/aarch64/aarch64-modes.def (BF): Move definition before
1033 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1034 (SVE_MODES): Handle BF modes.
1035 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1037 (aarch64_full_sve_mode): Likewise.
1038 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1040 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1041 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1042 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1043 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1045 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1047 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1048 (TYPES_all_data): Add bf16.
1049 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1050 (register_tuple_type): Increase buffer size.
1051 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1052 (bf16): New type suffix.
1053 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1054 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1055 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1056 Change type from all_data to all_arith.
1057 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1060 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1061 Matthew Malcomson <matthew.malcomson@arm.com>
1062 Richard Sandiford <richard.sandiford@arm.com>
1064 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1065 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1066 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1067 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1068 __ARM_FEATURE_MATMUL_FP64.
1069 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1070 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1071 be disabled at the same time.
1072 (f32mm): New extension.
1073 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1074 (AARCH64_FL_F64MM): Bump to the next bit up.
1075 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1076 (TARGET_SVE_F64MM): New macros.
1077 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1078 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1079 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1080 (UNSPEC_ZIP2Q): New unspeccs.
1081 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1082 (optab, sur, perm_insn): Handle the new unspecs.
1083 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1084 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1085 TARGET_SVE_F64MM instead of separate tests.
1086 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1087 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1088 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1089 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1090 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1091 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1092 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1093 (TYPES_s_signed): New macro.
1094 (TYPES_s_integer): Use it.
1095 (TYPES_d_float): New macro.
1096 (TYPES_d_data): Use it.
1097 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1098 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1099 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1100 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1101 (svmmla): New shape.
1102 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1103 template parameters.
1104 (ternary_resize2_lane_base): Likewise.
1105 (ternary_resize2_base): New class.
1106 (ternary_qq_lane_base): Likewise.
1107 (ternary_intq_uintq_lane_def): Likewise.
1108 (ternary_intq_uintq_lane): New shape.
1109 (ternary_intq_uintq_opt_n_def): New class
1110 (ternary_intq_uintq_opt_n): New shape.
1111 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1112 (ternary_uintq_intq_def): New class.
1113 (ternary_uintq_intq): New shape.
1114 (ternary_uintq_intq_lane_def): New class.
1115 (ternary_uintq_intq_lane): New shape.
1116 (ternary_uintq_intq_opt_n_def): New class.
1117 (ternary_uintq_intq_opt_n): New shape.
1118 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1119 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1120 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1121 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1123 (svdotprod_lane_impl): ...this new class.
1124 (svmmla_impl, svusdot_impl): New classes.
1125 (svdot_lane): Update to use svdotprod_lane_impl.
1126 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1127 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1129 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1130 function, with no types defined.
1131 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1132 AARCH64_FL_I8MM functions.
1133 (svmmla): New AARCH64_FL_F32MM function.
1134 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1135 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1136 AARCH64_FL_F64MM function.
1137 (REQUIRED_EXTENSIONS):
1139 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1141 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1144 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1146 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1147 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1148 (*movti_internal): Do not check for
1149 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1150 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1151 just after check for TARGET_AVX.
1152 (*movdf_internal): Ditto.
1153 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1154 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1155 * config/i386/sse.md (mov<mode>_internal): Only check
1156 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1157 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1158 (<sse>_andnot<mode>3<mask_name>): Move check for
1159 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1160 (<code><mode>3<mask_name>): Ditto.
1161 (*andnot<mode>3): Ditto.
1162 (*andnottf3): Ditto.
1163 (*<code><mode>3): Ditto.
1164 (*<code>tf3): Ditto.
1165 (*andnot<VI:mode>3): Remove
1166 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1167 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1168 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1169 (sse4_1_blendv<ssemodesuffix>): Ditto.
1170 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1171 Explain that tune applies to 128bit instructions only.
1173 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1175 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1176 to definition of hsa_kernel_description. Parse assembly to find SGPR
1177 and VGPR count of kernel and store in hsa_kernel_description.
1179 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1181 PR rtl-optimization/91838
1182 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1183 to truncate if allowed or reject combination.
1185 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1187 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1188 (find_inv_vars_cb): Likewise.
1190 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1192 * calls.c (special_function_p): Split out the check for DECL_NAME
1193 being non-NULL and fndecl being extern at file scope into a
1194 new maybe_special_function_p and call it. Drop check for fndecl
1195 being non-NULL that was after a usage of DECL_NAME (fndecl).
1196 * tree.h (maybe_special_function_p): New inline function.
1198 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1200 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1201 (mask_gather_load<mode>): ... here, and zero-initialize the
1203 (maskload<mode>di): Zero-initialize the destination.
1206 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1209 * doc/analyzer.texi (Limitations): Note that constraints on
1210 floating-point values are currently ignored.
1212 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1215 * symtab.c (symtab_node::noninterposable_alias): If localalias
1216 already exists, but is not usable, append numbers after it until
1217 a unique name is found. Formatting fix.
1220 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1223 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1225 * config/gcn/gcn.c (print_operand): Handle LTGT.
1226 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1228 2020-01-30 Richard Biener <rguenther@suse.de>
1230 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1231 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1233 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1235 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1236 without a DECL in .data.rel.ro.local.
1238 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1241 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1245 * config/i386/sse.md
1246 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1247 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1248 any_extend code iterator instead of always zero_extend.
1249 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1250 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1251 Use any_extend code iterator instead of always zero_extend.
1252 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1253 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1254 Use any_extend code iterator instead of always zero_extend.
1255 (*sse2_pmovmskb_ext): New define_insn.
1256 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1259 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1260 (*popcountsi2_zext_falsedep): New define_insn.
1262 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1264 * config.in: Regenerated.
1265 * configure: Regenerated.
1267 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1270 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1271 LLVM's assembler changed the default in version 9.
1273 2020-01-24 Jeff Law <law@redhat.com>
1275 PR tree-optimization/89689
1276 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1278 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1282 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1284 PR rtl-optimization/87763
1285 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1286 simplification to handle subregs as well as bare regs.
1287 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1289 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1292 * ira.c (ira): Revert use of simplified LRA algorithm.
1294 2020-01-29 Martin Jambor <mjambor@suse.cz>
1296 PR tree-optimization/92706
1297 * tree-sra.c (struct access): Fields first_link, last_link,
1298 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1299 next_rhs_queued and grp_rhs_queued respectively, new fields
1300 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1301 (struct assign_link): Field next renamed to next_rhs, new field
1302 next_lhs. Updated comment.
1303 (work_queue_head): Renamed to rhs_work_queue_head.
1304 (lhs_work_queue_head): New variable.
1305 (add_link_to_lhs): New function.
1306 (relink_to_new_repr): Also relink LHS lists.
1307 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1308 (add_access_to_lhs_work_queue): New function.
1309 (pop_access_from_work_queue): Renamed to
1310 pop_access_from_rhs_work_queue.
1311 (pop_access_from_lhs_work_queue): New function.
1312 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1314 (child_would_conflict_in_lacc): Renamed to
1315 child_would_conflict_in_acc. Adjusted parameter names.
1316 (create_artificial_child_access): New parameter set_grp_read, use it.
1317 (subtree_mark_written_and_enqueue): Renamed to
1318 subtree_mark_written_and_rhs_enqueue.
1319 (propagate_subaccesses_across_link): Renamed to
1320 propagate_subaccesses_from_rhs.
1321 (propagate_subaccesses_from_lhs): New function.
1322 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1325 2020-01-29 Martin Jambor <mjambor@suse.cz>
1327 PR tree-optimization/92706
1328 * tree-sra.c (struct access): Adjust comment of
1329 grp_total_scalarization.
1330 (find_access_in_subtree): Look for single children spanning an entire
1332 (scalarizable_type_p): Allow register accesses, adjust callers.
1333 (completely_scalarize): Remove function.
1334 (scalarize_elem): Likewise.
1335 (create_total_scalarization_access): Likewise.
1336 (sort_and_splice_var_accesses): Do not track total scalarization
1338 (analyze_access_subtree): New parameter totally, adjust to new meaning
1339 of grp_total_scalarization.
1340 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1341 (can_totally_scalarize_forest_p): New function.
1342 (create_total_scalarization_access): Likewise.
1343 (create_total_access_and_reshape): Likewise.
1344 (total_should_skip_creating_access): Likewise.
1345 (totally_scalarize_subtree): Likewise.
1346 (analyze_all_variable_accesses): Perform total scalarization after
1347 subaccess propagation using the new functions above.
1348 (initialize_constant_pool_replacements): Output initializers by
1349 traversing the access tree.
1351 2020-01-29 Martin Jambor <mjambor@suse.cz>
1353 * tree-sra.c (verify_sra_access_forest): New function.
1354 (verify_all_sra_access_forests): Likewise.
1355 (create_artificial_child_access): Set parent.
1356 (analyze_all_variable_accesses): Call the verifier.
1358 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1360 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1361 if called on indirect edge.
1362 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1363 speculative call if needed.
1365 2020-01-29 Richard Biener <rguenther@suse.de>
1367 PR tree-optimization/93428
1368 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1369 permutation when the load node is created.
1370 (vect_analyze_slp_instance): Re-use it here.
1372 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1374 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1376 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1378 PR rtl-optimization/93272
1379 * ira-lives.c (process_out_of_region_eh_regs): New function.
1380 (process_bb_node_lives): Call it.
1382 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1384 * coverage.c (read_counts_file): Make error message lowercase.
1386 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1388 * profile-count.c (profile_quality_display_names): Fix ordering.
1390 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1393 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1394 hash only when edge is first within the sequence.
1395 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1396 (symbol_table::create_edge): Do not set target_prob.
1397 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1399 (cgraph_edge::make_speculative): Drop target_prob parameter.
1400 (cgraph_edge::speculative_call_info): Remove.
1401 (cgraph_edge::first_speculative_call_target): New member function.
1402 (update_call_stmt_hash_for_removing_direct_edge): New function.
1403 (cgraph_edge::resolve_speculation): Rewrite to new API.
1404 (cgraph_edge::speculative_call_for_target): New member function.
1405 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1406 multiple speculation targets.
1407 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1409 (verify_speculative_call): Verify that targets form an interval.
1410 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1411 (cgraph_edge::first_speculative_call_target): New member function.
1412 (cgraph_edge::next_speculative_call_target): New member function.
1413 (cgraph_edge::speculative_call_target_ref): New member function.
1414 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1415 (cgraph_edge): Remove target_prob.
1416 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1417 Fix handling of speculative calls.
1418 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1419 * ipa-fnsummary.c (analyze_function_body): Likewise.
1420 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1421 * ipa-profile.c (dump_histogram): Fix formating.
1422 (ipa_profile_generate_summary): Watch for overflows.
1423 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1424 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1425 (update_indirect_edges_after_inlining): Update to new API.
1426 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1428 * profile-count.h: (profile_probability::adjusted): New.
1429 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1430 updating of profile.
1431 * value-prof.c (gimple_ic_transform): Rename to ...
1432 (dump_ic_profile): ... this one; update dumping.
1433 (stream_in_histogram_value): Fix formating.
1434 (gimple_value_profile_transformations): Update.
1436 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1439 * config/i386/i386.md (*movoi_internal_avx): Remove
1440 TARGET_SSE_TYPELESS_STORES check.
1441 (*movti_internal): Prefer TARGET_AVX over
1442 TARGET_SSE_TYPELESS_STORES.
1443 (*movtf_internal): Likewise.
1444 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1445 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1446 from TARGET_SSE_TYPELESS_STORES.
1448 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1450 * diagnostic-core.h (warning_at): Rename overload to...
1451 (warning_meta): ...this.
1452 (emit_diagnostic_valist): Delete decl of overload taking
1453 diagnostic_metadata.
1454 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1455 (warning_at): Rename overload taking diagnostic_metadata to...
1456 (warning_meta): ...this.
1458 2020-01-28 Richard Biener <rguenther@suse.de>
1460 PR tree-optimization/93439
1461 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1462 * tree-cfg.c (move_sese_region_to_fn): ... here.
1463 (verify_types_in_gimple_reference): Verify used cliques are
1466 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1469 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1470 argument of a pointer to struct gcc_options and pass it to
1471 parse_mtune_ctrl_str.
1472 (ix86_function_specific_restore): Pass opts to
1473 set_ix86_tune_features.
1474 (ix86_option_override_internal): Likewise.
1475 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1476 gcc_options and use it for x_ix86_tune_ctrl_string.
1478 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1480 PR rtl-optimization/87763
1481 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1482 simplification to handle subregs as well as bare regs.
1483 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1485 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1487 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1488 for reduction chains that (now) include a call.
1490 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1492 PR tree-optimization/92822
1493 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1494 out the don't-care elements of a vector whose significant elements
1495 are duplicates, make the don't-care elements duplicates too.
1497 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1499 PR tree-optimization/93434
1500 * tree-predcom.c (split_data_refs_to_components): Record which
1501 components have had aliasing loads removed. Prevent store-store
1502 commoning for all such components.
1504 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1507 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1508 -1 or is_vshift is true, use new_vector with number of elts npatterns
1509 rather than new_unary_operation.
1511 PR tree-optimization/93454
1512 * gimple-fold.c (fold_array_ctor_reference): Perform
1513 elt_size.to_uhwi () just once, instead of calling it in every
1514 iteration. Punt if that value is above size of the temporary
1515 buffer. Decrease third native_encode_expr argument when
1516 bufoff + elt_sz is above size of buf.
1518 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1520 * config/mips/mips.c (mips_declare_object_name)
1521 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1523 2020-01-27 Martin Liska <mliska@suse.cz>
1525 PR gcov-profile/93403
1526 * tree-profile.c (gimple_init_gcov_profiler): Generate
1527 both __gcov_indirect_call_profiler_v4 and
1528 __gcov_indirect_call_profiler_v4_atomic.
1530 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1533 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1535 (@aarch64_split_simd_mov<mode>): Use it.
1536 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1537 Leave the vec_extract patterns to handle 2-element vectors.
1538 (aarch64_simd_mov_from_<mode>high): Likewise.
1539 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1540 (vec_extractv2dfv1df): Likewise.
1542 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1544 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1545 jump conditions for *compare_condjump<GPI:mode>.
1547 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1550 * digraph.cc (test_edge::test_edge): Specify template for base
1553 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1555 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1557 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1559 * config/arc/arc-protos.h (gen_mlo): Remove.
1560 (gen_mhi): Likewise.
1561 * config/arc/arc.c (AUX_MULHI): Define.
1562 (arc_must_save_reister): Special handling for r58/59.
1563 (arc_compute_frame_size): Consider mlo/mhi registers.
1564 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1566 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1567 mlo/mhi name selection.
1568 (arc_restore_callee_saves): Don't early restore blink when ISR.
1569 (arc_expand_prologue): Add mlo/mhi saving.
1570 (arc_expand_epilogue): Add mlo/mhi restoring.
1573 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1574 numbering when MUL64 option is used.
1575 (DWARF2_FRAME_REG_OUT): Define.
1576 * config/arc/arc.md (arc600_stall): New pattern.
1577 (VUNSPEC_ARC_ARC600_STALL): Define.
1578 (mulsi64): Use correct mlo/mhi registers.
1579 (mulsi_600): Clean it up.
1580 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1582 (mhi_operand): Likewise.
1584 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1585 Petro Karashchenko <petro.karashchenko@ring.com>
1587 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1588 attributes if needed.
1589 (prepare_move_operands): Generate special unspec instruction for
1591 (arc_isuncached_mem_p): Propagate uncached attribute to each
1593 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1594 (VUNSPEC_ARC_STDI): Likewise.
1595 (ALLI): New mode iterator.
1596 (mALLI): New mode attribute.
1597 (lddi): New instruction pattern.
1599 (stdidi_split): Split instruction for architectures which are not
1600 supporting ll64 option.
1601 (lddidi_split): Likewise.
1603 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1605 PR rtl-optimization/92989
1606 * lra-lives.c (process_bb_lives): Update the live-in set before
1607 processing additional clobbers.
1609 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1611 PR rtl-optimization/93170
1612 * cselib.c (cselib_invalidate_regno_val): New function, split out
1614 (cselib_invalidate_regno): ...here.
1615 (cselib_invalidated_by_call_p): New function.
1616 (cselib_process_insn): Iterate over all the hard-register entries in
1617 REG_VALUES and invalidate any that cross call-clobbered registers.
1619 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1621 * dojump.c (split_comparison): Use HONOR_NANS rather than
1622 HONOR_SNANS when splitting LTGT.
1624 2020-01-27 Martin Liska <mliska@suse.cz>
1627 * opts.c (print_filtered_help): Exclude language-specific
1628 options from --help=common unless enabled in all FEs.
1630 2020-01-27 Martin Liska <mliska@suse.cz>
1632 * opts.c (print_help): Exclude params from
1633 all except --help=param.
1635 2020-01-27 Martin Liska <mliska@suse.cz>
1638 * config/i386/i386-features.c (make_resolver_func):
1639 Align the code with ppc64 target implementation.
1640 Do not generate a unique name for resolver function.
1642 2020-01-27 Richard Biener <rguenther@suse.de>
1644 PR tree-optimization/93397
1645 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1646 converted reduction chain SLP graph adjustment.
1648 2020-01-26 Marek Polacek <polacek@redhat.com>
1651 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1654 2020-01-26 Jason Merrill <jason@redhat.com>
1657 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1660 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1662 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1663 (rx_setmem): Likewise.
1665 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1668 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1669 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1670 drop <di> from constraint of last operand.
1673 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1674 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1675 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1677 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1680 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1683 2020-01-24 Jeff Law <law@redhat.com>
1685 PR tree-optimization/92788
1686 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1689 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1692 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1693 *avx_vperm_broadcast_<mode>,
1694 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1695 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1696 Move before avx2_perm<mode>/avx512f_perm<mode>.
1699 * simplify-rtx.c (simplify_const_unary_operation,
1700 simplify_const_binary_operation): Punt for mode precision above
1701 MAX_BITSIZE_MODE_ANY_INT.
1703 2020-01-24 Andrew Pinski <apinski@marvell.com>
1705 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1708 2020-01-24 Jeff Law <law@redhat.com>
1711 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1712 for REGs. Call output_operand_lossage to get more reasonable
1715 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1717 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1718 gcn_fp_compare_operator.
1719 (vec_cmpu<mode>di): Use gcn_compare_operator.
1720 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1721 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1722 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1723 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1724 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1725 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1726 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1727 gcn_fp_compare_operator.
1728 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1729 gcn_fp_compare_operator.
1730 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1731 gcn_fp_compare_operator.
1732 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1733 gcn_fp_compare_operator.
1735 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1737 * doc/install.texi (Cross-Compiler-Specific Options): Document
1738 `--with-toolexeclibdir' option.
1740 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1742 * target.def (flags_regnum): Also mention effect on delay slot filling.
1743 * doc/tm.texi: Regenerate.
1745 2020-01-23 Jeff Law <law@redhat.com>
1747 PR translation/90162
1748 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1750 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1753 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1756 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1758 PR rtl-optimization/93402
1759 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1762 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1764 * config.in: Regenerated.
1765 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1766 for TARGET_LIBC_GNUSTACK.
1767 * configure: Regenerated.
1768 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1769 found to be 2.31 or greater.
1771 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1773 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1775 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1776 (mips_asm_file_end): New function. Delegate to
1777 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1778 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1780 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1783 * config/i386/i386-modes.def (POImode): New mode.
1784 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1785 * config/i386/i386.md (DPWI): New mode attribute.
1786 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1788 (QPWI): ... this. Use POI instead of OI for TImode.
1789 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1790 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1793 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1796 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1798 (speculation_tracker_rev): New pattern.
1799 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1800 Use speculation_tracker_rev to track the inverse condition.
1802 2020-01-23 Richard Biener <rguenther@suse.de>
1804 PR tree-optimization/93381
1805 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1806 alias-set of the def as argument and record the first one.
1807 (vn_walk_cb_data::first_set): New member.
1808 (vn_reference_lookup_3): Pass the alias-set of the current def
1809 to push_partial_def. Fix alias-set used in the aggregate copy
1811 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1812 * real.c (clear_significand_below): Fix out-of-bound access.
1814 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1817 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1818 New define_insn patterns.
1820 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1822 * doc/sourcebuild.texi (check-function-bodies): Add an
1823 optional target/xfail selector.
1825 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1827 PR rtl-optimization/93124
1828 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1829 bare USE and CLOBBER insns.
1831 2020-01-22 Andrew Pinski <apinski@marvell.com>
1833 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1835 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1838 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1839 diagnostic_manager into "ana" namespace.
1840 * selftest-run-tests.c (selftest::run_tests): Update for move of
1841 selftest::run_analyzer_selftests to
1842 ana::selftest::run_analyzer_selftests.
1844 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1846 * cfgexpand.c (union_stack_vars): Update the size.
1848 2020-01-22 Richard Biener <rguenther@suse.de>
1850 PR tree-optimization/93381
1851 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1852 throughout, handle all conversions the same.
1854 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1857 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1858 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1859 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1860 Call force_reg on high_in2 unconditionally.
1862 2020-01-22 Martin Liska <mliska@suse.cz>
1864 PR tree-optimization/92924
1865 * profile.c (compute_value_histograms): Divide
1868 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1871 * output.h (assemble_name_resolve): Declare.
1872 * varasm.c (assemble_name_resolve): New function.
1873 (assemble_name): Use it.
1874 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1876 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1878 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1879 update_web_docs_git instead of update_web_docs_svn.
1881 2020-01-21 Andrew Pinski <apinski@marvell.com>
1884 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1885 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1886 (*tlsgd_small_<mode>): Likewise.
1887 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1888 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1889 register. Convert that register back to dest using convert_mode.
1891 2020-01-21 Jim Wilson <jimw@sifive.com>
1893 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1896 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1897 Uros Bizjak <ubizjak@gmail.com>
1900 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1902 (legitimize_tls_address): Do GNU2 TLS address computation in
1903 ptr_mode and zero-extend result to Pmode.
1904 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1905 :P with :PTR and Pmode with ptr_mode.
1906 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1907 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1908 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1910 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1913 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1914 the last two operands are CONST_INT_P before using them as such.
1916 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1918 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1919 to get the integer element types.
1921 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1923 * config/aarch64/aarch64-sve-builtins.h
1924 (function_expander::convert_to_pmode): Declare.
1925 * config/aarch64/aarch64-sve-builtins.cc
1926 (function_expander::convert_to_pmode): New function.
1927 (function_expander::get_contiguous_base): Use it.
1928 (function_expander::prepare_gather_address_operands): Likewise.
1929 * config/aarch64/aarch64-sve-builtins-sve2.cc
1930 (svwhilerw_svwhilewr_impl::expand): Likewise.
1932 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1935 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1936 cfun->machine->label_is_assembled.
1937 (aarch64_print_patchable_function_entry): New.
1938 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1939 * config/aarch64/aarch64.h (struct machine_function): New field,
1942 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1945 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1948 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1951 * cgraph.c (cgraph_edge::resolve_speculation,
1952 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1953 call_stmt_site_hash.
1955 2020-01-21 Martin Liska <mliska@suse.cz>
1957 * config/rs6000/rs6000.c (common_mode_defined): Remove
1960 2020-01-21 Richard Biener <rguenther@suse.de>
1962 PR tree-optimization/92328
1963 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1964 type when value-numbering same-sized store by inserting a
1966 (eliminate_dom_walker::eliminate_stmt): When eliminating
1967 a redundant store handle bit-reinterpretation of the same value.
1969 2020-01-21 Andrew Pinski <apinski@marvel.com>
1972 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1974 (prepare_block_for_update): This. Use a worklist instead of
1977 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1979 * gcc/config/arm/arm.c (clear_operation_p):
1980 Initialise last_regno, skip first iteration
1981 based on the first_set value and use ints instead
1982 of the unnecessary HOST_WIDE_INTs.
1984 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1987 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1988 compare_mode other than SFmode or DFmode.
1990 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1993 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1994 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1995 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1997 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1999 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2001 2020-01-20 Andrew Pinski <apinski@marvell.com>
2004 * targhooks.c (default_print_patchable_function_entry): Use
2005 output_asm_insn to emit the nop instruction.
2007 2020-01-20 Fangrui Song <maskray@google.com>
2010 * targhooks.c (default_print_patchable_function_entry): Align to
2013 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2016 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2017 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2018 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2019 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2020 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2021 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2022 Remove the {q} suffix from lea.
2023 (*tls_dynamic_gnu2_call_64): Renamed to ...
2024 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2025 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2026 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2027 Pass Pmode to gen_tls_dynamic_gnu2_64.
2029 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2031 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2033 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2035 * config/aarch64/aarch64-sve-builtins-base.cc
2036 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2038 2020-01-20 Richard Biener <rguenther@suse.de>
2041 * dwarf2out.c (prune_unused_types): Unconditionally mark
2042 called function DIEs.
2044 2020-01-20 Martin Liska <mliska@suse.cz>
2046 PR tree-optimization/93199
2047 * tree-eh.c (struct leh_state): Add
2048 new field outer_non_cleanup.
2049 (cleanup_is_dead_in): Pass leh_state instead
2050 of eh_region. Add a checking that state->outer_non_cleanup
2051 points to outer non-clean up region.
2052 (lower_try_finally): Record outer_non_cleanup
2054 (lower_catch): Likewise.
2055 (lower_eh_filter): Likewise.
2056 (lower_eh_must_not_throw): Likewise.
2057 (lower_cleanup): Likewise.
2059 2020-01-20 Richard Biener <rguenther@suse.de>
2061 PR tree-optimization/93094
2062 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2063 (vect_transform_loop): Likewise.
2064 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2065 loop_vectorized_call to vect_transform_loop.
2066 * tree-vect-loop.c (vect_transform_loop): Pass down
2067 loop_vectorized_call to vect_loop_versioning.
2068 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2069 the earlier discovered loop_vectorized_call.
2071 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2073 * doc/contribute.texi: Update for SVN -> Git transition.
2074 * doc/install.texi: Likewise.
2076 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2079 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2080 speculative targets.
2081 (verify_speculative_call): New function
2082 (cgraph_node::verify_node): Use it.
2083 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2086 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2089 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2090 (cgraph_edge::make_direct): Remove all indirect targets.
2091 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2092 (cgraph_node::verify_node): Verify that only one call_stmt or
2093 lto_stmt_uid is set.
2094 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2096 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2097 (lto_output_ref): Simplify streaming of stmt.
2098 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2100 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2102 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2103 Mark parameter unused.
2105 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2107 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2109 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2111 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2113 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2115 * Makefile.in: Add coroutine-passes.o.
2116 * builtin-types.def (BT_CONST_SIZE): New.
2117 (BT_FN_BOOL_PTR): New.
2118 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2119 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2120 * coroutine-builtins.def: New file.
2121 * coroutine-passes.cc: New file.
2122 * function.h (struct GTY function): Add a bit to indicate that the
2123 function is a coroutine component.
2124 * internal-fn.c (expand_CO_FRAME): New.
2125 (expand_CO_YIELD): New.
2126 (expand_CO_SUSPN): New.
2127 (expand_CO_ACTOR): New.
2128 * internal-fn.def (CO_ACTOR): New.
2132 * passes.def: Add pass_coroutine_lower_builtins,
2133 pass_coroutine_early_expand_ifns.
2134 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2135 (make_pass_coroutine_early_expand_ifns): New.
2136 * doc/invoke.texi: Document the fcoroutines command line
2139 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2141 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2144 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2145 after checking the argument is a REG. Don't use REGNO (reg)
2146 again to set last_regno, reuse regno variable instead.
2148 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2150 * doc/analyzer.texi (Limitations): Add note about NaN.
2152 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2153 Sudakshina Das <sudi.das@arm.com>
2155 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2156 and valid immediate.
2157 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2158 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2159 * config/arm/constraints.md (Pg): New.
2160 * config/arm/predicates.md (long_shift_imm): New.
2161 (arm_reg_or_long_shift_imm): Likewise.
2162 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2163 (thumb2_lsll): Likewise.
2166 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2167 Sudakshina Das <sudi.das@arm.com>
2169 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2170 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2171 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2172 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2173 * config/arm/thumb2.md (thumb2_asrl): New.
2174 (thumb2_lsll): Likewise.
2176 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2178 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2181 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2183 * gdbinit.in (help-gcc-hooks): New command.
2184 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2185 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2188 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2190 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2191 correct target macro.
2193 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2195 * config/aarch64/aarch64-protos.h
2196 (aarch64_sve_ld1ro_operand_p): New.
2197 * config/aarch64/aarch64-sve-builtins-base.cc
2198 (class load_replicate): New.
2199 (class svld1ro_impl): New.
2200 (class svld1rq_impl): Change to inherit from load_replicate.
2201 (svld1ro): New sve intrinsic function base.
2202 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2203 New DEF_SVE_FUNCTION.
2204 * config/aarch64/aarch64-sve-builtins-base.h
2205 (svld1ro): New decl.
2206 * config/aarch64/aarch64-sve-builtins.cc
2207 (function_expander::add_mem_operand): Modify assert to allow
2209 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2211 * config/aarch64/aarch64.c
2212 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2213 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2214 (aarch64_sve_ld1ro_operand_p): New.
2215 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2216 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2217 * config/aarch64/predicates.md
2218 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2220 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2222 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2223 Introduce this ACLE specified predefined macro.
2224 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2225 (fp): Disabling this disables f64mm.
2226 (simd): Disabling this disables f64mm.
2227 (fp16): Disabling this disables f64mm.
2228 (sve): Disabling this disables f64mm.
2229 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2230 (AARCH64_ISA_F64MM): New.
2231 (TARGET_F64MM): New.
2232 * doc/invoke.texi (f64mm): Document new option.
2234 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2236 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2237 (neoversen1_tunings): Likewise.
2239 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2242 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2243 Add assert to ensure prolog has been emitted.
2244 (aarch64_split_atomic_op): Likewise.
2245 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2246 Use epilogue_completed rather than reload_completed.
2247 (aarch64_atomic_exchange<mode>): Likewise.
2248 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2249 (atomic_nand<mode>): Likewise.
2250 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2251 (atomic_fetch_nand<mode>): Likewise.
2252 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2253 (atomic_nand_fetch<mode>): Likewise.
2255 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2258 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2260 (REVERSE_CONDITION): Delete.
2261 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2262 (CCFP_CCFPE): Likewise.
2263 (e): New mode attribute.
2264 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2265 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2266 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2267 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2268 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2269 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2270 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2271 name of generator from gen_ccmpdi to gen_ccmpccdi.
2272 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2273 the previous comparison but aren't able to, use the new ccmp_rev
2276 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2278 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2279 than testing directly for INTEGER_CST.
2280 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2282 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2284 PR tree-optimization/93292
2285 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2286 get_vectype_for_scalar_type returns NULL.
2288 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2290 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2291 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2293 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2295 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2297 * params.opt: (max-predicted-iterations): Set bounds.
2298 * predict.c (real_almost_one, real_br_prob_base,
2299 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2300 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2301 probabilities; do not truncate to reg_br_prob_bases.
2302 (estimate_loops_at_level): Pass max_cyclic_prob.
2303 (estimate_loops): Compute max_cyclic_prob.
2304 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2306 * profile-count.c (profile_probability::to_sreal): New.
2307 * profile-count.h (class sreal): Move up in file.
2308 (profile_probability::to_sreal): Declare.
2310 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2313 (arm_invalid_conversion): New function for target hook.
2314 (arm_invalid_unary_op): New function for target hook.
2315 (arm_invalid_binary_op): New function for target hook.
2317 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2319 * config.gcc: Add arm_bf16.h.
2320 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2321 (arm_simd_builtin_std_type): Add BFmode.
2322 (arm_init_simd_builtin_types): Define element types for vector types.
2323 (arm_init_bf16_types): New function.
2324 (arm_init_builtins): Add arm_init_bf16_types function call.
2325 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2326 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2327 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2328 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2329 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2330 (arm_mangle_type): Add __bf16.
2331 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2332 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2333 arm_bf16_ptr_type_node.
2334 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2335 define_split between ARM registers.
2336 * config/arm/arm_bf16.h: New file.
2337 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2338 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2340 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2341 * config/arm/vfp.md: Add BFmode to movhf patterns.
2343 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2344 Andre Vieira <andre.simoesdiasvieira@arm.com>
2346 * config/arm/arm-cpus.in (mve, mve_float): New features.
2347 (dsp, mve, mve.fp): New options.
2348 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2349 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2350 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2352 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2353 Thomas Preud'homme <thomas.preudhomme@arm.com>
2355 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2357 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2358 error for using -mcmse when targeting Armv8.1-M Mainline.
2360 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2361 Thomas Preud'homme <thomas.preudhomme@arm.com>
2363 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2364 address in r4 when targeting Armv8.1-M Mainline.
2365 (nonsecure_call_value_internal): Likewise.
2366 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2367 a register match_operand again. Emit BLXNS when targeting
2369 (nonsecure_call_value_reg_thumb2): Likewise.
2371 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2372 Thomas Preud'homme <thomas.preudhomme@arm.com>
2374 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2375 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2376 variable as true when floating-point ABI is not hard. Replace
2377 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2378 Generate VLSTM and VLLDM instruction respectively before and
2379 after a function call to cmse_nonsecure_call function.
2380 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2381 (VUNSPEC_VLLDM): Likewise.
2382 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2383 (lazy_load_multiple_insn): Likewise.
2385 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2386 Thomas Preud'homme <thomas.preudhomme@arm.com>
2388 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2389 (arm_emit_vfp_multi_reg_pop): Likewise.
2390 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2391 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2392 restore callee-saved VFP registers.
2394 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2395 Thomas Preud'homme <thomas.preudhomme@arm.com>
2397 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2398 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2399 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2400 callee-saved GPRs as well as clear ip register before doing a nonsecure
2401 call then restore callee-saved GPRs after it when targeting
2403 (arm_reorg): Adapt to function rename.
2405 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2406 Thomas Preud'homme <thomas.preudhomme@arm.com>
2408 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2409 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2410 clear_vfp_multiple pattern based on a new vfp parameter.
2411 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2412 targeting Armv8.1-M Mainline.
2413 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2414 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2415 whether VFP registers are available before looking call_used_regs for a
2417 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2418 of prototype of clear_operation_p.
2419 (clear_vfp_multiple_operation): New predicate.
2420 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2421 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2423 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2424 Thomas Preud'homme <thomas.preudhomme@arm.com>
2426 * config/arm/arm-protos.h (clear_operation_p): Declare.
2427 * config/arm/arm.c (clear_operation_p): New function.
2428 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2429 targeting Armv8.1-M Mainline or successor.
2430 (output_return_instruction): Only output APSR register clearing if
2431 Armv8.1-M Mainline instructions not available.
2432 (thumb_exit): Likewise.
2433 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2434 * config/arm/thumb2.md (clear_apsr): New define_insn.
2435 (clear_multiple): Likewise.
2436 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2438 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2439 Thomas Preud'homme <thomas.preudhomme@arm.com>
2441 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2442 (use_return_insn): Also return false for Armv8.1-M Mainline.
2443 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2444 Mainline instructions are available.
2445 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2446 when targeting Armv8.1-M Mainline Security Extensions.
2447 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2448 Mainline entry function.
2449 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2450 targeting Armv8.1-M Mainline or successor.
2451 (arm_expand_epilogue): Fix indentation of caller-saved register
2452 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2454 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2455 (FP_SYSREGS): Likewise.
2456 (enum vfp_sysregs_encoding): Define enum.
2457 (fp_sysreg_names): Declare.
2458 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2459 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2460 (pop_fpsysreg_insn): Likewise.
2462 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2463 Thomas Preud'homme <thomas.preudhomme@arm.com>
2465 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2466 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2467 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2468 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2469 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2470 (ARMv8_1m_main): New feature group.
2471 (armv8.1-m.main): New architecture.
2472 * config/arm/arm-tables.opt: Regenerate.
2473 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2474 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2475 (arm_options_perform_arch_sanity_checks): Error out when targeting
2476 Armv8.1-M Mainline Security Extensions.
2477 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2479 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2481 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2482 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2483 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2484 aarch64_bfdot_laneq): New.
2485 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2486 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2487 vbfdotq_laneq_f32): New.
2488 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2489 VBFMLA_W, VBF): New.
2490 (isquadop): Add V4BF, V8BF.
2492 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2494 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2495 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2496 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2497 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2498 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2499 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2500 usdot_laneq, sudot_lane,sudot_laneq): New.
2501 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2502 (aarch64_<sur>dot_lane): New.
2503 * config/aarch64/arm_neon.h (vusdot_s32): New.
2505 (vusdot_lane_s32): New.
2506 (vsudot_lane_s32): New.
2507 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2508 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2510 2020-01-16 Martin Liska <mliska@suse.cz>
2512 * value-prof.c (dump_histogram_value): Fix
2513 obvious spacing issue.
2515 2020-01-16 Andrew Pinski <apinski@marvell.com>
2517 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2518 !storage_order_barrier_p.
2520 2020-01-16 Andrew Pinski <apinski@marvell.com>
2522 * sched-int.h (_dep): Add unused bit-field field for the padding.
2523 * sched-deps.c (init_dep_1): Init unused field.
2525 2020-01-16 Andrew Pinski <apinski@marvell.com>
2527 * optabs.h (create_expand_operand): Initialize target field also.
2529 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2531 PR tree-optimization/92429
2532 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2533 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2535 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2538 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2540 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2541 aarch64_sve_int_mode to each mode.
2543 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2545 * doc/analyzer.texi (Overview): Add note about
2546 -fdump-ipa-analyzer.
2548 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2550 PR tree-optimization/93231
2551 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2552 input_type is unsigned. Use tree_to_shwi for shift constant.
2553 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2554 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2555 inputs more efficiently.
2557 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2559 * config/i386/i386.md (*movsf_internal): Do not require
2560 SSE2 ISA for alternatives 14 and 15.
2562 2020-01-15 Richard Biener <rguenther@suse.de>
2565 * tree-eh.c (sink_clobbers): If we already visited the destination
2566 block do not defer insertion.
2567 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2568 the purpose of defered insertion.
2570 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2572 * BASE-VER: Bump to 10.0.1.
2574 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2576 PR tree-optimization/93247
2577 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2578 type of the stmt that we're going to vectorize.
2580 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2582 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2583 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2586 2020-01-15 Martin Liska <mliska@suse.cz>
2588 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2589 2 calls of streamer_read_hwi in a function call.
2591 2020-01-15 Richard Biener <rguenther@suse.de>
2593 * alias.c (record_alias_subset): Avoid redundant work when
2594 subset is already recorded.
2596 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2598 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2599 the analyzer options provide CWE identifiers.
2601 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2603 * tree-diagnostic-path.cc (path_summary::event_range::print):
2604 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2605 using get_pure_location.
2607 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2609 PR tree-optimization/93262
2610 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2611 perform head trimming only if the last argument is constant,
2612 either all ones, or larger or equal to head trim, in the latter
2613 case decrease the last argument by head_trim.
2615 PR tree-optimization/93249
2616 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2617 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2618 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2619 perform head trim unless we can prove there are no '\0' chars
2620 from the source among the first head_trim chars.
2622 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2624 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2626 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2629 * config/i386/sse.md
2630 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2631 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2632 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2633 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2634 just a single alternative instead of two, make operands 1 and 2
2637 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2640 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2643 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2645 * Makefile.in (lang_opt_files): Add analyzer.opt.
2646 (ANALYZER_OBJS): New.
2647 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2648 tristate.o and ANALYZER_OBJS.
2649 (TEXI_GCCINT_FILES): Add analyzer.texi.
2650 * common.opt (-fanalyzer): New driver option.
2651 * config.in: Regenerate.
2652 * configure: Regenerate.
2653 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2654 (gccdepdir): Also create depdir for "analyzer" subdir.
2655 * digraph.cc: New file.
2656 * digraph.h: New file.
2657 * doc/analyzer.texi: New file.
2658 * doc/gccint.texi ("Static Analyzer") New menu item.
2659 (analyzer.texi): Include it.
2660 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2661 ("Warning Options"): Add static analysis warnings to the list.
2662 (-Wno-analyzer-double-fclose): New option.
2663 (-Wno-analyzer-double-free): New option.
2664 (-Wno-analyzer-exposure-through-output-file): New option.
2665 (-Wno-analyzer-file-leak): New option.
2666 (-Wno-analyzer-free-of-non-heap): New option.
2667 (-Wno-analyzer-malloc-leak): New option.
2668 (-Wno-analyzer-possible-null-argument): New option.
2669 (-Wno-analyzer-possible-null-dereference): New option.
2670 (-Wno-analyzer-null-argument): New option.
2671 (-Wno-analyzer-null-dereference): New option.
2672 (-Wno-analyzer-stale-setjmp-buffer): New option.
2673 (-Wno-analyzer-tainted-array-index): New option.
2674 (-Wno-analyzer-use-after-free): New option.
2675 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2676 (-Wno-analyzer-use-of-uninitialized-value): New option.
2677 (-Wanalyzer-too-complex): New option.
2678 (-fanalyzer-call-summaries): New warning.
2679 (-fanalyzer-checker=): New warning.
2680 (-fanalyzer-fine-grained): New warning.
2681 (-fno-analyzer-state-merge): New warning.
2682 (-fno-analyzer-state-purge): New warning.
2683 (-fanalyzer-transitivity): New warning.
2684 (-fanalyzer-verbose-edges): New warning.
2685 (-fanalyzer-verbose-state-changes): New warning.
2686 (-fanalyzer-verbosity=): New warning.
2687 (-fdump-analyzer): New warning.
2688 (-fdump-analyzer-callgraph): New warning.
2689 (-fdump-analyzer-exploded-graph): New warning.
2690 (-fdump-analyzer-exploded-nodes): New warning.
2691 (-fdump-analyzer-exploded-nodes-2): New warning.
2692 (-fdump-analyzer-exploded-nodes-3): New warning.
2693 (-fdump-analyzer-supergraph): New warning.
2694 * doc/sourcebuild.texi (dg-require-dot): New.
2695 (dg-check-dot): New.
2696 * gdbinit.in (break-on-saved-diagnostic): New command.
2697 * graphviz.cc: New file.
2698 * graphviz.h: New file.
2699 * ordered-hash-map-tests.cc: New file.
2700 * ordered-hash-map.h: New file.
2701 * passes.def (pass_analyzer): Add before
2702 pass_ipa_whole_program_visibility.
2703 * selftest-run-tests.c (selftest::run_tests): Call
2704 selftest::ordered_hash_map_tests_cc_tests.
2705 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2707 * shortest-paths.h: New file.
2708 * timevar.def (TV_ANALYZER): New timevar.
2709 (TV_ANALYZER_SUPERGRAPH): Likewise.
2710 (TV_ANALYZER_STATE_PURGE): Likewise.
2711 (TV_ANALYZER_PLAN): Likewise.
2712 (TV_ANALYZER_SCC): Likewise.
2713 (TV_ANALYZER_WORKLIST): Likewise.
2714 (TV_ANALYZER_DUMP): Likewise.
2715 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2716 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2717 * tree-pass.h (make_pass_analyzer): New decl.
2718 * tristate.cc: New file.
2719 * tristate.h: New file.
2721 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2724 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2725 alternatives 9 and 10.
2727 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2729 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2730 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2731 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2732 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2733 (selftest::hash_map_tests_c_tests): Call it.
2734 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2735 New static constant, using the value of = H::empty_zero_p.
2736 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2737 from default_hash_traits <Value>.
2738 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2740 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2741 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2742 calls to mark_empty with !Descriptor::empty_zero_p.
2743 (hash_table::empty_slow): Conditionalize the memset call with a
2744 check that Descriptor::empty_zero_p; otherwise, loop through the
2745 entries calling mark_empty on them.
2746 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2747 (pointer_hash::empty_zero_p): Likewise.
2748 (pair_hash::empty_zero_p): Likewise.
2749 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2751 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2752 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2753 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2754 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2755 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2756 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2757 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2758 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2760 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2763 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2765 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2766 fix typo on return value.
2768 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2771 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2773 (cgraph_edge::make_speculative): Add param for setting speculative_id
2775 (cgraph_edge::speculative_call_info): Update comments and find reference
2776 by speculative_id for multiple indirect targets.
2777 (cgraph_edge::resolve_speculation): Decrease the speculations
2778 for indirect edge, drop it's speculative if not direct target
2779 left. Update comments.
2780 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2781 (cgraph_node::dump): Print num_speculative_call_targets.
2782 (cgraph_node::verify_node): Don't report error if speculative
2783 edge not include statement.
2784 (cgraph_edge::num_speculative_call_targets_p): New function.
2785 * cgraph.h (int common_target_id): Remove.
2786 (int common_target_probability): Remove.
2787 (num_speculative_call_targets): New variable.
2788 (make_speculative): Add param for setting speculative_id.
2789 (cgraph_edge::num_speculative_call_targets_p): New declare.
2790 (target_prob): New variable.
2791 (speculative_id): New variable.
2792 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2793 call summaries for multiple speculative call targets.
2794 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2795 * ipa-profile.c (struct speculative_call_target): New struct.
2796 (class speculative_call_summary): New class.
2797 (class speculative_call_summaries): New class.
2798 (call_sums): New variable.
2799 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2800 (ipa_profile_write_edge_summary): New function.
2801 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2802 (ipa_profile_dump_all_summaries): New function.
2803 (ipa_profile_read_edge_summary): New function.
2804 (ipa_profile_read_summary_section): New function.
2805 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2806 (ipa_profile): Generate num_speculative_call_targets from
2808 * ipa-ref.h (speculative_id): New variable.
2809 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2810 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2811 common_target_probability. Stream out speculative_id and
2812 num_speculative_call_targets.
2813 (input_edge): Likewise.
2814 * predict.c (dump_prediction): Remove edges count assert to be
2816 * symtab.c (symtab_node::create_reference): Init speculative_id.
2817 (symtab_node::clone_references): Clone speculative_id.
2818 (symtab_node::clone_referring): Clone speculative_id.
2819 (symtab_node::clone_reference): Clone speculative_id.
2820 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2821 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2822 if indirect call contains multiple speculative targets.
2823 * value-prof.h (check_ic_target): Remove.
2824 * value-prof.c (gimple_value_profile_transformations):
2825 Use void function gimple_ic_transform.
2826 * value-prof.c (gimple_ic_transform): Handle topn case.
2827 Fix comment typos. Change it to a void function.
2829 2020-01-13 Andrew Pinski <apinski@marvell.com>
2831 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2832 (octeontx2t98): New define.
2833 (octeontx2t96): New define.
2834 (octeontx2t93): New define.
2835 (octeontx2f95): New define.
2836 (octeontx2f95n): New define.
2837 (octeontx2f95mm): New define.
2838 * config/aarch64/aarch64-tune.md: Regenerate.
2839 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2841 2020-01-13 Jason Merrill <jason@redhat.com>
2843 PR c++/33799 - destroy return value if local cleanup throws.
2844 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2846 2020-01-13 Martin Liska <mliska@suse.cz>
2848 * ipa-cp.c (get_max_overall_size): Use newly
2849 renamed param param_ipa_cp_unit_growth.
2850 * params.opt: Remove legacy param name.
2852 2020-01-13 Martin Sebor <msebor@redhat.com>
2854 PR tree-optimization/93213
2855 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2856 stores to be eliminated.
2858 2020-01-13 Martin Liska <mliska@suse.cz>
2860 * opts.c (print_help): Do not print CL_PARAM
2861 and CL_WARNING for CL_OPTIMIZATION.
2863 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2866 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2867 depending on optimization settings.
2869 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2871 PR tree-optimization/90838
2872 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2873 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2874 argument rather than to initialize temporary for targets that
2875 don't use the mode argument at all. Initialize ctzval to avoid
2878 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2880 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2881 * tree-core.h: Document it.
2882 * gimplify.c (gimplify_omp_workshare): Set it.
2883 * omp-low.c (lower_omp_target): Use it.
2884 * tree-pretty-print.c (dump_omp_clause): Print it.
2886 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2887 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2889 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2891 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2892 * common.opt (fdiagnostics-path-format=): New option.
2893 (diagnostic_path_format): New enum.
2894 (fdiagnostics-show-path-depths): New option.
2895 * coretypes.h (diagnostic_event_id_t): New forward decl.
2896 * diagnostic-color.c (color_dict): Add "path".
2897 * diagnostic-event-id.h: New file.
2898 * diagnostic-format-json.cc (json_from_expanded_location): Make
2900 (json_end_diagnostic): Call context->make_json_for_path if it
2901 exists and the diagnostic has a path.
2902 (diagnostic_output_format_init): Clear context->print_path.
2903 * diagnostic-path.h: New file.
2904 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2905 when printing a run of events in a diagnostic_path so that they
2906 all get the same color.
2907 (layout::m_diagnostic_path_p): New field.
2908 (layout::layout): Initialize it.
2909 (layout::print_any_labels): Don't colorize the label text for an
2910 event in a diagnostic_path.
2911 (gcc_rich_location::add_location_if_nearby): Add
2912 "restrict_to_current_line_spans" and "label" params. Pass the
2913 former to layout.maybe_add_location_range; pass the latter
2914 when calling add_range.
2915 * diagnostic.c: Include "diagnostic-path.h".
2916 (diagnostic_initialize): Initialize context->path_format and
2917 context->show_path_depths.
2918 (diagnostic_show_any_path): New function.
2919 (diagnostic_path::interprocedural_p): New function.
2920 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2921 (simple_diagnostic_path::num_events): New function.
2922 (simple_diagnostic_path::get_event): New function.
2923 (simple_diagnostic_path::add_event): New function.
2924 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2925 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2926 (debug): New overload taking a diagnostic_path *.
2927 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2928 * diagnostic.h (enum diagnostic_path_format): New enum.
2929 (json::value): New forward decl.
2930 (diagnostic_context::path_format): New field.
2931 (diagnostic_context::show_path_depths): New field.
2932 (diagnostic_context::print_path): New callback field.
2933 (diagnostic_context::make_json_for_path): New callback field.
2934 (diagnostic_show_any_path): New decl.
2935 (json_from_expanded_location): New decl.
2936 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2937 (-fdiagnostics-show-path-depths): New option.
2938 (-fdiagnostics-color): Add "path" to description of default
2939 GCC_COLORS; describe it.
2940 (-fdiagnostics-format=json): Document how diagnostic paths are
2941 represented in the JSON output format.
2942 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2943 Add optional params "restrict_to_current_line_spans" and "label".
2944 * opts.c (common_handle_option): Handle
2945 OPT_fdiagnostics_path_format_ and
2946 OPT_fdiagnostics_show_path_depths.
2947 * pretty-print.c: Include "diagnostic-event-id.h".
2948 (pp_format): Implement "%@" format code for printing
2949 diagnostic_event_id_t *.
2950 (selftest::test_pp_format): Add tests for "%@".
2951 * selftest-run-tests.c (selftest::run_tests): Call
2952 selftest::tree_diagnostic_path_cc_tests.
2953 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2954 * toplev.c (general_init): Initialize global_dc->path_format and
2955 global_dc->show_path_depths.
2956 * tree-diagnostic-path.cc: New file.
2957 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2958 non-static. Drop "diagnostic" param in favor of storing the
2959 original value of "where" and re-using it.
2960 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2961 maybe_unwind_expanded_macro_loc.
2962 (tree_diagnostics_defaults): Initialize context->print_path and
2963 context->make_json_for_path.
2964 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2966 (default_tree_make_json_for_path): New decl.
2967 (maybe_unwind_expanded_macro_loc): New decl.
2969 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2971 PR tree-optimization/93210
2972 * fold-const.h (native_encode_initializer,
2973 can_native_interpret_type_p): Declare.
2974 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2976 (native_encode_initializer): New function, moved from dwarf2out.c.
2977 Adjust to native_encode_expr compatible arguments, including dry-run
2978 and partial extraction modes. Don't handle STRING_CST.
2979 (can_native_interpret_type_p): No longer static.
2980 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2981 offset / BITS_PER_UNIT fits into int and don't call it if
2982 can_native_interpret_type_p fails. If suboff is NULL and for
2983 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2984 native_encode_initializer.
2985 (fold_const_aggregate_ref_1): Formatting fix.
2986 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2987 (tree_add_const_value_attribute): Adjust caller.
2989 PR tree-optimization/90838
2990 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2991 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2992 CTZ_DEFINED_VALUE_AT_ZERO.
2994 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2997 * lra-constraints.c (match_reload): Permit input operands have the
2998 same mode as output while other input operands have a different
3001 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3003 PR tree-optimization/90838
3004 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3005 (check_ctz_string): Likewise.
3006 (optimize_count_trailing_zeroes): Likewise.
3007 (simplify_count_trailing_zeroes): Likewise.
3008 (pass_forwprop::execute): Try ctz simplification.
3009 * match.pd: Add matching for ctz idioms.
3011 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3013 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3015 (aarch64_invalid_unary_op): New function for target hook.
3016 (aarch64_invalid_binary_op): New function for target hook.
3018 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3020 * config.gcc: Add arm_bf16.h.
3021 * config/aarch64/aarch64-builtins.c
3022 (aarch64_simd_builtin_std_type): Add BFmode.
3023 (aarch64_init_simd_builtin_types): Define element types for vector
3025 (aarch64_init_bf16_types): New function.
3026 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3027 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3029 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3030 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3032 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3033 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3034 * config/aarch64/aarch64.c
3035 (aarch64_classify_vector_mode): Add support for BF types.
3036 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3037 (aarch64_vq_mode): Add support for BF types.
3038 (aarch64_simd_container_mode): Add support for BF types.
3039 (aarch64_mangle_type): Add support for BF scalar type.
3040 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3041 * config/aarch64/arm_bf16.h: New file.
3042 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3043 * config/aarch64/iterators.md: Add BF types to mode attributes.
3044 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3046 2020-01-10 Jason Merrill <jason@redhat.com>
3048 PR c++/93173 - incorrect tree sharing.
3049 * gimplify.c (copy_if_shared): No longer static.
3050 * gimplify.h: Declare it.
3052 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3054 * doc/invoke.texi (-msve-vector-bits=): Document that
3055 -msve-vector-bits=128 now generates VL-specific code for
3056 little-endian targets.
3057 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3058 build_vector_type_for_mode to construct the data vector types.
3059 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3060 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3061 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3062 for 128-bit vectors.
3064 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3066 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3069 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3071 * config/aarch64/aarch64-builtins.c
3072 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3073 rather than checking the number of elements and the element mode.
3075 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3077 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3078 get_related_vectype_for_scalar_type rather than build_vector_type
3079 to create the index type for a conditional reduction.
3081 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3083 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3084 for any type of gather or scatter, including strided accesses.
3086 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3088 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3091 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3093 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3095 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3096 parameter and its use to reset DR_OFFSET's.
3097 (vect_transform_loop): Remove orig_drs_init argument.
3098 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3099 member of dr_vec_info rather than the offset of the associated
3100 data_reference's innermost_loop_behavior.
3101 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3102 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3103 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3104 get_dr_vinfo_offset.
3105 (vectorizable_store): Likewise.
3106 (vectorizable_load): Likewise.
3108 2020-01-10 Richard Biener <rguenther@suse.de>
3110 * gimple-ssa-store-merging
3111 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3113 2020-01-10 Martin Liska <mliska@suse.cz>
3116 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3117 encapsulation that was there before r280040.
3119 2020-01-10 Richard Biener <rguenther@suse.de>
3122 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3123 sequences to avoid walking them again for secondary opportunities.
3124 (pass_lower_eh_dispatch::execute): Instead actually insert
3127 2020-01-10 Richard Biener <rguenther@suse.de>
3130 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3131 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3132 avoid quadraticness.
3134 2020-01-10 Martin Jambor <mjambor@suse.cz>
3136 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3137 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3138 to get param_ipa_sra_max_replacements.
3139 (param_splitting_across_edge): Pass the caller to
3140 pull_accesses_from_callee.
3142 2020-01-10 Martin Jambor <mjambor@suse.cz>
3144 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3145 * ipa-cp.c (max_new_size): Removed.
3146 (orig_overall_size): New variable.
3147 (get_max_overall_size): New function.
3148 (estimate_local_effects): Use it. Adjust dump.
3149 (decide_about_value): Likewise.
3150 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3151 orig_overall_size. Adjust dump.
3152 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3154 2020-01-10 Martin Jambor <mjambor@suse.cz>
3156 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3157 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3158 instead of param_ipa_max_agg_items.
3159 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3160 optimization info for the callee.
3162 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3164 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3165 markers if debug_inline_points is false.
3167 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3169 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3171 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3172 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3173 aarch64-sve-builtins-sve2.h.
3174 (aarch64-sve-builtins-sve2.o): New rule.
3175 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3176 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3177 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3178 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3179 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3180 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3182 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3183 instructions that are handled here.
3184 (@cond_asrd<mode>): Generalize to...
3185 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3186 (*cond_asrd<mode>_2): Generalize to...
3187 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3188 (*cond_asrd<mode>_z): Generalize to...
3189 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3190 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3191 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3192 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3193 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3195 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3196 (@aarch64_scatter_stnt<mode>): Likewise.
3197 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3198 (@aarch64_mul_lane_<mode>): Likewise.
3199 (@aarch64_sve_suqadd<mode>_const): Likewise.
3200 (*<sur>h<addsub><mode>): Generalize to...
3201 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3203 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3204 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3205 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3206 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3207 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3208 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3209 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3210 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3211 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3212 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3213 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3214 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3215 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3216 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3217 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3218 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3219 (@aarch64_sve2_xar<mode>): Likewise.
3220 (@aarch64_sve2_bcax<mode>): Likewise.
3221 (*aarch64_sve2_eor3<mode>): Rename to...
3222 (@aarch64_sve2_eor3<mode>): ...this.
3223 (@aarch64_sve2_bsl<mode>): New expander.
3224 (@aarch64_sve2_nbsl<mode>): Likewise.
3225 (@aarch64_sve2_bsl1n<mode>): Likewise.
3226 (@aarch64_sve2_bsl2n<mode>): Likewise.
3227 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3228 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3229 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3230 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3231 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3232 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3233 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3234 (<su>mull<bt><Vwide>): Generalize to...
3235 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3237 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3238 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3239 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3240 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3241 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3242 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3243 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3244 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3245 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3246 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3247 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3248 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3249 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3250 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3251 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3252 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3253 (<SHRNB:r>shrnb<mode>): Generalize to...
3254 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3256 (<SHRNT:r>shrnt<mode>): Generalize to...
3257 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3259 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3260 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3261 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3262 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3263 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3264 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3265 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3266 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3267 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3268 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3269 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3270 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3271 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3272 (@aarch64_sve2_cvtnt<mode>): Likewise.
3273 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3274 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3275 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3276 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3277 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3278 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3279 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3280 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3281 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3282 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3283 (@aarch64_sve2_pmul<mode>): Likewise.
3284 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3285 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3286 (@aarch64_sve2_tbl2<mode>): Likewise.
3287 (@aarch64_sve2_tbx<mode>): Likewise.
3288 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3289 (@aarch64_sve2_histcnt<mode>): Likewise.
3290 (@aarch64_sve2_histseg<mode>): Likewise.
3291 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3292 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3293 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3294 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3295 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3296 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3297 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3298 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3299 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3300 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3301 (SVE2_PMULL_PAIR_I): New mode iterators.
3302 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3303 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3304 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3305 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3306 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3307 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3308 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3309 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3310 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3311 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3312 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3313 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3314 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3315 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3316 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3317 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3318 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3319 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3320 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3321 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3322 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3323 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3324 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3325 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3326 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3327 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3328 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3329 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3330 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3331 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3332 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3334 (VNARROW, Ventype): New mode attributes.
3335 (Vewtype): Handle VNx2DI. Fix typo in comment.
3336 (VDOUBLE): New mode attribute.
3337 (sve_lane_con): Handle VNx8HI.
3338 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3339 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3340 (sve_int_op, sve_int_op_rev): Handle the above codes.
3341 (sve_pred_int_rhs2_operand): Likewise.
3342 (MULLBT, SHRNB, SHRNT): Delete.
3343 (SVE_INT_SHIFT_IMM): New int iterator.
3344 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3345 and UNSPEC_WHILEHS for TARGET_SVE2.
3346 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3347 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3348 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3349 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3350 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3351 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3352 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3353 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3354 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3355 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3356 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3357 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3358 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3359 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3360 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3361 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3362 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3363 (optab): Handle the new unspecs.
3364 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3366 (lr): Handle the new unspecs.
3368 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3369 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3370 (sve_int_qsub_op): New int attributes.
3371 (sve_fp_op, rot): Handle the new unspecs.
3372 * config/aarch64/aarch64-sve-builtins.h
3373 (function_resolver::require_matching_pointer_type): Declare.
3374 (function_resolver::resolve_unary): Add an optional boolean argument.
3375 (function_resolver::finish_opt_n_resolution): Add an optional
3376 type_suffix_index argument.
3377 (gimple_folder::redirect_call): Declare.
3378 (gimple_expander::prepare_gather_address_operands): Add an optional
3380 * config/aarch64/aarch64-sve-builtins.cc: Include
3381 aarch64-sve-builtins-sve2.h.
3382 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3383 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3384 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3385 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3386 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3387 (TYPES_s_unsigned): Likewise.
3388 (TYPES_s_integer): Use TYPES_s_unsigned.
3389 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3390 (TYPES_sd_integer): Use them.
3391 (TYPES_d_unsigned): New macro.
3392 (TYPES_d_integer): Use it.
3393 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3394 (TYPES_cvt_narrow): Likewise.
3395 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3396 (preds_mx): New variable.
3397 (function_builder::add_overloaded_function): Allow the new feature
3398 set to be more restrictive than the original one.
3399 (function_resolver::infer_pointer_type): Remove qualifiers from
3400 the pointer type before printing it.
3401 (function_resolver::require_matching_pointer_type): New function.
3402 (function_resolver::resolve_sv_displacement): Handle functions
3403 that don't support 32-bit vector indices or svint32_t vector offsets.
3404 (function_resolver::finish_opt_n_resolution): Take the inferred type
3405 as a separate argument.
3406 (function_resolver::resolve_unary): Optionally treat all forms in
3407 the same way as normal merging functions.
3408 (gimple_folder::redirect_call): New function.
3409 (function_expander::prepare_gather_address_operands): Add an argument
3410 that says whether scaled forms are available. If they aren't,
3411 handle scaling of vector indices and don't add the extension and
3413 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3414 fall back to using cond_* instead.
3415 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3416 Split out the member variables into...
3417 (rtx_code_function_base): ...this new base class.
3418 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3419 (unspec_based_function): Split out the member variables into...
3420 (unspec_based_function_base): ...this new base class.
3421 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3422 (unspec_based_function_exact_insn): New class.
3423 (unspec_based_add_function, unspec_based_add_lane_function)
3424 (unspec_based_lane_function, unspec_based_pred_function)
3425 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3426 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3427 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3429 (unspec_based_fused_function): New class.
3430 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3431 (unspec_based_fused_lane_function): New class.
3432 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3434 (CODE_FOR_MODE1): New macro.
3435 (fixed_insn_function): New class.
3436 (while_comparison): Likewise.
3437 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3438 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3439 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3440 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3441 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3442 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3443 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3444 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3445 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3446 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3447 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3448 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3449 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3450 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3451 Also add an initial argument for unary_convert_narrowt, regardless
3452 of the predication type.
3453 (build_32_64): Allow loads and stores to specify MODE_none.
3454 (build_sv_index64, build_sv_uint_offset): New functions.
3455 (long_type_suffix): New function.
3456 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3457 (binary_imm_long_base, load_gather_sv_base): Likewise.
3458 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3459 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3460 (unary_narrowb_base, unary_narrowt_base): Likewise.
3461 (binary_long_lane_def, binary_long_lane): New shape.
3462 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3463 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3464 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3465 (binary_to_uint_def, binary_to_uint): Likewise.
3466 (binary_wide_def, binary_wide): Likewise.
3467 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3468 (compare_def, compare): Likewise.
3469 (compare_ptr_def, compare_ptr): Likewise.
3470 (load_ext_gather_index_restricted_def,
3471 load_ext_gather_index_restricted): Likewise.
3472 (load_ext_gather_offset_restricted_def,
3473 load_ext_gather_offset_restricted): Likewise.
3474 (load_gather_sv_def): Inherit from load_gather_sv_base.
3475 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3476 (shift_left_imm_def, shift_left_imm): Likewise.
3477 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3478 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3479 (store_scatter_index_restricted_def,
3480 store_scatter_index_restricted): Likewise.
3481 (store_scatter_offset_restricted_def,
3482 store_scatter_offset_restricted): Likewise.
3483 (tbl_tuple_def, tbl_tuple): Likewise.
3484 (ternary_long_lane_def, ternary_long_lane): Likewise.
3485 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3486 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3487 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3488 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3489 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3490 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3491 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3492 (ternary_uint_def, ternary_uint): Likewise.
3493 (unary_convert): Fix typo in comment.
3494 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3495 (unary_long_def, unary_long): Likewise.
3496 (unary_narrowb_def, unary_narrowb): Likewise.
3497 (unary_narrowt_def, unary_narrowt): Likewise.
3498 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3499 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3500 (unary_to_int_def, unary_to_int): Likewise.
3501 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3502 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3503 (svasrd_impl): Delete.
3504 (svcadd_impl::expand): Handle integer operations too.
3505 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3506 new functions to derive the unspec numbers.
3507 (svmla_svmls_lane_impl): Replace with...
3508 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3509 integer operations too.
3510 (svwhile_impl): Rename to...
3511 (svwhilelx_impl): ...this and inherit from while_comparison.
3512 (svasrd): Use unspec_based_function.
3513 (svmla_lane): Use svmla_lane_impl.
3514 (svmls_lane): Use svmls_lane_impl.
3515 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3516 (svwhilele, svwhilelt): Use svwhilelx_impl.
3517 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3518 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3519 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3520 * config/aarch64/aarch64-sve-builtins.def: Include
3521 aarch64-sve-builtins-sve2.def.
3523 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3525 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3526 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3527 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3528 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3529 immediates as well as vector ones.
3530 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3531 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3532 (aarch64_sve_qsub_immediate): Update calls accordingly.
3534 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3536 * config/aarch64/aarch64-sve2.md: Add banner comments.
3537 (<su>mulh<r>s<mode>3): Move further up file.
3538 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3539 (*aarch64_sve2_sra<mode>): Move further down file.
3540 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3542 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3544 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3546 (while_optab_cmp): Handle them.
3547 * config/aarch64/aarch64-sve.md
3548 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3549 and add a "@" marker.
3550 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3551 instead of gen_aarch64_sve2_while_ptest.
3552 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3554 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3556 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3557 (UNSPEC_WHILELE): ...this.
3558 (UNSPEC_WHILE_LO): Rename to...
3559 (UNSPEC_WHILELO): ...this.
3560 (UNSPEC_WHILE_LS): Rename to...
3561 (UNSPEC_WHILELS): ...this.
3562 (UNSPEC_WHILE_LT): Rename to...
3563 (UNSPEC_WHILELT): ...this.
3564 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3565 (cmp_op, while_optab_cmp): Likewise.
3566 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3567 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3568 (svwhilelt): Likewise.
3570 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3572 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3573 (unary_to_uint): Define.
3574 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3575 (unary_count): Rename to...
3576 (unary_to_uint_def, unary_to_uint): ...this.
3577 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3579 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3581 * config/aarch64/aarch64-sve-builtins-functions.h
3582 (code_for_mode_function): New class.
3583 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3584 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3585 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3586 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3587 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3589 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3591 * config/aarch64/iterators.md (addsub): New code attribute.
3592 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3594 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3595 in the asm string and attributes. Fix indentation.
3596 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3598 (@aarch64_sve_<optab><mode>): ...this.
3599 * config/aarch64/aarch64-sve-builtins.h
3600 (function_expander::expand_signed_unpred_op): Delete.
3601 * config/aarch64/aarch64-sve-builtins.cc
3602 (function_expander::expand_signed_unpred_op): Likewise.
3603 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3604 try using code_for_aarch64_sve instead.
3605 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3606 (svqsub_impl): Likewise.
3607 (svqadd, svqsub): Use rtx_code_function instead.
3609 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3611 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3612 (HADDSUB, sur, addsub): Remove them.
3614 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3616 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3617 functions the same way, rather than singling out those that
3618 aren't mapped directly to optabs.
3620 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3622 * target.def (compatible_vector_types_p): New target hook.
3623 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3624 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3625 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3626 * doc/tm.texi: Regenerate.
3627 * gimple-expr.c: Include target.h.
3628 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3629 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3631 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3632 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3633 Use the original predicate if it already has a suitable type.
3635 2020-01-09 Martin Jambor <mjambor@suse.cz>
3637 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3638 resolve_speculation and redirect_call_stmt_to_callee static. Change
3639 return type of set_call_stmt to cgraph_edge *.
3640 * auto-profile.c (afdo_indirect_call): Adjust call to
3641 redirect_call_stmt_to_callee.
3642 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3643 make the this pointer explicit, adjust self-recursive calls and the
3644 call top make_direct. Return the resulting edge.
3645 (cgraph_edge::remove): Make this pointer explicit.
3646 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3647 (cgraph_edge::make_direct): Likewise, adjust call to
3648 resolve_speculation.
3649 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3650 call to set_call_stmt.
3651 (cgraph_update_edges_for_call_stmt_node): Update call to
3652 set_call_stmt and remove.
3653 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3654 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3655 (cgraph_node::create_edge_including_clones): Moved "first" definition
3656 of edge to the block where it was used. Adjusted calls to
3658 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3659 cgraph_edge::remove.
3660 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3661 make_direct and redirect_call_stmt_to_callee.
3662 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3663 resolve_speculation and make_direct.
3664 * ipa-inline-transform.c (inline_transform): Adjust call to
3665 redirect_call_stmt_to_callee.
3666 (check_speculations_1):: Adjust call to resolve_speculation.
3667 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3668 resolve-speculation.
3669 (inline_small_functions): Adjust call to resolve_speculation.
3670 (ipa_inline): Likewise.
3671 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3673 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3674 safe with regards to edge removal, adjust calls to
3675 redirect_call_stmt_to_callee.
3676 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3677 and redirect_call_stmt_to_callee.
3678 * multiple_target.c (create_dispatcher_calls): Adjust call to
3679 redirect_call_stmt_to_callee
3680 (redirect_to_specific_clone): Likewise.
3681 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3682 Adjust calls to cgraph_edge::remove.
3683 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3684 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3685 (expand_call_inline): Adjust call to cgraph_edge::remove.
3687 2020-01-09 Martin Liska <mliska@suse.cz>
3689 * params.opt: Set Optimization for
3690 param_max_speculative_devirt_maydefs.
3692 2020-01-09 Martin Sebor <msebor@redhat.com>
3696 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3698 2020-01-09 Martin Liska <mliska@suse.cz>
3700 * auto-profile.c (auto_profile): Use opt_for_fn
3702 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3703 (propagate_vals_across_arith_jfunc): Likewise.
3704 (hint_time_bonus): Likewise.
3705 (incorporate_penalties): Likewise.
3706 (good_cloning_opportunity_p): Likewise.
3707 (perform_estimation_of_a_value): Likewise.
3708 (estimate_local_effects): Likewise.
3709 (ipcp_propagate_stage): Likewise.
3710 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3711 (set_switch_stmt_execution_predicate): Likewise.
3712 (analyze_function_body): Likewise.
3713 * ipa-inline-analysis.c (offline_size): Likewise.
3714 * ipa-inline.c (early_inliner): Likewise.
3715 * ipa-prop.c (ipa_analyze_node): Likewise.
3716 (ipcp_transform_function): Likewise.
3717 * ipa-sra.c (process_scan_results): Likewise.
3718 (ipa_sra_summarize_function): Likewise.
3719 * params.opt: Rename ipcp-unit-growth to
3720 ipa-cp-unit-growth. Add Optimization for various
3721 IPA-related parameters.
3723 2020-01-09 Richard Biener <rguenther@suse.de>
3726 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3728 2020-01-09 Richard Biener <rguenther@suse.de>
3730 PR tree-optimization/93040
3731 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3733 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3735 * common/config/avr/avr-common.c (avr_option_optimization_table)
3736 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3738 2020-01-09 Martin Liska <mliska@suse.cz>
3740 * cgraphclones.c (symbol_table::materialize_all_clones):
3741 Use cgraph_node::dump_name.
3743 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3746 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3747 output_operand_lossage instead of gcc_unreachable.
3748 * doc/md.texi (riscv f constraint): Fix typo.
3751 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3752 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3753 CONST_SCALAR_INT_P instead of CONST_INT_P.
3754 (*subv<mode>4_1): Rename to ...
3755 (subv<mode>4_1): ... this.
3756 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3757 define_insn_and_split patterns.
3758 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3761 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3763 * vec.c (class selftest::count_dtor): New class.
3764 (selftest::test_auto_delete_vec): New test.
3765 (selftest::vec_c_tests): Call it.
3766 * vec.h (class auto_delete_vec): New class template.
3767 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3769 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3771 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3773 2020-01-08 Jim Wilson <jimw@sifive.com>
3775 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3776 use of TLS_MODEL_LOCAL_EXEC when not pic.
3778 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3780 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3783 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3786 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3787 *stack_protect_set_3 peephole2): Also check that the second
3788 insns source is general_operand.
3791 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3792 predicate for output operand instead of register_operand.
3793 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3794 memory destination and non-memory operands[2].
3796 2020-01-08 Martin Liska <mliska@suse.cz>
3798 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3799 ::dump_asm_name instead of (::name or ::asm_name).
3800 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3801 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3802 (analyze_functions): Likewise.
3803 (expand_all_functions): Likewise.
3804 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3805 (propagate_bits_across_jump_function): Likewise.
3806 (dump_profile_updates): Likewise.
3807 (ipcp_store_bits_results): Likewise.
3808 (ipcp_store_vr_results): Likewise.
3809 * ipa-devirt.c (dump_targets): Likewise.
3810 * ipa-fnsummary.c (analyze_function_body): Likewise.
3811 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3812 (process_hsa_functions): Likewise.
3813 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3814 (set_alias_uids): Likewise.
3815 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3816 * ipa-inline.c (recursive_inlining): Likewise.
3817 (inline_to_all_callers_1): Likewise.
3818 (ipa_inline): Likewise.
3819 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3820 (ipa_propagate_frequency): Likewise.
3821 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3822 (remove_described_reference): Likewise.
3823 * ipa-pure-const.c (worse_state): Likewise.
3824 (check_retval_uses): Likewise.
3825 (analyze_function): Likewise.
3826 (propagate_pure_const): Likewise.
3827 (propagate_nothrow): Likewise.
3828 (dump_malloc_lattice): Likewise.
3829 (propagate_malloc): Likewise.
3830 (pass_local_pure_const::execute): Likewise.
3831 * ipa-visibility.c (optimize_weakref): Likewise.
3832 (function_and_variable_visibility): Likewise.
3833 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3834 (ipa_discover_variable_flags): Likewise.
3835 * lto-streamer-out.c (output_function): Likewise.
3836 (output_constructor): Likewise.
3837 * tree-inline.c (copy_bb): Likewise.
3838 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3839 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3841 2020-01-08 Richard Biener <rguenther@suse.de>
3844 * tree-eh.c (sink_clobbers): Update virtual operands for
3845 the first and last stmt only. Add a dry-run capability.
3846 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3847 after CFG manipulations and in RPO order to catch all
3848 secondary opportunities reliably.
3850 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3853 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3855 2019-01-08 Richard Biener <rguenther@suse.de>
3858 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3859 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3860 virtual operand, also updating SSA use.
3861 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3862 Update stmt after resetting virtual operand.
3863 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3864 * gimple-iterator.c (gsi_remove): When not removing the stmt
3865 permanently do not delink immediate uses or mark the stmt modified.
3867 2020-01-08 Martin Liska <mliska@suse.cz>
3869 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3870 (ipa_call_context::estimate_size_and_time): Likewise.
3871 (inline_analyze_function): Likewise.
3873 2020-01-08 Martin Liska <mliska@suse.cz>
3875 * cgraph.c (cgraph_node::dump): Use systematically
3878 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3880 Add -nodevicespecs option for avr.
3883 * config/avr/avr.opt (-nodevicespecs): New driver option.
3884 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3885 "-specs=device-specs/..." if that option is not set.
3886 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3888 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3890 Implement 64-bit double functions for avr.
3893 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3894 --with-double-comparison.
3895 * doc/install.texi: Document them.
3896 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3897 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3898 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3899 * doc/invoke.texi (AVR Built-in Macros): Document them.
3900 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3901 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3902 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3904 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3907 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3908 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3909 when only building rm-profile multilibs.
3911 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3914 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3915 lattice for a value to check.
3916 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3917 finite propagation in self-recursive scc.
3919 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3921 * ipa-inline.c (caller_growth_limits): Restore the AND.
3923 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3925 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3926 (VEC_ALLREG_ALT): New iterator.
3927 (VEC_ALLREG_INT_MODE): New iterator.
3928 (VCMP_MODE): New iterator.
3929 (VCMP_MODE_INT): New iterator.
3930 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3931 (vec_cmp<u>v64qidi): New define_expand.
3932 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3933 (vec_cmpu<mode>di_exec): New define_expand.
3934 (vec_cmp<u>v64qidi_exec): New define_expand.
3935 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3936 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3937 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3938 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3939 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3940 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3941 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3942 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3943 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3944 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3946 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3947 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3949 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3951 * config/gcn/constraints.md (DA): Update description and match.
3953 (Db): New constraint.
3954 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3956 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3957 Implement 'Db' mixed immediate type.
3958 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3959 (addcv64si3_dup<exec_vcc>): Delete.
3960 (subcv64si3<exec_vcc>): Rework constraints.
3961 (addv64di3): Rework constraints.
3962 (addv64di3_exec): Rework constraints.
3963 (subv64di3): Rework constraints.
3964 (addv64di3_dup): Delete.
3965 (addv64di3_dup_exec): Delete.
3966 (addv64di3_zext): Rework constraints.
3967 (addv64di3_zext_exec): Rework constraints.
3968 (addv64di3_zext_dup): Rework constraints.
3969 (addv64di3_zext_dup_exec): Rework constraints.
3970 (addv64di3_zext_dup2): Rework constraints.
3971 (addv64di3_zext_dup2_exec): Rework constraints.
3972 (addv64di3_sext_dup2): Rework constraints.
3973 (addv64di3_sext_dup2_exec): Rework constraints.
3975 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3977 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3978 existing target checks.
3980 2020-01-07 Richard Biener <rguenther@suse.de>
3982 * doc/install.texi: Bump minimal supported MPC version.
3984 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3986 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3987 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3988 * langhooks.c: Include stor-layout.h.
3989 (lhd_simulate_enum_decl): New function.
3990 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3991 handle_arm_sve_h for the LTO frontend.
3992 (register_vector_type): Cope with null returns from pushdecl.
3994 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3996 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3997 (aarch64_sve::nvectors_if_data_type): Replace with...
3998 (aarch64_sve::builtin_type_p): ...this.
3999 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4000 (find_vector_type): Delete.
4001 (add_sve_type_attribute): New function.
4002 (lookup_sve_type_attribute): Likewise.
4003 (register_builtin_types): Add an "SVE type" attribute to each type.
4004 (register_tuple_type): Likewise.
4005 (svbool_type_p, nvectors_if_data_type): Delete.
4006 (mangle_builtin_type): Use lookup_sve_type_attribute.
4007 (builtin_type_p): Likewise. Add an overload that returns the
4008 number of constituent vector and predicate registers.
4009 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4010 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4011 instead of aarch64_sve_argument_p.
4012 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4013 (aarch64_pass_by_reference): Likewise.
4014 (aarch64_function_value_1): Likewise.
4015 (aarch64_return_in_memory): Likewise.
4016 (aarch64_layout_arg): Likewise.
4018 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4020 PR tree-optimization/93156
4021 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4022 least significant bit is always clear.
4024 PR tree-optimization/93118
4025 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4026 simplifier with two intermediate conversions.
4028 2020-01-07 Martin Liska <mliska@suse.cz>
4030 * params.opt: Add Optimization for various parameters.
4032 2020-01-07 Martin Liska <mliska@suse.cz>
4035 * doc/extend.texi: Explain cloning for target_clone
4038 2020-01-07 Martin Liska <mliska@suse.cz>
4040 PR tree-optimization/92860
4041 * common.opt: Make in Optimization option
4042 as it is affected by -O0, which is an Optimization
4044 * tree-inline.c (tree_inlinable_function_p):
4045 Use opt_for_fn for warn_inline.
4046 (expand_call_inline): Likewise.
4048 2020-01-07 Martin Liska <mliska@suse.cz>
4050 PR tree-optimization/92860
4051 * common.opt: Make flag_ree as optimization
4054 2020-01-07 Martin Liska <mliska@suse.cz>
4056 PR optimization/92860
4057 * params.opt: Mark param_min_crossjump_insns with Optimization
4060 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4062 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4063 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4065 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4067 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4068 helper function to return the valid addressing formats for a given
4069 hard register and mode.
4070 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4072 * config/rs6000/constraints.md (Q constraint): Update
4074 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4077 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4078 Use 'Q' for doing vector extract from memory.
4079 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4081 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4082 doing vector extract from memory.
4083 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4084 extract from memory.
4086 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4087 for the offset being 34-bits when -mcpu=future is used.
4089 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4091 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4092 instead of cmpib_comparison_operator in cmpib patterns.
4093 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4094 of cmpib_comparison_operator. Revise comment.
4096 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4098 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4099 in an IFN_DIV_POW2 node to be equal.
4101 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4103 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4104 (vect_check_scalar_mask): ...this.
4105 (vectorizable_store, vectorizable_load): Update call accordingly.
4106 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4107 argument in calls to conditional internal functions.
4109 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4111 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4112 '0' matching inputs.
4113 (subv64di3_exec): Likewise.
4115 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4117 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4118 * doc/md.texi (movstr): Likewise.
4120 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4122 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4125 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4127 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4129 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4130 to a temporary file and use move-if-change to update the real
4131 file where necessary.
4133 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4135 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4136 rather than Upa for CPY /M.
4138 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4140 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4143 2020-01-06 Martin Liska <mliska@suse.cz>
4145 PR tree-optimization/92860
4146 * params.opt: Mark param_max_combine_insns with Optimization
4149 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4152 * config/i386/i386.md (SWIDWI): New mode iterator.
4153 (DWI, dwi): Add TImode variants.
4154 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4155 <general_hilo_operand> instead of <general_operand>. Use
4156 CONST_SCALAR_INT_P instead of CONST_INT_P.
4157 (*addv<mode>4_1): Rename to ...
4158 (addv<mode>4_1): ... this.
4159 (QWI): New mode attribute.
4160 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4161 define_insn_and_split patterns.
4162 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4164 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4165 <general_hilo_operand> instead of <general_operand>.
4166 (*addcarry<mode>_1): New define_insn.
4167 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4169 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4171 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4172 Use "call" instead of "set".
4174 2020-01-03 Martin Jambor <mjambor@suse.cz>
4177 * ipa-cp.c (print_all_lattices): Skip functions without info.
4179 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4182 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4183 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4184 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4185 for 'e' simd clones.
4188 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4190 (mprefer-vector-width=): Add Save.
4191 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4192 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4193 (ix86_debug_options, ix86_function_specific_print): Adjust
4194 ix86_target_string callers.
4195 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4196 (ix86_valid_target_attribute_tree): Likewise.
4197 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4198 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4199 ix86_target_string caller.
4202 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4203 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4204 instead of gen_int_shift_amount + convert_modes.
4206 PR rtl-optimization/93088
4207 * loop-iv.c (find_single_def_src): Punt after looking through
4208 128 reg copies for regs with single definitions. Move definitions
4211 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4213 * config/arm/arm-c.c (arm_cpu_builtins): Define
4214 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4215 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4216 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4217 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4218 * config/arm/arm-tables.opt: Regenerated.
4219 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4220 arm_arch_i8mm and arm_arch_bf16 when enabled.
4221 * config/arm/arm.h (TARGET_I8MM): New macro.
4222 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4223 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4224 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4225 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4226 (v8_6_a_simd_variants): New.
4227 (v8_*_a_simd_variants): Add i8mm and bf16.
4228 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4230 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4233 * predict.c (compute_function_frequency): Don't call
4234 warn_function_cold on functions that already have cold attribute.
4236 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4239 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4240 COMDAT group function labels in .data.rel.ro.local section.
4241 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4244 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4245 comparison_operator in B and S integer comparisons. Likewise, use
4246 ordered_comparison_operator instead of cmpib_comparison_operator in
4248 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4250 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4252 Update copyright years.
4254 * gcc.c (process_command): Update copyright notice dates.
4255 * gcov-dump.c (print_version): Ditto.
4256 * gcov.c (print_version): Ditto.
4257 * gcov-tool.c (print_version): Ditto.
4258 * gengtype.c (create_file): Ditto.
4259 * doc/cpp.texi: Bump @copying's copyright year.
4260 * doc/cppinternals.texi: Ditto.
4261 * doc/gcc.texi: Ditto.
4262 * doc/gccint.texi: Ditto.
4263 * doc/gcov.texi: Ditto.
4264 * doc/install.texi: Ditto.
4265 * doc/invoke.texi: Ditto.
4267 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4269 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4272 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4274 PR tree-optimization/93098
4275 * match.pd (popcount): For shift amounts, use integer_onep
4276 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4277 tests. Make sure that precision is power of two larger than or equal
4278 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4279 instead of ULL suffixed constants. Formatting fixes.
4281 Copyright (C) 2020 Free Software Foundation, Inc.
4283 Copying and distribution of this file, with or without modification,
4284 are permitted in any medium without royalty provided the copyright
4285 notice and this notice are preserved.