1 2019-03-18 Martin Jambor <mjambor@suse.cz>
4 2019-03-18 Martin Jambor <mjambor@suse.cz>
6 PR tree-optimization/89546
7 * tree-sra.c (propagate_subaccesses_across_link): Requeue new_acc if
8 any propagation to its children took place.
10 2019-03-17 H.J. Lu <hongjiu.lu@intel.com>
12 Backport from mainline
13 2019-03-14 H.J. Lu <hongjiu.lu@intel.com>
16 * config/i386/i386.c (ix86_print_operand): Handle 'M' to add
17 addr32 prefix to VSIB address for X32.
18 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask): Prepend
20 (*avx512pf_gatherpf<mode>df_mask): Likewise.
21 (*avx512pf_scatterpf<mode>sf_mask): Likewise.
22 (*avx512pf_scatterpf<mode>df_mask): Likewise.
23 (*avx2_gathersi<mode>): Prepend "%M3" to opcode.
24 (*avx2_gathersi<mode>_2): Prepend "%M2" to opcode.
25 (*avx2_gatherdi<mode>): Prepend "%M3" to opcode.
26 (*avx2_gatherdi<mode>_2): Prepend "%M2" to opcode.
27 (*avx2_gatherdi<mode>_3): Prepend "%M3" to opcode.
28 (*avx2_gatherdi<mode>_4): Prepend "%M2" to opcode.`
29 (*avx512f_gathersi<mode>): Prepend "%M4" to opcode.
30 (*avx512f_gathersi<mode>_2): Prepend "%M3" to opcode.
31 (*avx512f_gatherdi<mode>): Prepend "%M4" to opcode.
32 (*avx512f_gatherdi<mode>_2): Prepend "%M3" to opcode.
33 (*avx512f_scattersi<mode>): Prepend "%M0" to opcode.
34 (*avx512f_scatterdi<mode>): Likewise.
36 2019-03-15 Richard Biener <rguenther@suse.de>
38 Backport from mainline
39 2018-11-30 Jakub Jelinek <jakub@redhat.com>
42 * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
43 * tree-ssa-ifcombine.c (ifcombine_ifandif): If
44 --param logical-op-non-short-circuit is present, override
45 LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
46 * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
48 2019-03-15 Martin Liska <mliska@suse.cz>
51 * doc/invoke.texi: Remove -fdump-class-hierarchy option.
53 2019-03-14 Martin Jambor <mjambor@suse.cz>
55 Backport from mainline
56 2019-03-07 Martin Jambor <mjambor@suse.cz>
59 * ipa-cp.c (perform_estimation_of_a_value): Account zero time benefit
60 for extern inline functions.
62 2019-03-14 Richard Biener <rguenther@suse.de>
65 * tree-scalar-evolution.c (get_loop_exit_condition): Use
67 * tree-ssa-loop-ivcanon.c (canonicalize_loop_induction_variables):
68 Use gimple_location_safe.
70 2019-03-13 Vladimir Makarov <vmakarov@redhat.com>
73 * lra-constraints.c (inherit_in_ebb): Update
74 potential_reload_hard_regs along with live_hard_regs.
76 2019-03-14 Richard Biener <rguenther@suse.de>
78 Backport from mainline
79 2019-03-13 Richard Biener <rguenther@suse.de>
82 * tree-scalar-evolution.c (simplify_peeled_chrec): Do not
83 throw FP expressions at tree-affine.
85 2019-03-01 Richard Biener <rguenther@suse.de>
88 * tree-cfgcleanup.h (cleanup_tree_cfg): Add SSA update flags
89 argument, defaulted to zero.
90 * passes.c (execute_function_todo): Pass down SSA update flags
92 * tree-cfgcleanup.c: Include tree-into-ssa.h and tree-cfgcleanup.h.
93 (cleanup_tree_cfg_1): After cleanup_control_flow_pre update SSA
95 (cleanup_tree_cfg_noloop): Pass down SSA update flags.
96 (cleanup_tree_cfg): Get and pass down SSA update flags.
98 2019-02-18 Richard Biener <rguenther@suse.de>
100 PR tree-optimization/89296
101 * tree-ssa-loop-ch.c (ch_base::copy_headers): Restrict setting
102 of no-warning flag to cases that might emit the bogus warning.
104 2019-03-13 Andreas Krebbel <krebbel@linux.ibm.com>
106 Backport from mainline
107 2019-03-11 Andreas Krebbel <krebbel@linux.ibm.com>
109 * config/s390/s390.c (s390_expand_builtin): Do the copy_to_reg not
110 only on the else branch.
112 2019-03-13 Andreas Krebbel <krebbel@linux.ibm.com>
114 Backport from mainline
115 2019-02-07 Andreas Krebbel <krebbel@linux.ibm.com>
117 * config/s390/s390-builtin-types.def: Add new types.
118 * config/s390/s390-builtins.def: (s390_vec_xl, s390_vec_xld2)
119 (s390_vec_xlw4): Make the memory operand into a const pointer.
120 (s390_vec_xld2, s390_vec_xlw4): Add a variant for single precision
122 * config/s390/s390-c.c (s390_expand_overloaded_builtin): Generate
123 a new vector type with the alignment of the scalar memory operand.
125 2019-03-12 Richard Biener <rguenther@suse.de>
127 PR tree-optimization/89664
128 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Properly
129 free the occurance tree after the early out.
131 2019-03-12 Andre Vieira <andre.simoesdiasvieira@arm.com>
133 Backport from mainline
134 2019-03-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
136 * config/arm/arm.h (TARGET_FP16_TO_DOUBLE): Add TARGET_VFP_DOUBLE
139 2019-03-11 Eric Botcazou <ebotcazou@adacore.com>
141 PR rtl-optimization/89588
142 * loop-unroll.c (decide_unroll_constant_iterations): Make guard for
143 explicit unrolling factor more robust.
145 2019-03-11 Martin Liska <mliska@suse.cz>
147 Backport from mainline
148 2019-03-08 Martin Liska <mliska@suse.cz>
151 * config/i386/i386.c (ix86_option_override_internal): Disable
152 jump tables when retpolines are used.
154 2019-03-11 Martin Liska <mliska@suse.cz>
156 Backport from mainline
157 2019-02-11 David Malcolm <dmalcolm@redhat.com>
160 * input.c (selftest::test_line_offset_overflow): New selftest.
161 (selftest::input_c_tests): Call it.
163 2019-03-08 Martin Jambor <mjambor@suse.cz>
165 Backport from mainline
166 2019-03-07 Martin Jambor <mjambor@suse.cz>
169 * cgraph.h (cgraph_node): New inline method former_thunk_p.
170 * cgraph.c (cgraph_node::dump): Dump a note if node is a former thunk.
171 (clone_of_p): Treat expanded thunks like thunks, be optimistic if they
172 have multiple callees. At the end check if declarations match as
173 opposed to cgraph_nodes.
175 2019-03-06 Tamar Christina <tamar.christina@arm.com>
178 2019-02-25 Tamar Christina <tamar.christina@arm.com>
181 * common/config/aarch64/aarch64-common.c
182 (struct aarch64_option_extension): Add is_synthetic.
183 (all_extensions): Use it.
184 (TARGET_OPTION_INIT_STRUCT): Define hook.
185 (struct gcc_targetm_common): Moved to end.
186 (all_extensions_by_on): New.
187 (opt_ext_cmp, typedef opt_ext): New.
188 (aarch64_option_init_struct): New.
189 (aarch64_contains_opt): New.
190 (aarch64_get_extension_string_for_isa_flags): Output smallest set.
191 * config/aarch64/aarch64-option-extensions.def
192 (AARCH64_OPT_EXTENSION): Explicitly include AES and SHA2 in crypto.
193 (fp, simd, crc, lse, fp16, rcpc, rdma, dotprod, aes, sha2, sha3,
195 Set is_synthetic to false.
196 (crypto): Set is_synthetic to true.
197 * config/aarch64/driver-aarch64.c (AARCH64_OPT_EXTENSION): Add
200 2019-03-06 Tamar Christina <tamar.christina@arm.com>
203 2019-02-28 Tamar Christina <tamar.christina@arm.com>
206 * config/aarch64/aarch64-option-extensions.def: Document it.
207 * config/aarch64/driver-aarch64.c (host_detect_local_cpu): Skip feature
210 2019-03-06 Xiong Hu Luo <luoxhu@linux.ibm.com>
212 Backport of r268834 from mainline to gcc-8-branch.
213 2019-02-13 Xiong Hu Luo <luoxhu@linux.vnet.ibm.com>
215 * config/rs6000/altivec.h (vec_sbox_be, vec_cipher_be,
216 vec_cipherlast_be, vec_ncipher_be, vec_ncipherlast_be): New #defines.
217 * config/rs6000/crypto.md (CR_vqdi): New define_mode_iterator.
218 (crypto_vsbox_<mode>, crypto_<CR_insn>_<mode>): New define_insns.
219 * config/rs6000/rs6000-builtin.def (VSBOX_BE): New BU_CRYPTO_1.
220 (VCIPHER_BE, VCIPHERLAST_BE, VNCIPHER_BE, VNCIPHERLAST_BE):
222 * config/rs6000/rs6000.c (builtin_function_type)
223 <CRYPTO_BUILTIN_VSBOX_BE, CRYPTO_BUILTIN_VCIPHER_BE,
224 CRYPTO_BUILTIN_VCIPHERLAST_BE, CRYPTO_BUILTIN_VNCIPHER_BE,
225 CRYPTO_BUILTIN_VNCIPHERLAST_BE>: New switch options.
226 * doc/extend.texi (vec_sbox_be, vec_cipher_be, vec_cipherlast_be,
227 vec_ncipher_be, vec_ncipherlast_be): New builtin functions.
229 2019-03-05 Richard Biener <rguenther@suse.de>
231 Backport from mainline
232 2019-02-26 Richard Biener <rguenther@suse.de>
234 PR tree-optimization/89505
235 * tree-ssa-structalias.c (compute_dependence_clique): Make sure
236 to handle restrict pointed-to vars with multiple subvars
239 2019-03-01 Tamar Christina <tamar.christina@arm.com>
242 * config/aarch64/aarch64-option-extensions.def (fp, simd, crypto,
243 fp16): Collapse line.
245 2019-03-01 Richard Biener <rguenther@suse.de>
247 Backport from mainline
248 2019-03-01 Jakub Jelinek <jakub@redhat.com>
251 * dwarf2out.c (output_comdat_type_unit): Add ATTRIBUTE_UNUSED to
252 early_lto_debug argument.
254 2019-02-27 Richard Biener <rguenther@suse.de>
257 * dwarf2out.c (use_debug_types): Disable when in_lto_p.
259 2019-02-27 Richard Biener <rguenther@suse.de>
262 * dwarf2out.c (output_comdat_type_unit): Add early_lto_debug
263 parameter, prefix section name with .gnu.debuglto_ if true.
264 (dwarf2out_finish): Pass false to output_comdat_type_unit.
265 (dwarf2out_early_finish): Pass true to output_comdat_type_unit.
267 2019-02-27 Richard Biener <rguenther@suse.de>
270 * dwarf2out.c (size_of_die): Key on AT_ref (a)->comdat_type_p
271 rather than on use_debug_types, doing what output_die does.
272 (value_format): Likewise.
274 2019-02-28 Eric Botcazou <ebotcazou@adacore.com>
276 PR tree-optimization/89536
277 * tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_NOT_EXPR>: Test
278 only whether bit #0 of the value is 0 instead of the entire value.
280 2019-02-28 Li Jia He <helijia@linux.ibm.com>
283 2019-02-20 Li Jia He <helijia@linux.ibm.com>
286 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin)
287 <case ALTIVEC_BUILTIN_VSPLTISB, ALTIVEC_BUILTIN_VSPLTISH,
288 ALTIVEC_BUILTIN_VSPLTISW>: Don't convert the operand before
291 2019-02-27 Uroš Bizjak <ubizjak@gmail.com>
294 * config/i386/i386.c (ix86_atomic_assign_expand_fenv): Check
295 TARGET_SSE in addition to TARGET_SSE_MATH.
297 (ix86_excess_precision): Ditto.
298 (ix86_float_exceptions_rounding_supported_p): Ditto.
299 (use_rsqrt_p): Ditto.
300 * config/i386/sse.md (rsqrt<mode>2): Ditto.
302 2019-02-27 Jakub Jelinek <jakub@redhat.com>
304 Backported from mainline
305 2019-02-18 Jakub Jelinek <jakub@redhat.com>
308 * config/s390/s390.c (s390_indirect_branch_attrvalue,
309 s390_indirect_branch_settings): Define unconditionally.
310 (s390_set_current_function): Likewise, but guard the whole body except
311 the s390_indirect_branch_settings call with
312 #if S390_USE_TARGET_ATTRIBUTE.
313 (TARGET_SET_CURRENT_FUNCTION): Redefine unconditionally.
315 2019-02-26 Eric Botcazou <ebotcazou@adacore.com>
317 Backport from mainline
318 2019-02-19 Eric Botcazou <ebotcazou@adacore.com>
320 * rtlanal.c (get_initial_register_offset): Fall back to the estimate
321 as long as the epilogue isn't completed.
323 2019-02-26 Eric Botcazou <ebotcazou@adacore.com>
325 * tree-ssa-dom.c (edge_info::derive_equivalences) <BIT_IOR_EXPR>: Fix
326 and move around comment.
327 <BIT_AND_EXPR>: Likewise.
328 <BIT_NOT_EXPR>: Add specific handling for boolean types.
330 2019-02-24 Alan Modra <amodra@gmail.com>
333 * config/rs6000/rs6000.md (<bd>_<mode> split): Check for an int
334 output reg on add insn.
335 (<bd>tf_<mode> split): Likewise. Match predicates with insn.
337 2019-02-22 Jakub Jelinek <jakub@redhat.com>
339 * BASE-VER: Set to 8.3.1.
341 2019-02-22 Release Manager
343 * GCC 8.3.0 released.
345 2019-02-15 Richard Biener <rguenther@suse.de>
346 Jakub Jelinek <jakub@redhat.com>
348 PR tree-optimization/89278
349 * tree-loop-distribution.c: Include tree-eh.h.
350 (generate_memset_builtin, generate_memcpy_builtin): Call
351 rewrite_to_non_trapping_overflow on builtin->size before passing it
352 to force_gimple_operand_gsi.
354 2019-02-15 Jakub Jelinek <jakub@redhat.com>
357 * optc-save-gen.awk: Handle optimize_fast like optimize_size or
359 * opth-gen.awk: Likewise.
361 2019-02-14 David Malcolm <dmalcolm@redhat.com>
363 Backport of r267957 from trunk.
364 2019-01-15 David Malcolm <dmalcolm@redhat.com>
367 * tree.c (build_function_type): Assert that arg_types is not
370 2019-02-14 Jakub Jelinek <jakub@redhat.com>
372 PR rtl-optimization/89354
373 * combine.c (make_extraction): Punt if extraction_mode is narrower
376 2019-02-14 David Malcolm <dmalcolm@redhat.com>
378 Backport of r263339 from trunk.
379 2018-08-06 David Malcolm <dmalcolm@redhat.com>
381 * function-tests.c (selftest::test_expansion_to_rtl): Call
382 free_after_compilation.
384 2019-02-14 David Malcolm <dmalcolm@redhat.com>
386 Backport of r263295 from trunk.
387 2018-08-03 David Malcolm <dmalcolm@redhat.com>
389 * doc/gcov.texi (-x): Remove duplicate "to".
390 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
391 (-Wif-not-aligned): Remove duplicate "is".
392 (-flto): Remove duplicate "the".
393 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
395 (MSP430 Options): Remove duplicate "and" from the description
396 of "-mgprel-sec=regexp".
397 (x86 Options): Remove duplicate copies of "vmldLog102" and
398 vmlsLog104 from description of "-mveclibabi=type".
400 2019-02-14 David Malcolm <dmalcolm@redhat.com>
402 Backport of r263275 from trunk.
403 2018-08-02 David Malcolm <dmalcolm@redhat.com>
405 * diagnostic-show-locus.c (diagnostic_show_locus): Use
406 pp_take_prefix when saving the existing prefix.
407 * diagnostic.c (diagnostic_append_note): Likewise.
408 * langhooks.c (lhd_print_error_function): Likewise.
409 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
410 param's type. Free the existing prefix.
411 (pp_take_prefix): New function.
412 (pretty_printer::pretty_printer): Drop the prefix parameter.
413 Rename the length parameter to match the comment.
414 (pretty_printer::~pretty_printer): Free the prefix.
415 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
417 (struct pretty_printer): Drop the "const" from "prefix" field's
418 type and clarify memory management.
419 (pp_set_prefix): Drop the "const" from the 2nd param.
420 (pp_take_prefix): New decl.
422 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
425 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
429 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
431 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
434 2019-01-18 Segher Boessenkool <segher@kernel.crashing.org>
437 * config/rs6000/rs6000.md (*movsi_from_df): Allow only register
440 2019-02-14 Segher Boessenkool <segher@kernel.crashing.org>
443 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
445 PR rtl-optimization/85805
446 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
447 value for hard registers if that was written in the same mode.
449 2019-02-14 Martin Liska <mliska@suse.cz>
451 Backport from mainline
452 2019-02-14 Martin Liska <mliska@suse.cz>
454 PR rtl-optimization/89242
455 * dce.c (delete_unmarked_insns): Call free_dominance_info we
456 process a transformation.
458 2019-02-14 Martin Liska <mliska@suse.cz>
460 Backport from mainline
461 2019-02-13 Martin Liska <mliska@suse.cz>
464 * cfgrtl.c (remove_barriers_from_footer): New function.
465 (try_redirect_by_replacing_jump): Use it.
466 (cfg_layout_redirect_edge_and_branch): Likewise.
468 2019-02-14 Martin Liska <mliska@suse.cz>
470 Backport from mainline
471 2019-02-11 Martin Liska <mliska@suse.cz>
474 * ipa-cp.c (build_toporder_info): Remove usage of a param.
475 * ipa-inline.c (inline_small_functions): Likewise.
476 * ipa-pure-const.c (propagate_pure_const): Likewise.
477 (propagate_nothrow): Likewise.
478 * ipa-reference.c (propagate): Likewise.
479 * ipa-utils.c (struct searchc_env): Remove unused field.
480 (searchc): Always search across AVAIL_INTERPOSABLE.
481 (ipa_reduced_postorder): Always allow AVAIL_INTERPOSABLE as
482 the only called IPA pure const can properly not propagate
483 across interposable boundary.
484 * ipa-utils.h (ipa_reduced_postorder): Remove param.
486 2019-02-14 Jakub Jelinek <jakub@redhat.com>
488 PR tree-optimization/89314
489 * fold-const.c (fold_binary_loc): Cast strlen argument to
490 const char * before dereferencing it. Formatting fixes.
492 Backported from mainline
493 2019-02-13 Jakub Jelinek <jakub@redhat.com>
496 * tree-ssa-structalias.c (set_uids_in_ptset): Or in vi->is_heap_var
497 into pt->vars_contains_escaped_heap instead of setting
498 pt->vars_contains_escaped_heap to it.
501 * optabs.c (prepare_cmp_insn): Use UINTVAL (size) instead of
502 INTVAL (size), compare it to GET_MODE_MASK instead of
503 1 << GET_MODE_BITSIZE.
506 * config/i386/predicates.md (x86_64_immediate_operand): Allow
507 TLS UNSPECs offsetted by signed 32-bit CONST_INT even with
510 2019-02-09 Jakub Jelinek <jakub@redhat.com>
513 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
514 If !node->definition and TYPE_ARG_TYPES is non-NULL, use
515 TYPE_ARG_TYPES instead of DECL_ARGUMENTS.
517 2019-02-13 Bill Schmidt <wschmidt@linux.ibm.com>
519 Backport from mainline
520 2019-02-11 Bill Schmidt <wschmidt@linux.ibm.com>
522 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Shift-right
523 and shift-left vector built-ins need to include a TRUNC_MOD_EXPR
524 for correct semantics.
526 2019-02-13 Richard Biener <rguenther@suse.de>
528 Backport from mainline
529 2019-01-08 Richard Biener <rguenther@suse.de>
531 PR tree-optimization/86554
532 * tree-ssa-sccvn.c (visit_nary_op): When value-numbering to
533 expressions with different overflow behavior make sure there's an
534 available expression on the path.
536 2019-02-13 Martin Liska <mliska@suse.cz>
539 * lto-streamer.h (LTO_minor_version): Bump version due
542 2019-02-13 Richard Biener <rguenther@suse.de>
544 Backport from mainline
545 2019-02-12 Richard Biener <rguenther@suse.de>
547 PR tree-optimization/89253
548 * tree-ssa-loop-split.c (tree_ssa_split_loops): Check we can
551 2019-02-08 Richard Biener <rguenther@suse.de>
554 * tree-data-ref.c (initialize_matrix_A): Fail if constant
556 (analyze_subscript_affine_affine): Handle failure from
559 2019-01-28 Richard Biener <rguenther@suse.de>
561 PR tree-optimization/88739
562 * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid generating
563 BIT_FIELD_REFs of non-mode-precision integral operands.
565 2019-02-12 Jan Hubicka <hubicka@ucw.cz>
567 Backport from mainline:
568 2019-02-09 Jan Hubicka <hubicka@ucw.cz>
570 * ipa-visibility.c (localize_node): Also do not localize
571 LDPR_PREVAILING_DEF_IRONLY_EXP.
573 2019-02-11 Stefan Agner <stefan@agner.ch>
575 Backport from mainline.
576 2019-01-10 Stefan Agner <stefan@agner.ch>
579 * config/arm/arm.c (arm_option_override_internal): Force
580 opts->x_inline_asm_unified to true only if TARGET_THUMB2_P.
582 2019-02-09 Jan Hubicka <hubicka@ucw.cz>
584 Backport from mainline:
585 2019-01-05 Jan Hubicka <hubicka@ucw.cz>
589 2017-05-22 Jan Hubicka <hubicka@ucw.cz>
591 * ipa-inline.c (edge_badness): Use inlined_time instead of
592 inline_summaries->get.
594 2019-02-09 Aaron Sawdey <acsawdey@linux.ibm.com>
596 Backported from mainline
597 2019-02-05 Aaron Sawdey <acsawdey@linux.ibm.com>
600 * config/rs6000/rs6000.md (<bd>tf_<mode>): Generate a local label
601 for the long branch case.
603 2019-02-05 Aaron Sawdey <acsawdey@linux.ibm.com>
606 * config/rs6000/rs6000-string.c (do_ifelse, expand_cmp_vec_sequence,
607 expand_compare_loop, expand_block_compare_gpr,
608 expand_strncmp_align_check, expand_strncmp_gpr_sequence): Insert
609 REG_BR_PROB notes in inline expansion of memcmp/strncmp. Add
610 #include "profile-count.h" and "predict.h" for types and functions
611 needed to work with REG_BR_PROB notes.
613 2019-02-09 Aaron Sawdey <acsawdey@linux.ibm.com>
615 * config/rs6000/rs6000-string.c (expand_compare_loop,
616 expand_block_compare): Insert REG_BR_PROB notes in inline expansion of
619 2019-02-09 Jakub Jelinek <jakub@redhat.com>
621 Backported from mainline
622 2019-01-16 David Malcolm <dmalcolm@redhat.com>
625 * combine.c (delete_noop_moves): Convert to "bool" return,
626 returning true if any edges are eliminated.
627 (combine_instructions): Also return true if delete_noop_moves
630 2019-02-08 Jakub Jelinek <jakub@redhat.com>
632 PR rtl-optimization/89234
633 * except.c (copy_reg_eh_region_note_forward): Return if note_or_insn
634 is a NOTE, CODE_LABEL etc. - rtx_insn * other than INSN_P.
635 (copy_reg_eh_region_note_backward): Likewise.
637 2019-02-09 Alan Modra <amodra@gmail.com>
640 * config/rs6000/rs6000.c (save_reg_p): Match logic in
641 rs6000_emit_prologue emitting pic_offset_table setup.
643 2019-02-08 Richard Biener <rguenther@suse.de>
645 Backport from mainline
646 2018-11-13 Richard Biener <rguenther@suse.de>
648 PR tree-optimization/86991
649 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
650 group building until we have successfully detected the SLP
652 (vect_is_simple_reduction): Remove fixup code here.
654 2019-02-08 Richard Biener <rguenther@suse.de>
656 Backport from mainline
657 2019-02-03 Richard Biener <rguenther@suse.de>
660 * dwarf2out.c (copy_ancestor_tree): Register non-stubs as
663 2019-01-29 Richard Biener <rguenther@suse.de>
666 * dwarf2out.c (collect_skeleton_dies): New helper.
667 (copy_decls_for_unworthy_types): Call it.
668 (build_abbrev_table): Do not try to replace
669 DW_AT_signature refs with local refs.
671 2019-01-17 Richard Biener <rguenther@suse.de>
674 * dwarf2out.c (want_pubnames): Never generate pubnames sections
675 and friends for the LTO part of debug info.
677 2019-01-24 Richard Biener <rguenther@suse.de>
680 * tree-streamer-out.c (write_ts_decl_common_tree_pointers):
681 When in "legacy" debug mode make sure to reset self-origins.
683 2019-01-15 Richard Biener <rguenther@suse.de>
686 * dwarf2out.c (gen_member_die): Do not generate inheritance
689 2018-11-22 Richard Biener <rguenther@suse.de>
693 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
694 which can appear in size expressions.
695 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
696 CALL_EXPR_BY_DESCRIPTOR.
697 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
698 * tree-streamer-out.c (pack_ts_base_value_fields): Stream
699 CALL_EXPR_BY_DESCRIPTOR.
700 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
702 2019-02-08 Christophe Lyon <christophe.lyon@linaro.org>
704 Backport from mainline
705 2019-01-07 Christophe Lyon <christophe.lyon@linaro.org>
708 * doc/install.texi (with-multilib-list): Document for aarch64.
710 2019-02-08 Richard Biener <rguenther@suse.de>
712 Backport from mainline
713 2018-12-10 Richard Biener <rguenther@suse.de>
715 PR tree-optimization/88427
716 * vr-values.c (vr_values::extract_range_from_phi_node):
717 Handle symbolic ranges conservatively when trying to drop
720 2018-11-28 Richard Biener <rguenther@suse.de>
722 PR tree-optimization/88217
723 * vr-values.c (vr_values::extract_range_from_phi_node): Make
724 sure to handle results > +INF and < -INF correctly when
725 trying to drop down to +INF - 1 or -INF + 1.
727 2018-11-23 Richard Biener <rguenther@suse.de>
729 PR tree-optimization/88149
730 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
731 the case where there are two different def types for the
732 same operand at different operand position in the same stmt.
734 2019-02-07 Jakub Jelinek <jakub@redhat.com>
736 Backported from mainline
737 2019-02-05 Jakub Jelinek <jakub@redhat.com>
740 * dce.c (delete_unmarked_insns): Don't remove no-op moves if they
741 can throw, non-call exceptions are enabled and we can't delete
742 dead exceptions or alter cfg. Set must_clean if
743 delete_insn_and_edges returns true, don't set it blindly for calls.
745 PR rtl-optimization/89195
746 * combine.c (make_extraction): For MEMs, don't extract bytes outside
750 * optabs.c (prepare_cmp_insn): Pass x and y to
751 emit_block_comp_via_libcall rather than XEXP (x, 0) and XEXP (y, 0).
753 2019-02-02 Jakub Jelinek <jakub@redhat.com>
756 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
757 Punt with warning on aggregate return or argument types. Ignore
758 type/mode checking for uniform arguments.
760 2019-02-01 Jakub Jelinek <jakub@redhat.com>
762 PR tree-optimization/88107
763 * tree-cfg.c (find_outermost_region_in_block): Add ALL argument,
764 instead of assertion that eh_region_outermost is non-NULL, if it
765 is NULL, set *ALL to true and return NULL.
766 (move_sese_region_to_fn): Adjust caller, if all is set, call
767 duplicate_eh_regions with NULL region.
769 2019-01-31 Jakub Jelinek <jakub@redhat.com>
772 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Allow inlining
773 always_inline callees into no_sanitize_address callers.
775 2019-01-30 Jakub Jelinek <jakub@redhat.com>
778 * config/i386/i386.c (ix86_warn_parameter_passing_abi): Don't warn
779 for arguments to functions that are TU-local and shouldn't be
780 referenced by assembly.
782 2019-01-29 Jakub Jelinek <jakub@redhat.com>
786 * omp-simd-clone.c (simd_clone_clauses_extract)
787 <case OMP_CLAUSE_ALIGNED>: Ignore clauses with NULL
788 OMP_CLAUSE_ALIGNED_ALIGNMENT.
790 2019-01-28 Jakub Jelinek <jakub@redhat.com>
793 * gimplify.c (gimplify_omp_for): When adding OMP_CLAUSE_*_GIMPLE_SEQ
794 for lastprivate/linear IV, push gimplify context around gimplify_assign
795 and, if it needed any temporaries, pop it into a gimple bind around the
799 * doc/invoke.texi (-mclwb, -mprfchw, -mrdpid, -mrdseed, -msgx,
800 -madx, -mhle, -mavx5124fmaps, -mavx512vnni, -mavx5124vnniw): Document
802 (bmi2): Add missing @opindex.
803 * doc/extend.texi (x86 target attribute): Move fma4, lwp, ssse3
804 options alphabetically. Add missing 3dnow, 3dnowa, adx, avx, avx2,
805 avx5124fmaps, avx5124vnniw, avx512bitalg, avx512bw, avx512cd,
806 avx512dq, avx512er, avx512f, avx512ifma, avx512pf, avx512vbmi,
807 avx512vbmi2, avx512vl, avx512vnni, avx512vpopcntdq, bmi, bmi2,
808 clflushopt, clwb, clzero, crc32, cx16, f16c, fma, fsgsbase,
809 fxsr, gfni, hle, lzcnt, movbe, movdir64b, movdiri, mwaitx, pconfig,
810 pku, prefetchwt1, prfchw, rdpid, rdrnd, rdseed, rtm, sahf,
811 sgx, sha, shstk, tbm, vaes, vpclmulqdq, wbnoinvd, xsave,
812 xsavec, xsaveopt and xsaves options.
814 2019-01-27 Jakub Jelinek <jakub@redhat.com>
818 (<mask_codefor>avx512dq_shuf_<shuffletype>64x2_1<mask_name>,
819 avx512f_shuf_<shuffletype>64x2_1<mask_name>): Ensure the
820 first constants in pairs are multiples of 2. Formatting fixes.
821 (avx512vl_shuf_<shuffletype>32x4_1<mask_name>,
822 avx512vl_shuf_<shuffletype>32x4_1<mask_name>): Ensure the
823 first constants in each quadruple are multiples of 4. Formatting fixes.
825 2019-01-24 Jakub Jelinek <jakub@redhat.com>
827 PR tree-optimization/88964
828 * gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Also
829 punt if HONOR_SNANS (chrec).
831 2019-01-23 Jakub Jelinek <jakub@redhat.com>
833 PR tree-optimization/88964
834 * gimple-loop-interchange.cc (loop_cand::analyze_induction_var): Use
835 build_zero_cst instead of build_int_cst. Return false for loop
836 invariants which honor signed zeros.
838 2019-01-22 Jakub Jelinek <jakub@redhat.com>
841 * config/rs6000/rs6000.c: Include tree-vrp.h and tree-ssanames.h.
842 (rs6000_gimple_fold_builtin): If MEM_REF address doesn't satisfy
843 is_gimple_mem_ref_addr predicate, force it into a SSA_NAME first.
846 * gimplify.c (gimplify_omp_atomic): Handle bitfield atomics with
847 non-integral DECL_BIT_FIELD_REPRESENTATIVEs.
850 * optabs.c (add_equal_note): Add op0_mode argument, use it instead of
852 (expand_binop_directly, expand_doubleword_clz,
853 expand_doubleword_popcount, expand_ctz, expand_ffs,
854 expand_unop_direct, maybe_emit_unop_insn): Adjust callers.
856 PR rtl-optimization/49429
858 PR rtl-optimization/86334
860 * expr.c (emit_block_move_hints): Move marking of MEM_EXPRs
861 addressable from here...
862 (emit_block_op_via_libcall): ... to here.
864 2019-01-17 Jakub Jelinek <jakub@redhat.com>
867 * config/aarch64/arm_neon.h: Fix #pragma GCC target syntax - replace
868 (("..."))) with ("..."). Use arch=armv8.2-a+sha3 instead of
869 arch=armv8.2-a+crypto for vsha512hq_u64 etc. intrinsics.
871 PR rtl-optimization/88870
872 * dce.c (deletable_insn_p): Never delete const/pure calls that can
873 throw if we can't alter the cfg or delete dead exceptions.
874 (mark_insn): Don't call find_call_stack_args for such calls.
876 2019-01-11 Jakub Jelinek <jakub@redhat.com>
879 * passes.c (finish_optimization_passes): Call print_combine_total_stats
880 inside of pass_combine_1 dump rather than pass_profile_1.
882 2019-01-10 Jakub Jelinek <jakub@redhat.com>
885 * attribs.c (handle_dll_attribute): Clear TREE_STATIC after setting
888 2019-01-07 Jakub Jelinek <jakub@redhat.com>
891 * dwarf2out.c (const_ok_for_output_1): Remove redundant call to
892 const_not_ok_for_debug_p target hook.
893 (mem_loc_descriptor) <case UNSPEC>: Only call const_ok_for_output_1
894 on UNSPEC and subexpressions thereof if all subexpressions of the
895 UNSPEC are CONSTANT_P.
897 2019-01-05 Jakub Jelinek <jakub@redhat.com>
900 * dwarf2out.c (const_ok_for_output_1): Reject MINUS that contains
901 SYMBOL_REF, CODE_LABEL or UNSPEC in subexpressions of second argument.
902 Reject PLUS that contains SYMBOL_REF, CODE_LABEL or UNSPEC in
903 subexpressions of both operands.
904 (mem_loc_descriptor): Handle UNSPEC if target hook acks it and all the
905 subrtxes are CONSTANT_P.
907 2019-02-07 Richard Biener <rguenther@suse.de>
909 Backport from mainline
910 2018-11-20 Richard Biener <rguenther@suse.de>
912 PR tree-optimization/88105
913 * tree-ssa-dom.c (pass_dominator::execute): Do not walk
916 2018-11-28 Richard Biener <rguenther@suse.de>
918 PR tree-optimization/88223
919 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
920 over a stored-same value may-alias store make sure to consider
921 partial overlaps which are valid when TBAA reasonings do not
922 apply and byte-granular overlaps are possible at all.
924 2019-02-07 Richard Biener <rguenther@suse.de>
926 Backport from mainline
927 2018-12-11 Richard Biener <rguenther@suse.de>
931 * tree-complex.c (update_complex_assignment): Properly transfer
932 or clean EH info around gimple_assign_set_rhs_with_ops.
934 2018-11-15 Richard Biener <rguenther@suse.de>
936 PR tree-optimization/88030
937 * tree-complex.c (need_eh_cleanup): New global.
938 (update_complex_assignment): Mark blocks that need EH update.
939 (expand_complex_comparison): Likewise.
940 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
941 perform EH cleanup and schedule CFG cleanup if that did anything.
943 2018-11-08 Richard Biener <rguenther@suse.de>
945 PR tree-optimization/87929
946 * tree-complex.c (expand_complex_comparison): Clean EH.
948 2019-02-06 Eric Botcazou <ebotcazou@adacore.com>
950 * config/i386/i386.c (ix86_expand_prologue): Emit a memory blockage
951 after restoring registers saved to allocate the frame on Windows.
953 2019-02-06 Richard Biener <rguenther@suse.de>
955 Backport from mainline
956 2019-01-31 Richard Biener <rguenther@suse.de>
958 PR tree-optimization/89135
959 * tree-ssa-phiprop.c (pass_phiprop::execute): Skip blocks
962 2019-01-18 Richard Biener <rguenther@suse.de>
964 PR tree-optimization/88903
965 * tree-vect-stmts.c (vectorizable_shift): Verify we see all
966 scalar stmts a SLP shift amount is composed of when detecting
969 2019-02-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
971 Backport from mainline.
972 2019-01-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
973 * config/rs6000/rs6000-c.c (altivec-resolve_overloaded_builtin):
974 Change handling of ALTIVEC_BUILTIN_VEC_EXTRACT. Coerce result to
975 type of vector element when vec_extract is implemented by direct
978 2019-02-05 Andreas Krebbel <krebbel@linux.ibm.com>
980 Backport from mainline
981 2019-02-05 Andreas Krebbel <krebbel@linux.ibm.com>
984 * config/s390/s390.md: Remove load and test FP splitter.
986 2019-02-04 Bill Schmidt <wshmidt@linux.ibm.com>
989 Backport from mainline
991 2019-01-30 Bill Schmidt <wschmidt@linux.ibm.com>
994 * config/rs6000/vsx.md (*vsx_reduc_<VEC_reduc_name>_v4sf_scalar):
995 Disable for little-endian.
997 2019-01-22 Jakub Jelinek <jakub@redhat.com>
1000 * config/rs6000/vsx.md (*vsx_reduc_<VEC_reduc_name>_v2df_scalar):
1001 Disable for little endian.
1003 2019-02-04 Eric Botcazou <ebotcazou@adacore.com>
1005 * config/sparc/sparc.h: Remove superfluous blank lines.
1006 * config/sparc/sparc.c (global_offset_table_rtx): Rename into...
1007 (got_register_rtx): ...this.
1008 (sparc_got): Adjust to above renaming.
1009 (sparc_tls_got): Likewise.
1010 (sparc_output_mi_thunk): Likewise.
1011 (sparc_init_pic_reg): Likewise.
1012 (save_local_or_in_reg_p): Fix test on the GOT register.
1013 (USE_HIDDEN_LINKONCE): Move around.
1014 (get_pc_thunk_name): Likewise.
1015 (gen_load_pcrel_sym): Likewise.
1016 (load_got_register): Likewise.
1018 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1020 Backport from mainline
1021 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
1023 PR tree-optimization/87022
1024 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1025 bits in dist vector rather than the first one.
1027 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1029 Backport from mainline
1030 2019-02-01 Bin Cheng <bin.cheng@linux.alibaba.com>
1032 PR tree-optimization/88932
1033 * tree-predcom.c (try_combine_chains): Get loop bbs in dom order.
1035 2018-01-31 Bill Schmidt <wschmidt@linux.ibm.com>
1037 Backport from mainline
1038 2018-01-31 Bill Schmidt <wschmidt@linux.ibm.com>
1040 PR tree-optimization/89008
1041 * gimple-ssa-strength-reduction.c (slsr_process_mul): Don't
1042 process anything of the form X * 0.
1044 2019-01-31 Richard Biener <rguenther@suse.de>
1046 Backport from mainline
1047 2019-01-31 Richard Biener <rguenther@suse.de>
1049 PR rtl-optimization/89115
1050 * lra.c (lra_rtx_hash): Properly hash CONST_INT values.
1052 2019-01-30 Richard Biener <rguenther@suse.de>
1054 PR rtl-optimization/89115
1055 * opts.c (default_options_optimization): Reduce
1056 PARAM_MAX_DSE_ACTIVE_LOCAL_STORES by a factor of 10 at -O1.
1057 Make PARAM_LOOP_INVARIANT_MAX_BBS_IN_LOOP reduction relative
1060 2019-01-30 Kewen Lin <linkw@gcc.gnu.org>
1062 Backport from mainline.
1063 2019-01-17 Kewen Lin <linkw@gcc.gnu.org>
1065 * doc/extend.texi: Add four new prototypes for vec_ld and seven new
1066 prototypes for vec_st.
1067 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add entries
1068 for scalar address type variants of altivec_vec_ld/altivec_vec_st,
1069 mainly on signed/unsigned long long and double.
1071 2019-01-29 Vineet Gupta <vgupta@synopsys.com>
1073 Backport from mainline.
1074 * gcc/config.gcc: Force .init_array for ARC.
1076 2019-01-29 Vineet Gupta <vgupta@synopsys.com>
1078 Backport from mainline.
1079 * config/arc/atomic.md: Add operand to DMB instruction.
1081 2019-01-27 Uroš Bizjak <ubizjak@gmail.com>
1084 * rtl.h (prepare_copy_insn): New prototype.
1085 * gcse.c (prepare_copy_insn): New function, split out from
1086 process_insert_insn.
1087 (process_insert_insn): Use prepare_copy_insn.
1088 * store-motion.c (replace_store_insn): Use prepare_copy_insn
1089 instead of gen_move_insn.
1091 2019-01-25 Richard Biener <rguenther@suse.de>
1093 PR tree-optimization/86865
1094 * graphite-scop-detection.c (scop_detection::can_represent_loop):
1095 Reject non-do-while loops.
1097 2019-01-24 Uroš Bizjak <ubizjak@gmail.com>
1100 * config/i386/sse.md (sse2_cvtpi2pd): Add SSE alternatives.
1101 Disparage MMX alternative.
1102 (sse2_cvtpd2pi): Ditto.
1103 (sse2_cvttpd2pi): Ditto.
1105 2019-01-24 Richard Earnshaw <rearnsha@arm.com>
1108 * profile-count.h (profile_count): On ARM systems using GCC 6/7/8
1109 force the alignment of m_val.
1111 2019-01-24 Richard Biener <rguenther@suse.de>
1113 Backport from mainline
1114 2019-01-23 Richard Biener <rguenther@suse.de>
1116 PR tree-optimization/89008
1117 * tree-ssa-reassoc.c (eliminate_using_constants): For * 0 do
1118 not leave another stray operand.
1120 2019-01-22 Uroš Bizjak <ubizjak@gmail.com>
1123 * config/i386/i386.c (ix86_expand_builtin) [case IX86_BUILTIN_BEXTRI32,
1124 case IX86_BUILTIN_BEXTRI64]: Sanitize operands.
1126 2019-01-18 Richard Earnshaw <rearnsha@arm.com>
1129 * config/arm/arm-cpus.in (mp): New feature.
1131 (fgroup ARMv7ve): Add mp and sec features.
1132 (arch armv7-a): Add options to allow mp and sec extensions.
1133 (cpu generic-armv7-a): Add options to allow mp and sec extensions.
1134 (cpu cortex-a5, cpu cortex-7, cpu cortex-a9): Add mp and sec
1135 extenstions to the base architecture.
1136 (cpu cortex-a8): Add sec extension to the base architecture.
1137 (cpu marvell-pj4): Add mp and sec extensions to the base architecture.
1138 * config/arm/t-aprofile (MULTILIB_MATCHES): Map all armv7-a arch
1139 variants down to the base v7-a varaint.
1140 * config/arm/t-multilib (v7_a_arch_variants): New variable.
1141 * doc/invoke.texi (ARM Options): Add +mp and +sec to the list
1142 of permitted extensions for -march=armv7-a and for
1143 -mcpu=generic-armv7-a.
1145 2019-01-17 Uroš Bizjak <ubizjak@gmail.com>
1147 * config/alpha/alpha.c (alpha_gimplify_va_arg):
1148 Handle split indirect COMPLEX_TYPE arguments.
1150 2019-01-17 Eric Botcazou <ebotcazou@adacore.com>
1152 * gimple-ssa-isolate-paths.c (stmt_uses_name_in_undefined_way): Replace
1153 flag_non_call_exceptions with cfun->can_throw_non_call_exceptions.
1154 (stmt_uses_0_or_null_in_undefined_way): Likewise.
1155 * tree-ssa-alias.c (same_addr_size_stores_p): Likewise.
1157 2019-01-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1159 Backported from mainline
1160 2019-01-03 Aaron Sawdey <acsawdey@linux.ibm.com>
1161 * config/rs6000/rs6000-string.c (expand_block_move): Don't use
1162 unaligned vsx and avoid lxvd2x/stxvd2x.
1163 (gen_lvx_v4si_move): New function.
1165 2019-01-16 Martin Jambor <mjambor@suse.cz>
1167 Backported from mainline
1168 2018-12-10 Martin Jambor <mjambor@suse.cz>
1171 * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
1172 we check pointers against pointers.
1174 2019-01-10 Jan Hubicka <hubicka@ucw.cz>
1176 Backported from mainline
1177 2018-12-27 Jan Hubicka <hubicka@ucw.cz>
1179 * ipa-devirt.c (polymorphic_call_target_d): Add n_odr_types.
1180 (polymorphic_call_target_hasher::hash): Hash it.
1181 (polymorphic_call_target_hasher::equal): Compare it.
1182 (possible_polymorphic_call_targets): Set it.
1183 * tree.c (free_lang_data): Rebuild type inheritance graph even on
1186 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1188 Backported from mainline
1189 2019-01-02 Richard Biener <rguenther@suse.de>
1192 * ipa-icf.h (sem_item_optimizer::sort_congruence_split): Declare.
1193 * ipa-icf.c (sem_item_optimizer::sort_congruence_split): New
1195 (sem_item_optimizer::do_congruence_step_f): Sort the congruence
1196 set after UIDs before splitting them.
1198 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1200 Backported from mainline
1201 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1203 PR tree-optimization/85574
1204 * tree-ssa-uncprop.c (struct equiv_hash_elt): Remove unused
1206 (struct ssa_equip_hash_traits): Declare.
1207 (val_ssa_equiv): Use custom hash traits using operand_equal_p.
1209 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1211 * config/sparc/sparc.md (*tablejump_sp32): Merge into...
1212 (*tablejump_sp64): Likewise.
1213 (*tablejump<P:mode>): ...this.
1214 (*call_address_sp32): Merge into...
1215 (*call_address_sp64): Likewise.
1216 (*call_address<P:mode>): ...this.
1217 (*call_symbolic_sp32): Merge into...
1218 (*call_symbolic_sp64): Likewise.
1219 (*call_symbolic<P:mode>): ...this.
1220 (call_value): Remove constraint and add predicate.
1221 (*call_value_address_sp32): Merge into...
1222 (*call_value_address_sp64): Likewise.
1223 (*call_value_address<P:mode>): ...this.
1224 (*call_value_symbolic_sp32): Merge into...
1225 (*call_value_symbolic_sp64): Likewise.
1226 (*call_value_symbolic<P:mode>): ...this.
1227 (*sibcall_symbolic_sp32): Merge into...
1228 (*sibcall_symbolic_sp64): Likewise.
1229 (*sibcall_symbolic<P:mode>): ...this.
1230 (sibcall_value): Remove constraint and add predicate.
1231 (*sibcall_value_symbolic_sp32): Merge into...
1232 (*sibcall_value_symbolic_sp64): Likewise.
1233 (*sibcall_value_symbolic<P:mode>): ...this.
1234 (window_save): Minor tweak.
1235 (*branch_sp32): Merge into...
1236 (*branch_sp64): Likewise.
1237 (*branch<P:mode>): ...this.
1239 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1240 James Clarke <jrtc27@jrtc27.com>
1243 * config/sparc/sparc.c (sparc_legitimize_tls_address): Only use Pmode
1244 consistently in TLS address generation and adjust code to the renaming
1245 of patterns. Mark calls to __tls_get_addr as const.
1246 * config/sparc/sparc.md (tgd_hi22): Turn into...
1247 (tgd_hi22<P:mode>): ...this and use Pmode throughout.
1248 (tgd_lo10): Turn into...
1249 (tgd_lo10<P:mode>): ...this and use Pmode throughout.
1250 (tgd_add32): Merge into...
1251 (tgd_add64): Likewise.
1252 (tgd_add<P:mode>): ...this and use Pmode throughout.
1253 (tldm_hi22): Turn into...
1254 (tldm_hi22<P:mode>): ...this and use Pmode throughout.
1255 (tldm_lo10): Turn into...
1256 (tldm_lo10<P:mode>): ...this and use Pmode throughout.
1257 (tldm_add32): Merge into...
1258 (tldm_add64): Likewise.
1259 (tldm_add<P:mode>): ...this and use Pmode throughout.
1260 (tldm_call32): Merge into...
1261 (tldm_call64): Likewise.
1262 (tldm_call<P:mode>): ...this and use Pmode throughout.
1263 (tldo_hix22): Turn into...
1264 (tldo_hix22<P:mode>): ...this and use Pmode throughout.
1265 (tldo_lox10): Turn into...
1266 (tldo_lox10<P:mode>): ...this and use Pmode throughout.
1267 (tldo_add32): Merge into...
1268 (tldo_add64): Likewise.
1269 (tldo_add<P:mode>): ...this and use Pmode throughout.
1270 (tie_hi22): Turn into...
1271 (tie_hi22<P:mode>): ...this and use Pmode throughout.
1272 (tie_lo10): Turn into...
1273 (tie_lo10<P:mode>): ...this and use Pmode throughout.
1274 (tie_ld64): Use DImode throughout.
1275 (tie_add32): Merge into...
1276 (tie_add64): Likewise.
1277 (tie_add<P:mode>): ...this and use Pmode throughout.
1278 (tle_hix22_sp32): Merge into...
1279 (tle_hix22_sp64): Likewise.
1280 (tle_hix22<P:mode>): ...this and use Pmode throughout.
1281 (tle_lox22_sp32): Merge into...
1282 (tle_lox22_sp64): Likewise.
1283 (tle_lox22<P:mode>): ...this and use Pmode throughout.
1284 (*tldo_ldub_sp32): Merge into...
1285 (*tldo_ldub_sp64): Likewise.
1286 (*tldo_ldub<P:mode>): ...this and use Pmode throughout.
1287 (*tldo_ldub1_sp32): Merge into...
1288 (*tldo_ldub1_sp64): Likewise.
1289 (*tldo_ldub1<P:mode>): ...this and use Pmode throughout.
1290 (*tldo_ldub2_sp32): Merge into...
1291 (*tldo_ldub2_sp64): Likewise.
1292 (*tldo_ldub2<P:mode>): ...this and use Pmode throughout.
1293 (*tldo_ldsb1_sp32): Merge into...
1294 (*tldo_ldsb1_sp64): Likewise.
1295 (*tldo_ldsb1<P:mode>): ...this and use Pmode throughout.
1296 (*tldo_ldsb2_sp32): Merge into...
1297 (*tldo_ldsb2_sp64): Likewise.
1298 (*tldo_ldsb2<P:mode>): ...this and use Pmode throughout.
1299 (*tldo_ldub3_sp64): Use DImode throughout.
1300 (*tldo_ldsb3_sp64): Likewise.
1301 (*tldo_lduh_sp32): Merge into...
1302 (*tldo_lduh_sp64): Likewise.
1303 (*tldo_lduh<P:mode>): ...this and use Pmode throughout.
1304 (*tldo_lduh1_sp32): Merge into...
1305 (*tldo_lduh1_sp64): Likewise.
1306 (*tldo_lduh1<P:mode>): ...this and use Pmode throughout.
1307 (*tldo_ldsh1_sp32): Merge into...
1308 (*tldo_ldsh1_sp64): Likewise.
1309 (*tldo_ldsh1<P:mode>): ...this and use Pmode throughout.
1310 (*tldo_lduh2_sp64): Use DImode throughout.
1311 (*tldo_ldsh2_sp64): Likewise.
1312 (*tldo_lduw_sp32): Merge into...
1313 (*tldo_lduw_sp64): Likewise.
1314 (*tldo_lduw<P:mode>): ...this and use Pmode throughout.
1315 (*tldo_lduw1_sp64): Use DImode throughout.
1316 (*tldo_ldsw1_sp64): Likewise.
1317 (*tldo_ldx_sp64): Likewise.
1318 (*tldo_stb_sp32): Merge into...
1319 (*tldo_stb_sp64): Likewise.
1320 (*tldo_stb<P:mode>): ...this and use Pmode throughout.
1321 (*tldo_sth_sp32): Merge into...
1322 (*tldo_sth_sp64): Likewise.
1323 (*tldo_sth<P:mode>): ...this and use Pmode throughout.
1324 (*tldo_stw_sp32): Merge into...
1325 (*tldo_stw_sp64): Likewise.
1326 (*tldo_stw<P:mode>): ...this and use Pmode throughout.
1327 (*tldo_stx_sp64): Use DImode throughout.
1329 2019-01-09 Eric Botcazou <ebotcazou@adacore.com>
1331 * doc/invoke.texi (-Os): Add reference to -finline-functions.
1332 (-finline-functions): Remove reference to -O2, add references
1333 to -fprofile-use and -fauto-profile.
1335 2019-01-08 Jakub Jelinek <jakub@redhat.com>
1337 Backported from mainline
1338 2019-01-05 Jakub Jelinek <jakub@redhat.com>
1342 * expr.c (expand_assignment): For calls returning VLA structures
1343 if to_rtx is not a MEM, force it into a stack temporary.
1345 2019-01-04 Jakub Jelinek <jakub@redhat.com>
1348 * config/i386/i386.c (ix86_expand_divmod_libfunc): Use mode instead
1349 of GET_MODE (opN) as modes of the libcall arguments.
1351 2019-01-03 Jakub Jelinek <jakub@redhat.com>
1354 * dwarf2out.c (modified_type_die): If type is equal to sizetype,
1355 change it to qualified_type.
1357 2018-12-21 Jakub Jelinek <jakub@redhat.com>
1361 * omp-expand.c (extract_omp_for_update_vars): Regimplify the condition
1363 (expand_omp_for_generic): Don't clobber t temporary for ordered loops.
1366 * config/i386/sse.md (*avx512pf_gatherpf<mode>sf_mask,
1367 *avx512pf_gatherpf<mode>df_mask, *avx512pf_scatterpf<mode>sf_mask,
1368 *avx512pf_scatterpf<mode>df_mask): Use %X5 instead of %5 for
1370 (gatherq_mode): Remove mode iterator.
1371 (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2): Use X instead
1373 (*avx512f_gatherdi<mode>): Use X instead of <gatherq_mode>.
1374 (*avx512f_gatherdi<mode>_2, *avx512f_scattersi<mode>,
1375 *avx512f_scatterdi<mode>): Use %X5 for -masm=intel.
1377 PR rtl-optimization/88563
1378 * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Swap innermode
1379 and mode arguments to convert_modes. Likewise swap mode and word_mode
1380 arguments. Handle both arguments with VOIDmode before convert_modes
1381 of one of them. Formatting fixes.
1383 2018-12-19 Jakub Jelinek <jakub@redhat.com>
1386 * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable
1387 for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq.
1389 2018-12-14 Jakub Jelinek <jakub@redhat.com>
1393 * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
1394 * expr.c (categorize_ctor_elements_1): Likewise. Compute it like
1395 p_nz_elts, except don't multiply it by mult. Adjust recursive call.
1396 Fix up COMPLEX_CST handling.
1397 (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
1398 it and pass it through to categorize_ctor_elements_1.
1399 (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
1400 * gimplify.c (gimplify_init_constructor): Likewise. Don't force
1401 ctor into readonly data section if num_unique_nonzero_elements is
1402 smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
1405 2018-12-13 Jakub Jelinek <jakub@redhat.com>
1407 PR rtl-optimization/88470
1408 * cfgcleanup.c (outgoing_edges_match): If the function is
1409 shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
1410 edge to EXIT, return false.
1412 PR rtl-optimization/88416
1413 * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
1414 even if !AUTO_INC_DEC.
1416 2018-12-07 Jakub Jelinek <jakub@redhat.com>
1419 * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
1420 for functions with naked attribute.
1422 2018-12-06 Jakub Jelinek <jakub@redhat.com>
1425 * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
1426 call output_operand_lossage on VOIDmode CONST_INTs. After
1427 output_operand_lossage do return false.
1429 2018-12-05 Jakub Jelinek <jakub@redhat.com>
1431 PR tree-optimization/87360
1432 * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
1433 data dependencies, don't return false, just continue.
1435 2018-11-29 Jakub Jelinek <jakub@redhat.com>
1438 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
1439 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
1440 in unsigned_type_for instead of vector integral type where overflow
1443 2018-11-26 Jakub Jelinek <jakub@redhat.com>
1446 * dwarf2out.c (secname_for_decl): For functions with
1447 DECL_SECTION_NAME if in_cold_section_p, try to return
1448 current_function_section's name if it is a named section.
1450 2018-11-22 Jakub Jelinek <jakub@redhat.com>
1454 * config/i386/i386.c (ix86_option_override_internal): Default
1455 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
1457 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
1458 stack_protect_test, stack_protect_test_<mode>): Use empty condition
1459 instead of TARGET_SSP_TLS_GUARD.
1461 2018-11-20 Jakub Jelinek <jakub@redhat.com>
1463 PR tree-optimization/87895
1464 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
1465 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
1466 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
1467 redirects edges to EXIT to edges to incr_bb, iterate while EXIT
1468 has any preds and always use EDGE_PRED (, 0).
1470 2018-11-19 Jakub Jelinek <jakub@redhat.com>
1472 PR tree-optimization/88071
1473 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
1474 of false as last argument to gsi_remove.
1475 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
1476 false as last argument to gsi_replace.
1479 * omp-expand.c: Don't include debug.h.
1480 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT
1481 from innermost outer parallel, task, or target that has a
1482 child_fn set, or, if there is no such outer region, use
1483 current_function_decl. Do the DECL_CONTEXT adjustment regardless of
1484 whether a suitable BLOCK is found or not.
1485 (expand_parallel_call): Don't call adjust_context_and_scope here.
1486 (grid_expand_target_grid_body): Revert 2017-01-25 changes.
1487 (expand_omp_taskreg, expand_omp_target): Likewise. Call
1488 adjust_context_and_scope.
1489 * dwarf2out.c (dwarf2out_early_global_decl): For
1490 decl_function_context recurse instead of calling dwarf2out_decl.
1492 2018-11-16 Jakub Jelinek <jakub@redhat.com>
1494 PR rtl-optimization/87475
1495 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
1496 CROSSING_JUMP_P insns.
1497 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
1499 2018-11-15 Jakub Jelinek <jakub@redhat.com>
1501 PR rtl-optimization/88018
1502 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
1503 on the presence of fallthru edge, rather than if it is a USE or not.
1505 2018-11-13 Jakub Jelinek <jakub@redhat.com>
1507 PR tree-optimization/87898
1508 * omp-simd-clone.c (ipa_simd_modify_function_body): Remove debug stmts
1509 where the first argument was changed into a non-decl.
1511 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1513 Backport from mainline
1514 2019-01-02 Jan Hubicka <hubicka@ucw.cz>
1517 * varpool.c (varpool_node::ctor_useable_for_folding_p): Also return
1518 false at WPA time when body was removed.
1520 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1522 Backport from mainline
1523 2018-08-29 Jan Hubicka <jh@suse.cz>
1527 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1528 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1530 2019-01-04 Aaron Sawdey <acsawdey@linux.ibm.com>
1532 Backport from mainline
1533 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
1535 * config/rs6000/rs6000-string.c (expand_block_clear): Change how
1536 we determine if unaligned vsx is ok.
1538 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1540 Backport from mainline
1541 2019-01-03 Jan Hubicka <hubicka@ucw.cz>
1543 * ipa-utils.c (scale_ipa_profile_for_fn): Break out from ...
1544 (ipa_merge_profiles): ... here; do not ICE on thunks and aliases.
1546 2019-01-03 Martin Liska <mliska@suse.cz>
1548 Backport from mainline
1549 2019-01-01 Jan Hubicka <hubicka@ucw.cz>
1551 * coverage.c (get_coverage_counts): Use current_function_decl.
1552 * profile.c (read_thunk_profile): New function.
1553 (branch_prob): Add THUNK parameter.
1554 * tree-profile.c (tree_profiling): Handle thunks.
1555 * value-prof.c (init_node_map): Handle thunks.
1556 * value-prof.h (branch_prob): Upate prototype.
1557 (read_thunk_profile): Declare.
1559 2019-01-03 Iain Sandoe <iain@sandoe.co.uk>
1562 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1564 backport from mainline.
1565 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
1566 Iain Sandoe <iain@sandoe.co.uk>
1569 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
1570 unless it has been used.
1571 (first_reg_to_save): Remove dead code.
1573 2019-01-02 Segher Boessenkool <segher@kernel.crashing.org>
1576 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1579 * doc/extend.texi (Basic Asm): Update grammar.
1580 (Extended Asm): Update grammar.
1583 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1585 * doc/extend.texi (Using Assembly Language with C): Document asm inline.
1586 (Size of an asm): Fix typo. Document asm inline.
1587 * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
1588 * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
1589 (gimple_asm_set_volatile): Fix typo.
1590 (gimple_asm_inline_p): New.
1591 (gimple_asm_set_inline): New.
1592 * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
1594 * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
1595 gimple_asm_inline_p flag, too.
1596 * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
1598 * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
1599 a minimum size for an asm.
1600 * tree.h (ASM_INLINE_P): New.
1602 2018-12-28 Thomas Schwinge <thomas@codesourcery.com>
1603 Julian Brown <julian@codesourcery.com>
1605 * gimplify.c (gimplify_scan_omp_clauses): Fix known_eq typo/bug.
1607 2018-12-27 Martin Liska <mliska@suse.cz>
1609 Backport from mainline
1610 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
1613 * ipa-polymorphic-call.c
1614 (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Handle
1615 arguments of thunks correctly.
1616 (ipa_polymorphic_call_context::get_dynamic_context): Be ready for
1617 NULL instance pinter.
1618 * lto-cgraph.c (lto_output_node): Always stream thunk info.
1620 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1622 backport from mainline.
1623 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
1624 Iain Sandoe <iain@sandoe.co.uk>
1627 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
1628 unless it has been used.
1629 (first_reg_to_save): Remove dead code.
1631 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1633 Backport from mainline
1634 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1637 * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
1639 * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
1641 2018-12-23 Iain Sandoe <iain@sandoe.co.uk>
1643 Backport from mainline
1644 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1647 * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
1648 * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
1649 stack alignment in non-leaf functions.
1651 2018-12-21 Uros Bizjak <ubizjak@gmail.com>
1653 Backport from mainline
1654 2018-12-10 Uros Bizjak <ubizjak@gmail.com>
1657 * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
1658 check operand 1 with vector_operand predicate.
1659 (ix86_expand_sse_movcc): For vector modes, check op_true with
1660 vector_operand, not nonimmediate_operand.
1662 2018-12-19 Bill Schmidt <wschmidt@linux.ibm.com>
1664 Backport from mainline
1665 2018-12-18 Bill Schmidt <wschmidt@linux.ibm.com>
1667 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
1668 Describe when a typedef name can be used as the type specifier for
1669 a vector type, and when it cannot.
1671 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
1674 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
1677 * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load):
1678 Require TARGET_POWERPC64.
1680 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
1683 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
1685 PR rtl-optimization/88253
1686 * combine.c (combine_simplify_rtx): Test for side-effects before
1687 substituting by zero.
1689 2018-12-15 Segher Boessenkool <segher@kernel.crashing.org>
1692 2018-12-14 Segher Boessenkool <segher@kernel.crashing.org>
1694 PR rtl-optimization/88001
1695 * function.c (match_asm_constraints_1): Don't invalidly share RTL.
1697 2018-12-15 Jan Hubicka <jh@suse.cz>
1699 * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
1701 (cgraph_build_static_cdtor): Update.
1702 (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
1705 2018-12-15 Jan Hubicka <jh@suse.cz>
1707 Backport from mainline
1708 2018-11-02 Richard Biener <rguenther@suse.de>
1710 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
1713 2018-12-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1715 Backport from mainline
1716 2018-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1718 * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
1721 2018-12-14 Andre Vieira <andre.simoesdiasvieira@arm.com>
1723 Backport from mainline
1725 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
1726 (cortex-r7, cortex-r8): Update fpu and add new configuration.
1727 * doc/invoke.texi (armv7-r): Add two new vfp options.
1728 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
1729 support this option.
1731 2018-12-13 H.J. Lu <hongjiu.lu@intel.com>
1733 Backport from mainline
1734 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
1737 * config/i386/emmintrin.h (__v16qs): New to cope with option
1739 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
1740 (_mm_cmplt_epi8): Likewise.
1741 (_mm_cmpgt_epi8): Likewise.
1743 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
1745 Backport from mainline
1746 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
1748 * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
1749 s390_vec_double_s64 instead of s390_vcdgb.
1750 (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
1753 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
1755 Backport from mainline
1756 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
1758 * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
1759 ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
1760 ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
1762 2018-12-13 Eric Botcazou <ebotcazou@adacore.com>
1764 Backport from mainline
1765 2018-12-11 Eric Botcazou <ebotcazou@adacore.com>
1767 * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
1768 accounting for STACK_BOUNDARY 128.
1769 (STACK_DYNAMIC_OFFSET): Likewise.
1771 2018-12-12 Peter Bergner <bergner@linux.ibm.com>
1773 Backport from mainline
1774 2018-12-07 Peter Bergner <bergner@linux.ibm.com>
1777 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1778 -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
1779 Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
1780 * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
1781 require -mlong-double-128.
1783 2018-12-05 Richard Biener <rguenther@suse.de>
1785 Backport from mainline
1786 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1788 PR tree-optimization/87288
1789 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1790 into account when determining PEELING_FOR_NITERS.
1792 2018-12-05 Iain Sandoe <iain@sandoe.co.uk>
1794 Backport from mainline
1795 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
1797 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
1798 emit a diagnostic that it is not supported and reset the option.
1799 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
1800 supported and consume the option. (ASM_FINAL_SPEC): New.
1802 2018-12-04 Iain Sandoe <iain@sandoe.co.uk>
1804 Backport from mainline
1805 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1808 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
1809 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
1810 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
1812 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
1814 Backport from mainline
1815 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
1818 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1819 -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
1821 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
1823 Backport from mainline
1824 2018-10-25 Aaron Sawdey <acsawdey@linux.ibm.com>
1826 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
1827 a shorter sequence with fewer branches.
1828 (emit_final_str_compare_gpr): Ditto.
1830 Backport from mainline to allow the above code to go in:
1831 2018-06-14 Aaron Sawdey <acsawdey@linux.ibm.com>
1833 * config/rs6000/rs6000-string.c (do_and3, do_and3_mask,
1834 do_cmpb3, do_rotl3): New functions.
1836 2018-11-28 Richard Biener <rguenther@suse.de>
1838 PR tree-optimization/79351
1839 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
1840 empty CONSTRUCTORs ensure the store is at a constant position.
1842 2018-11-28 Richard Biener <rguenther@suse.de>
1844 PR tree-optimization/88229
1845 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
1846 is an INTEGER_CST before accessing it so.
1848 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
1850 Backport from mainline
1851 2018-11-20 Andreas Krebbel <krebbel@linux.ibm.com>
1853 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
1854 DImode parts of the target operand.
1856 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
1858 Backport from mainline
1859 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
1861 * doc/invoke.texi: Document z14/arch12 -march option.
1863 2018-11-23 Tamar Christina <tamar.christina@arm.com>
1865 Backported from mainline.
1866 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1869 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1870 (mov<mov>): ..this and enable unconditionally.
1872 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
1874 Backport from mainline
1875 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
1878 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
1879 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
1881 2018-11-20 Eric Botcazou <ebotcazou@adacore.com>
1883 PR rtl-optimization/85925
1884 * rtl.h (word_register_operation_p): New predicate.
1885 * combine.c (record_dead_and_set_regs_1): Only apply specific handling
1886 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
1887 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments.
1888 (num_sign_bit_copies1): Likewise.
1890 2018-11-19 Martin Liska <mliska@suse.cz>
1892 PR gcov-profile/88045
1893 * coverage.c (coverage_begin_function): One can't relly on the
1894 fact that function ends on the same line it starts or later.
1896 2018-11-18 Uros Bizjak <ubizjak@gmail.com>
1898 Backport from mainline
1899 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
1902 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1903 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1904 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1905 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1907 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
1909 * tree-switch-conversion.c (emit_case_decision_tree): Pass location
1910 of switch statement to emit_case_nodes.
1911 (do_jump_if_equal): Add LOC parameter and set it on the newly built
1912 GIMPLE comparison statement.
1913 (emit_cmp_and_jump_insns): Likewise.
1914 (emit_case_nodes): Add LOC parameter and pass it to do_jump_if_equal
1915 as well as recursive calls.
1917 2018-11-15 Nathan Sidwell <nathan@acm.org>
1921 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
1924 2018-11-14 Paul A. Clarke <pc@us.ibm.com>
1926 Backport from mainline
1927 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
1929 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1931 Backport from mainline
1933 * config/arc/arc.c (arc_return_address_register): Fix
1936 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1938 Backport from mainline
1939 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
1940 * config/arc/elf.h (PROFILE_HOOK): Define.
1941 * config/arc/linux.h (PROFILE_HOOK): Likewise.
1943 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1945 Backport from mainline
1946 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
1947 jump table data in the text section.
1949 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1951 Backport from mainline
1952 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
1953 to fit the eh_return pattern.
1954 * config/arc/arc.md (eh_return): Define.
1955 (VUNSPEC_ARC_EH_RETURN): Likewise.
1956 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
1958 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
1960 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1962 Backport from mainline
1963 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
1965 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
1967 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1969 Backport from mainline
1970 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1972 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
1974 Backport from mainline
1975 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
1978 * cfgexpand.c (pass_expand::execute): Move the call to
1979 finish_eh_generation in front of the call to expand_stack_alignment.
1981 2018-11-07 Max Filippov <jcmvbkbc@gmail.com>
1983 Backport from mainline
1984 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
1986 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1988 2018-11-05 Jakub Jelinek <jakub@redhat.com>
1990 PR tree-optimization/87859
1991 * gimple-ssa-store-merging.c (struct merged_store_group): Add
1992 first_nonmergeable_order member.
1993 (merged_store_group::merged_store_group): Initialize them.
1994 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
1995 stores with order >= first_nonmergeable_order.
1996 Set merged_store->first_nonmergeable_order if we've skipped any
1997 stores. Attempt to merge overlapping INTEGER_CST stores that
1998 we would otherwise skip.
2001 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2003 Backported from mainline
2004 2018-10-20 Jakub Jelinek <jakub@redhat.com>
2007 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
2009 2018-10-19 Jakub Jelinek <jakub@redhat.com>
2013 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
2014 depend closely nested inside of loop with ordered clause with
2017 2018-10-29 Richard Biener <rguenther@suse.de>
2019 Backport from mainline
2020 2018-09-26 Richard Biener <rguenther@suse.de>
2024 * tree-inline.c (expand_call_inline): When the location
2025 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
2026 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
2027 the inserted BLOCK to make inlined_function_outer_scope_p
2029 * dwarf2out.c (add_call_src_coords_attributes): Do not add
2030 coords for reserved locations.
2032 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2034 Backport from mainline
2035 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2040 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2041 (dwarf2out_switch_text_section): Generate a local label for the second
2042 function sub-section and apply it as the second FDE start label.
2043 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2044 second sub-section start.
2046 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2048 Backport from mainline
2049 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2052 (darwin_function_switched_text_sections): Delete.
2053 * gcc/config/darwin.h
2054 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2056 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2058 backport from mainline.
2059 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2062 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2063 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2064 (STACK_DYNAMIC_OFFSET): Likewise.
2066 2018-10-26 Bill Schmidt <wschmidt@linux.ibm.com>
2068 Backport from mainline
2069 2018-10-19 Bill Schmidt <wschmidt@linux.ibm.com>
2071 PR tree-optimization/87473
2072 * gimple-ssa-strength-reduction.c (record_phi_increments_1): For
2073 phi arguments identical to the base expression of the phi
2074 candidate, record a phi-adjust increment of zero minus the index
2075 expression of the hidden basis.
2076 (phi_incr_cost_1): For phi arguments identical to the base
2077 expression of the phi candidate, the difference to compare against
2078 the increment is zero minus the index expression of the hidden
2079 basis, and there is no potential savings from replacing the (phi)
2081 (ncd_with_phi): For phi arguments identical to the base expression
2082 of the phi candidate, the difference to compare against the
2083 increment is zero minus the index expression of the hidden basis.
2084 (all_phi_incrs_profitable_1): For phi arguments identical to the
2085 base expression of the phi candidate, the increment to be checked
2086 for profitability is zero minus the index expression of the hidden
2089 2018-10-25 Richard Biener <rguenther@suse.de>
2091 PR tree-optimization/87665
2092 PR tree-optimization/87745
2093 * tree-vectorizer.h (get_earlier_stmt): Remove.
2094 (get_later_stmt): Pick up UID from the original non-pattern stmt.
2096 2018-10-24 Richard Biener <rguenther@suse.de>
2098 PR tree-optimization/87665
2099 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2102 2018-10-23 Richard Biener <rguenther@suse.de>
2104 PR tree-optimization/87700
2105 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2107 2018-10-19 Richard Biener <rguenther@suse.de>
2110 Backport from mainline
2111 2018-07-12 Richard Biener <rguenther@suse.de>
2113 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2114 (vn_lookup_simplify_result): Remove recursion limit applied
2116 (vn_nary_build_or_lookup_1): Adjust.
2117 (try_to_simplify): Likewise.
2118 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2120 (gimple_resimplify2): Likewise.
2121 (gimple_resimplify3): Likewise.
2122 (gimple_resimplify4): Likewise.
2124 2018-10-19 Andreas Krebbel <krebbel@linux.ibm.com>
2126 Backport from mainline
2127 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
2129 * config/s390/s390.c (s390_expand_vec_init): Force vector element
2130 into reg if it isn't a general operand.
2132 2018-10-18 Richard Biener <rguenther@suse.de>
2136 2018-02-07 Richard Biener <rguenther@suse.de>
2138 PR tree-optimization/84204
2139 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2142 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
2145 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
2146 bail out if both sides do not have the same storage order.
2148 2018-10-17 Richard Biener <rguenther@suse.de>
2150 Backport from mainline
2151 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
2154 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
2155 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
2157 2018-10-16 Richard Biener <rguenther@suse.de>
2159 Backport from mainline
2160 2018-10-08 Richard Biener <rguenther@suse.de>
2162 PR tree-optimization/63155
2163 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
2164 (ssa_propagation_engine::ssa_propagate): Remove redundant
2165 bitmap bit clearing.
2167 2018-10-05 Richard Biener <rguenther@suse.de>
2169 PR tree-optimization/63155
2170 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
2171 vertical space in dumpfiles.
2172 * tree-ssa-propagate.h
2173 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
2174 * tree-ssa-propagate.c (cfg_blocks_back): New global.
2175 (ssa_edge_worklist_back): Likewise.
2176 (curr_order): Likewise.
2177 (cfg_blocks_get): Remove abstraction.
2178 (cfg_blocks_add): Likewise.
2179 (cfg_blocks_empty_p): Likewise.
2180 (add_ssa_edge): Add to current or next worklist based on
2182 (add_control_edge): Likewise.
2183 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
2185 (ssa_propagation_engine::ssa_propagate): ... here. Unify
2186 iteration from CFG and SSA edge worklist so we process
2187 everything in RPO order, prioritizing forward progress
2189 (ssa_prop_init): Allocate new worklists, do not dump
2191 (ssa_prop_fini): Free new worklists.
2193 2018-09-24 Richard Biener <rguenther@suse.de>
2195 PR tree-optimization/63155
2196 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
2197 the worklist when the edge of the respective argument isn't
2200 2018-10-16 Wilco Dijkstra <wdijkstr@arm.com>
2202 Backported from mainline
2204 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
2205 Use HOST_WIDE_INT_1U for shift.
2207 2018-10-16 Richard Biener <rguenther@suse.de>
2209 Backport from mainline
2210 2018-09-18 Richard Biener <rguenther@suse.de>
2213 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2214 (compute_samebase_partition_bases): Likewise.
2215 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2216 (gimple_can_coalesce_p): Simplify.
2218 2018-10-16 Richard Biener <rguenther@suse.de>
2220 Backport from mainline
2221 2018-10-15 Richard Biener <rguenther@suse.de>
2224 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
2225 (visit_loadstore): When a used restrict tag escaped verify that
2226 the points-to solution of "other" pointers do not include
2228 (compute_dependence_clique): If a used restrict tag escaped
2229 communicated that down to visit_loadstore.
2231 2018-10-01 Richard Biener <rguenther@suse.de>
2233 PR tree-optimization/87465
2234 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2235 causing branch miscounts.
2237 2018-10-11 Jakub Jelinek <jakub@redhat.com>
2239 Backported from mainline
2240 2018-10-10 Jakub Jelinek <jakub@redhat.com>
2243 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
2244 to special_args set.
2246 2018-09-26 Jakub Jelinek <jakub@redhat.com>
2249 * config/i386/i386.c: Include debug.h and dwarf2out.h.
2250 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2253 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2254 Andreas Krebbel <krebbel@linux.ibm.com>
2256 PR tree-optimization/86844
2257 * gimple-ssa-store-merging.c
2258 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2259 there are any overlapping stores in between them, make sure they are
2260 also coalesced or we give up completely.
2262 2018-10-09 Will Schmidt <will_schmidt@vnet.ibm.com>
2264 Backport from trunk.
2265 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2268 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2269 around folding of vec_sl to handle out of range shift values.
2271 2018-10-09 H.J. Lu <hongjiu.lu@intel.com>
2273 Backport from mainline
2274 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
2277 * config/i386/i386.c (construct_container): Use TImode for
2278 BLKmode values in 2 integer registers.
2280 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2282 Backport from mainline
2283 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
2286 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
2287 Defined with __builtin_ia32_vfmaddsubpd512_mask.
2289 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2291 Backport from mainline
2292 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
2295 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
2296 assembler for -mavx.
2297 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
2299 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
2301 Backport from mainline
2302 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
2304 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2305 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
2307 2018-10-03 Jonathan Wakely <jwakely@redhat.com>
2310 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2312 2018-10-03 Martin Liska <mliska@suse.cz>
2314 Backport from mainline
2315 2018-10-03 Martin Liska <mliska@suse.cz>
2317 PR gcov-profile/86109
2318 * coverage.c (coverage_begin_function): Do not
2319 mark lambdas as artificial.
2320 * tree-core.h (struct GTY): Remove tm_clone_flag
2321 and introduce new lambda_function.
2322 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2324 2018-10-02 H.J. Lu <hongjiu.lu@intel.com>
2326 Backport from mainline
2327 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
2330 * config/i386/i386.c (rest_of_insert_endbranch): Set
2331 endbr_queued_at_entrance to true and don't insert ENDBR if
2332 x86_function_profiler will be called.
2333 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
2335 * config/i386/i386.h (machine_function): Add
2336 endbr_queued_at_entrance.
2338 2018-10-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2340 Backport from mainline
2341 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2343 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
2344 if starting source register is not even.
2346 2018-09-29 Jakub Jelinek <jakub@redhat.com>
2349 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2350 __m512d type for __A argument rather than __m512.
2352 2018-09-27 Michael Meissner <meissner@linux.ibm.com>
2354 Backport from mainline
2355 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2358 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2359 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2362 2018-09-25 Martin Liska <mliska@suse.cz>
2364 Backport from mainline
2365 2018-09-24 Martin Liska <mliska@suse.cz>
2368 * asan.c: Make asan_handled_variables extern.
2370 * cfgexpand.c (expand_stack_vars): Make sure
2371 a representative is unpoison if another
2372 variable in the partition is handled by
2373 use-after-scope sanitization.
2375 2018-09-24 Andrew Pinski <apinski@marvell.com>
2377 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
2378 access prev before checking it for NULLness in the
2379 AARCH64_FUSE_CMP_BRANCH case.
2381 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2383 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2384 if the call takes a static chain.
2386 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
2388 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2389 (atomic_storehi): Likewise.
2390 (atomic_storesi): Likewise.
2391 (atomic_loaddi): Restore compare and swap exchange loop code.
2393 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2396 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2398 PR rtl-optimization/86882
2399 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2401 2018-09-18 Martin Liska <mliska@suse.cz>
2403 Backport from mainline
2404 2018-09-17 Martin Liska <mliska@suse.cz>
2406 PR gcov-profile/85871
2407 * gcov.c (output_intermediate_file): Fix out of bounds
2410 2018-09-16 Hans-Peter Nilsson <hp@bitrange.com>
2413 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2414 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2415 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2416 leaf_function_p, instead use has_hard_reg_initial_val.
2418 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
2421 * dojump.c (do_compare_and_jump): Canonicalize function pointers
2422 when one operand is a function pointer. Use POINTER_TYPE_P and
2423 FUNC_OR_METHOD_TYPE_P.
2424 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2425 * fold-const.c (build_range_check): Likewise.
2426 * match.pd (simple_comparison): Likewise.
2428 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2431 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2434 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2437 2018-09-14 Carl Love <cel@us.ibm.com>
2440 2018-09-14 Carl Love <cel@us.ibm.com>
2442 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2443 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2445 2018-09-12 Segher Boessenkool <segher@kernel.crashing.org>
2448 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
2451 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
2454 2018-09-12 Segher Boessenkool <segher@kernel.crashing.org>
2457 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2459 PR rtl-optimization/86771
2460 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2461 of two SETs into those two SETs, one to be placed at i2, if that SETs
2462 destination is modified between i2 and i3.
2464 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2466 Backport from mainline
2467 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2469 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2471 ("trunc<BFP:mode><DFP_ALL:mode>2")
2472 ("trunc<DFP_ALL:mode><BFP:mode>2")
2473 ("extend<BFP:mode><DFP_ALL:mode>2")
2474 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2475 according to the target operand type.
2477 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2480 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2481 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
2483 2018-09-05 Nathan Sidwell <nathan@acm.org>
2486 * stor-layout.c (place_field): Scan forwards to check last
2487 bitfield when ms_bitfield_placement is in effect.
2489 2018-09-05 Richard Biener <rguenther@suse.de>
2492 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2495 2018-09-05 Jakub Jelinek <jakub@redhat.com>
2497 Backported from mainline
2498 2018-09-04 Jakub Jelinek <jakub@redhat.com>
2501 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2502 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2503 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2504 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2505 and OPTION_MASK_ISA_XSAVEC_UNSET.
2507 2018-08-31 Jakub Jelinek <jakub@redhat.com>
2510 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
2511 gen_int_mode. Formatting fixes.
2513 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2515 PR tree-optimization/86835
2516 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2517 new_stmt after def_gsi, make sure to insert new_square_stmt after
2518 that stmt, not 2 stmts before it.
2520 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2522 Backport from mainline
2523 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2525 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2526 XOR operations in NAND case.
2528 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2530 * doc/invoke.texi (Option Summary): Add -Waligned-new.
2532 2018-09-03 Tom de Vries <tdevries@suse.de>
2534 backport from trunk:
2535 2018-06-21 Tom de Vries <tdevries@suse.de>
2537 PR tree-optimization/85859
2538 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2539 test with comment from bb_no_side_effects_p.
2541 2018-09-01 Michael Matz <matz@suse.de>
2543 Backport from mainline
2544 PR tree-optimization/87074
2545 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
2546 PHIs for outer-loop uses.
2548 2018-08-31 Richard Biener <rguenther@suse.de>
2550 Backport from mainline
2551 2018-08-27 Richard Biener <rguenther@suse.de>
2553 PR tree-optimization/86927
2554 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2555 use const cond reduction code.
2557 2018-08-23 Richard Biener <rguenther@suse.de>
2560 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2563 2018-08-22 Richard Biener <rguenther@suse.de>
2565 PR tree-optimization/86945
2566 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2568 2018-08-17 Richard Biener <rguenther@suse.de>
2571 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2572 across a va-arg-pack using call adjust its return value accordingly.
2574 2018-08-02 Richard Biener <rguenther@suse.de>
2576 PR tree-optimization/86816
2577 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2578 which checks for value availability before querying it.
2579 (gvn_uses_equal): Use it.
2580 (same_succ_hash): Likewise.
2581 (gimple_equal_p): Likewise.
2583 2018-07-17 Richard Biener <rguenther@suse.de>
2586 * dwarf2out.c (init_sections_and_labels): Always generate
2587 a debug_line_str_section for early LTO debug.
2588 (dwarf2out_finish): Reset debug_line_str_hash output early.
2589 Bump counter for extra dwarf5 .debug_loc labels to not conflict
2591 (dwarf2out_early_finish): Output debug_line_str.
2593 2018-08-29 Luis Machado <luis.machado@linaro.org>
2595 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static global.
2596 (qdf24xx_tunings): Set vector cost structure to qdf24xx_vector_cost.
2598 2018-08-29 Luis Machado <luis.machado@linaro.org>
2600 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2601 <register_sextend>: Set to 3.
2603 2018-08-27 Martin Sebor <msebor@redhat.com>
2605 PR tree-optimization/86914
2606 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2608 2018-08-28 Jakub Jelinek <jakub@redhat.com>
2611 * calls.c (maybe_warn_nonstring_arg): Punt early if
2612 warn_stringop_overflow is zero. Don't call get_range_strlen
2613 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2614 Only use lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
2615 increment lenrng[0].
2617 2018-08-27 Jakub Jelinek <jakub@redhat.com>
2619 PR rtl-optimization/87065
2620 * combine.c (simplify_if_then_else): Formatting fix.
2621 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
2623 (known_cond): Don't return const_true_rtx for vector modes. Use
2624 CONST0_RTX instead of const0_rtx. Formatting fixes.
2626 2018-08-25 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2628 Backport from mainline
2630 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2631 with all enabled __intN types.
2633 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2635 2018-08-21 H.J. Lu <hongjiu.lu@intel.com>
2637 Backport from mainline
2638 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2641 * config/i386/i386.md (eh_return): Always update EH return
2642 address in word_mode.
2644 2018-08-17 John David Anglin <danglin@gcc.gnu.org>
2646 Backport from mainline
2647 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2649 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2650 Update comment for atomic instructions.
2651 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2652 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2654 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2655 load for __ATOMIC_SEQ_CST model.
2656 (atomic_loaddi_1): Remove float register target.
2657 (atomic_storedi): Handle CONST_INT values.
2658 (atomic_storedi_1): Remove float register source. Add special case
2660 (memory_barrier): New expander and insn.
2662 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2664 Backport from mainline
2665 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2668 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2671 2018-08-16 Martin Liska <mliska@suse.cz>
2673 Backport from mainline
2674 2018-08-02 Martin Liska <mliska@suse.cz>
2676 PR gcov-profile/86817
2677 * gcov.c (process_all_functions): New function.
2679 (process_file): Move functions processing to
2680 process_all_functions.
2682 2018-08-14 Siddhesh Poyarekar <siddhesh@sourceware.org>
2684 Backport from mainline
2685 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2687 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2689 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2691 2018-08-13 Liu Hao <lh_mouse@126.com>
2693 Backport from mainline
2694 2018-08-13 Liu Hao <lh_mouse@126.com>
2696 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2697 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2698 and clear it thereafter, as it only works for DBCS.
2700 2018-08-13 Liu Hao <lh_mouse@126.com>
2702 Backport from mainline
2703 2018-08-13 Liu Hao <lh_mouse@126.com>
2705 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2706 handle returned by _get_osf_handle().
2708 2018-08-14 Richard Sandiford <richard.sandiford@arm.com>
2710 Backport from mainline
2711 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2713 PR tree-optimization/86871
2714 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2715 instead of gimple_assign_lhs.
2717 2018-08-07 H.J. Lu <hongjiu.lu@intel.com>
2719 Backport from mainline
2720 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
2723 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
2724 cfun->machine->max_used_stack_alignment if needed.
2726 2018-08-02 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2728 Backport from mainline
2729 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2732 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
2733 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
2734 requested variable alignment.
2735 (expand_one_ssa_partition): Likewise.
2736 (expand_one_var): Likewise.
2738 2018-08-02 Andreas Schwab <schwab@linux-m68k.org>
2740 Backport from mainline
2741 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
2744 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
2745 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
2747 2018-08-01 Richard Biener <rguenther@suse.de>
2750 * graphite.h: Include isl/id.h and isl/space.h to allow build
2753 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
2756 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2757 mode for COFF targets.
2758 * defaults.h (TARGET_COFF): Define.
2759 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
2760 TARGET_COFF): Define.
2761 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
2762 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
2763 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
2765 2018-07-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
2767 Backport from mainline
2768 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
2771 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
2773 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2775 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2776 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2777 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2779 2018-07-26 Uros Bizjak <ubizjak@gmail.com>
2781 Backport from mainline
2782 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
2785 * expmed.c (emit_store_flag): Do not emit setcc followed by a
2786 conditional move when trapping comparison was split to a
2787 non-trapping one (and vice versa).
2789 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2791 Backported from mainline
2792 2018-07-24 Jakub Jelinek <jakub@redhat.com>
2795 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
2796 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
2797 and abs_d == d, do the power of two handling if profitable.
2799 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2802 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
2803 to _looptemp_ clauses, other than the first two.
2806 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
2807 and cond temporaries don't have reference type if iterator has
2808 pointer type. For init use &for_pre_body instead of pre_p if
2809 for_pre_body is non-empty.
2811 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2814 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2815 declare target to variables if they have always,{to,from,tofrom} map
2818 * BASE-VER: Set to 8.2.1.
2820 2018-07-26 Release Manager
2822 * GCC 8.2.0 released.
2824 2018-07-18 Martin Sebor <msebor@redhat.com>
2827 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
2828 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
2829 Handle integer subtraction.
2830 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
2831 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
2832 * doc/invoke.texi (-Wstringop-truncation): Update.
2834 gcc/testsuite/ChangeLog:
2836 2018-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2838 Backport from mainline
2839 2017-03-24 Claudiu Zissulescu <claziss@synopsys.com>
2841 * config/arc/arc-protos.h (arc_pad_return): Remove.
2842 * config/arc/arc.c (machine_function): Remove force_short_suffix
2844 (arc_print_operand): Adjust printing of '&'.
2845 (arc_verify_short): Remove conditional printing of short suffix.
2846 (arc_final_prescan_insn): Remove reference to size_reason.
2847 (pad_return): New function.
2848 (arc_reorg): Call pad_return.
2849 (arc_pad_return): Remove.
2850 (arc_init_machine_status): Remove reference to force_short_suffix.
2851 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
2852 (attr length): When attribute iscompact is true force to 2
2853 regardless; in the case of maybe check if we want to force the
2854 instruction to have 4 bytes length.
2855 (nopv): Change it to generate 4 byte long nop as well.
2856 (blockage): New pattern.
2857 (simple_return): Remove call to arc_pad_return.
2858 (p_return_i): Likewise.
2860 2018-07-19 Richard Biener <rguenther@suse.de>
2862 Backport from mainline
2863 2018-07-13 Richard Biener <rguenther@suse.de>
2866 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
2867 instead of get_context_die.
2869 2018-07-11 Richard Biener <rguenther@suse.de>
2872 * dwarf2out.c (init_sections_and_labels): Use
2873 output_asm_line_debug_info consistently.
2874 (dwarf2out_early_finish): Likewise.
2875 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
2878 2018-07-16 Richard Biener <rguenther@suse.de>
2880 Backport from mainline
2881 2018-07-13 Richard Biener <rguenther@suse.de>
2884 * match.pd (addr1 - addr2): Allow either of the operand to
2887 2018-06-15 Richard Biener <rguenther@suse.de>
2890 * tree-cfg.c (move_stmt_op): unshare invariant addresses
2891 before adjusting their block.
2893 2018-06-06 Richard Biener <rguenther@suse.de>
2895 PR tree-optimization/85935
2896 * graphite-scop-detection.c (find_params_in_bb): Analyze
2897 condition operands with respect to the correct loop. Assert
2898 the analysis doesn't fail.
2900 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
2902 PR tree-optimization/86514
2903 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
2904 conversion to a boolean type from a type with greater precision.
2906 2018-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
2908 Backport from mainline
2909 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
2910 Steve Munroe <munroesj52@gmail.com>
2912 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
2913 (_mm_andnot_si128): Likewise.
2914 (_mm_or_si128): Likewise.
2915 (_mm_xor_si128): Likewise.
2917 2018-07-14 Martin Sebor <msebor@redhat.com>
2919 PR tree-optimization/86274
2920 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2922 (format_floating): Correct handling of infinities and NaNs.
2924 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
2926 Backport from mainline
2927 2018-07-13 H.J. Lu <hongjiu.lu@intel.com>
2928 Sunil K Pandey <sunil.k.pandey@intel.com>
2931 * config/i386/i386.c (m_CORE_AVX512): New.
2932 (m_CORE_AVX2): Likewise.
2933 (m_CORE_ALL): Add m_CORE_AVX2.
2934 * config/i386/x86-tune.def: Replace m_HASWELL with m_CORE_AVX2.
2935 Replace m_SKYLAKE_AVX512 with m_CORE_AVX512 on avx256_optimal
2936 and remove the rest of m_SKYLAKE_AVX512.
2938 2018-07-13 Marek Polacek <polacek@redhat.com>
2941 2018-07-03 Marek Polacek <polacek@redhat.com>
2944 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2946 2018-07-12 Richard Biener <rguenther@suse.de>
2949 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
2950 Remove -mieee-fp handling.
2952 2018-07-12 Jakub Jelinek <jakub@redhat.com>
2954 PR tree-optimization/86492
2955 * gimple-ssa-store-merging.c
2956 (imm_store_chain_info::coalesce_immediate_stores): Call
2957 check_no_overlap even for the merge_overlapping case.
2959 2018-07-11 Jakub Jelinek <jakub@redhat.com>
2961 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
2962 Use __mmask64 type instead of __mmask8 for __M argument.
2963 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
2964 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
2966 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
2967 __mmask16 for __M argument.
2968 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
2969 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
2970 to __mmask16 instead of __mmask8.
2971 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
2972 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
2973 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
2974 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
2975 instead of __mmask16 for __U argument.
2976 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
2977 __mmask16 instead of __mmask8 for __U argument.
2978 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
2980 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
2982 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
2984 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
2986 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
2988 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
2990 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
2992 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
2994 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
2996 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
2998 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
2999 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
3000 return type as well as __M argument type and all casts from __mmask8
3002 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
3003 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
3004 return type as well as __M argument type and all casts from __mmask8
3006 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
3007 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
3008 return type as well as __M argument type and all casts from __mmask8
3010 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
3011 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
3012 return type as well as __M argument type and all casts from __mmask8
3014 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
3015 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
3018 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
3020 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
3021 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
3024 2018-07-05 Martin Sebor <msebor@redhat.com>
3027 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
3028 than its domain to compute its the upper bound of a char array.
3030 2018-07-03 Carl Love <cel@us.ibm.com>
3033 2018-07-03 Carl Love <cel@us.ibm.com>
3035 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
3036 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
3037 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
3038 VSX_BUILTIN_DOUBLEL_V4SF.
3040 2018-07-04 Luis Machado <luis.machado@linaro.org>
3042 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3044 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3046 2018-07-02 Segher Boessenkool <segher@kernel.crashing.org>
3049 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3052 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3053 ieee128_float_type_node to long_double_type_node unless
3054 TARGET_LONG_DOUBLE_128 is set.
3056 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3058 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3059 -fno-omit-frame-pointer when not optimizing.
3061 2018-06-30 Alexandre Oliva <aoliva@redhat.com>
3064 * dwarf2out.c (loc_list_has_views): Adjust comments.
3065 (dw_loc_list): Split single cross-partition range with
3068 2018-06-29 Kelvin Nilsen <kelvin@gcc.gnu.org>
3070 Backport from mainline
3071 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3073 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3074 behavior of vec_pack (vector double, vector double) to match
3075 behavior of vec_float2 (vector double, vector double).
3077 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3079 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3081 2018-06-28 Kelvin Nilsen <kelvin@gcc.gnu.org>
3083 Backport from mainline
3084 2017-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3086 * doc/sourcebuild.texi (Endianness): New subsubsection.
3088 2018-06-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
3090 Backport from mainline
3091 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
3093 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3094 behavior of vec_packsu (vector unsigned long long, vector unsigned
3095 long long) to match behavior of vec_packs with same signature.
3097 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3100 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3102 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3103 unaligned vsx for 16B memset.
3105 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3108 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3111 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3114 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3117 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
3119 PR rtl-optimization/85645
3120 * regrename.c (build_def_use): Also kill the chains that include the
3121 destination of a REG_CFA_REGISTER note.
3123 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3126 2018-05-08 Segher Boessenkool <segher@kernel.crashing.org>
3128 PR rtl-optimization/85645
3129 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
3130 insn that has a REG_CFA_REGISTER note.
3132 2018-06-26 Robin Dapp <rdapp@linux.ibm.com>
3134 * config/s390/s390.h (enum processor_flags): Do not use
3135 default tune parameter when -march was specified.
3137 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3140 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3141 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3144 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3146 Back port from trunk
3147 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3149 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3150 added IEEE/IBM long double multilib support on PowerPC little
3151 endian Linux systems.
3152 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3153 (MULTILIB_DEFAULTS): Likewise.
3154 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3156 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3157 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3158 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3160 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
3163 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
3165 * config/rs6000/rs6000.md (abs<mode>2): Handle IFmode.
3167 2018-05-02 Jan Hubicka <jh@suse.cz>
3169 Backport from mainline
3170 2018-04-30 Jan Hubicka <jh@suse.cz>
3172 * lto-wrapper.c (ltrans_priorities): New static var.
3173 (cmp_priority): New.
3174 (run_gcc): Read priorities and if doing parallel build order
3175 the Makefile by them.
3177 2018-06-25 Segher Boessenkool <segher@kernel.crashing.org>
3180 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3183 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3184 ieee128 argument takes up only one (vector) register, not two (floating
3187 2018-06-23 Richard Sandiford <richard.sandiford@linaro.org>
3189 PR tree-optimization/85989
3190 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
3192 (backprop::intersect_uses): Check it when deciding whether this
3193 is a backedge reference.
3194 (backprop::process_block): Add each phi to m_visited_phis
3195 after visiting it, then clear it at the end.
3197 2018-06-22 David Edelsohn <dje.gcc@gmail.com>
3199 Backport from mainline
3201 2018-06-19 Tony Reix <tony.reix@atos.com>
3202 Damien Bergamini <damien.bergamini@atos.com>
3203 David Edelsohn <dje.gcc@gmail.com>
3205 * collect2.c (static_obj): New variable.
3206 (static_libs): New variable.
3207 (is_in_list): Uncomment declaration.
3208 (main): Track AIX libraries linked statically.
3209 (is_in_list): Uncomment definition.
3210 (scan_prog_file): Don't add AIX shared libraries initializer
3211 to constructor list if linking statically.
3213 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
3215 Back port from trunk
3216 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3218 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3219 conversion insn that shows up when pr85657-3.c is compiled using
3220 IEEE 128-bit long double.
3222 Back port from trunk
3223 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3225 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3226 multiply and divide external functions from being created more
3229 Back port from trunk
3230 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3232 * config/rs6000/rs6000.md (neg<mode>2_internal): Use the correct
3233 mode to check whether the mode is IBM extended.
3235 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3238 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3239 floating point modes, so that IFmode is numerically greater than
3240 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3241 to declare the ordering. This prevents IFmode from being
3242 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3243 machine. Include rs6000-modes.h to share the fractional values
3244 between genmodes* and the rest of the compiler.
3248 * config/rs6000/rs6000-modes.h: New file.
3249 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3250 meaning of rs6000_long_double_size so that 126..128 selects an
3251 appropriate 128-bit floating point type.
3252 (rs6000_option_override_internal): Likewise.
3253 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3254 (TARGET_LONG_DOUBLE_128): Change the meaning of
3255 rs6000_long_double_size so that 126..128 selects an appropriate
3256 128-bit floating point type.
3257 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3258 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3259 source and destination to match the standard usage.
3260 (truncifkf2): Likewise.
3261 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3262 ISA 2.07 to use an explicit clobber, instead of passing in a
3264 (copysign<mode>3_soft): Likewise.
3266 Back port from trunk
3267 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3269 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3272 Back port from trunk
3273 2018-06-04 Michael Meissner <meissner@linux.ibm.com>
3275 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
3276 track if we pass or return IEEE 128-bit floating point.
3277 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
3278 C++ mangling that is compatible with GCC 8.1.
3279 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
3280 (init_cumulative_args): Note if we pass or return IEEE 128-bit
3281 floating point types.
3282 (rs6000_function_arg_advance_1): Likewise.
3283 (rs6000_mangle_type): Optionally generate mangled names that match
3284 what GCC 8.1 generated for IEEE 128-bit floating point types.
3285 (rs6000_globalize_decl_name): If we have an external function that
3286 passes or returns IEEE 128-bit types, generate a weak reference
3287 from the mangled name used in GCC 8.1 to the current mangled
3289 (rs6000_init_builtins): Make __ibm128 use the long double type if
3290 long double is IBM extended double. Make __float128 use the long
3291 double type if long double is IEEE 128-bit.
3294 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
3295 macro for __ibm128 built-in functions.
3296 (PACK_IF): Add __ibm128 pack/unpack functions.
3297 (UNPACK_IF): Likewise.
3298 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
3299 enable long double built-in functions if long double is IEEE
3300 128-bit floating point.
3301 (rs6000_invalid_builtin): Update long double built-in function
3303 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
3304 functions, adjust the built-in function to use the long double
3305 built-in function if __ibm128 and long double are the same type.
3306 * doc/extend.texi (PowerPC builtins): Update documention for
3307 __builtin_{,un}pack_longdouble. Add documentation for
3308 __builtin_{,un}pack_ibm128.
3310 Back port from trunk
3311 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
3313 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
3314 the 128-bit floating point types. Fix function comment.
3316 2018-06-22 Michael Meissner <meissner@linux.ibm.com>
3318 Back port from trunk
3319 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3322 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3323 define __ibm128 as long double.
3324 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3325 as a distinct type with IEEE 128-bit floating point is supported.
3326 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3327 128-bit types to use the correct functions.
3328 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3329 convert between 128-bit floating point types that have different
3330 modes but the same representation, instead of using gen_lowpart to
3332 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3334 (IFKF_reg): New attributes to give the register constraints for
3336 (extend<mode>tf2_internal): New insns to mark an explicit
3337 conversion between 128-bit floating point types that have a
3338 different mode but share the same representation.
3340 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3343 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3344 -x assembler-with-cpp.
3346 2018-06-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
3348 Backported from mainline
3349 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3351 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3352 multilibs for *-*-rtems*.
3353 * config/riscv/t-rtems: New file.
3355 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3357 Backported from mainline
3358 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3360 PR rtl-optimization/86108
3361 * bb-reorder.c (create_forwarder_block): Renamed to ...
3362 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3363 jump from new landing pad to the second part.
3364 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3367 2018-06-20 Jakub Jelinek <jakub@redhat.com>
3370 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
3373 PR tree-optimization/86231
3374 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
3375 anti-range don't overwrite *vr0min before using it to compute *vr0max.
3377 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3379 Backport from mainline
3380 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
3382 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
3384 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
3386 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3388 * gimplify.c (gimplify_init_constructor): Really never clear for an
3389 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3391 2018-06-18 Martin Sebor <msebor@redhat.com>
3394 * calls.c (alloc_max_size): Correct a logic error/typo.
3395 Treat excessive arguments as infinite. Warn for invalid arguments.
3396 * doc/invoke.texi (-Walloc-size-larger-than): Update.
3398 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3401 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3402 documented as preserved for backward compatibility only.
3403 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3405 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3408 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3409 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3410 Only call store_expr for halves if the mode is the same.
3413 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3414 Fix up comment formatting.
3416 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3419 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3420 unsigned_type_for returns NULL.
3423 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3424 subregs of multi-word pseudos unless the float mode has word size.
3426 2018-06-14 Richard Biener <rguenther@suse.de>
3429 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3430 duplicate and harmful type_hash_canon.
3431 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3433 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3435 Backport from mainline
3436 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3438 * config/rtems.h (STDINT_LONG32): Define.
3440 2018-05-05 Roland McGrath <mcgrathr@google.com>
3443 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3444 any section for which we don't know a specific type it should have,
3445 regardless of name. Previously this was done only for the exact
3446 names ".init_array", ".fini_array", and ".preinit_array".
3447 (default_elf_asm_named_section): Add comment about
3448 relationship with default_section_type_flags and SECTION_NOTYPE.
3449 (get_section): Don't consider it a type conflict if one side has
3450 SECTION_NOTYPE and the other doesn't, as long as neither has the
3451 SECTION_BSS et al used in the default_section_type_flags logic.
3453 2018-06-13 Martin Sebor <msebor@redhat.com>
3455 PR tree-optimization/86114
3456 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3458 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3460 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3463 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3464 offsets for register save directives. Emit a second batch of save
3465 directives, if need be, when the function accesses prior frames.
3467 2018-06-11 Jason Merrill <jason@redhat.com>
3469 PR c++/80485 - inline function non-zero address.
3470 * symtab.c (nonzero_address): Check DECL_COMDAT.
3472 2018-06-11 Peter Bergner <bergner@vnet.ibm.com>
3474 Backport from mainline
3475 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3478 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3481 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3484 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3487 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3488 on the correct operand.
3489 (*movdi_internal64): Ditto.
3491 2018-06-08 Carl Love <cel@us.ibm.com>
3493 Backport from mainline
3494 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
3495 vextract_fp_from_shortl): Add BE support.
3497 2018-06-08 Carl Love <cel@us.ibm.com>
3499 Backport from mainline
3500 * gcc/config/rs6000/vsx.md (first_match_index_<mode>):
3501 Calculate index using natureal element order.
3502 (first_match_or_eos_index_<mode>):
3503 Calculate index using natural element order.
3504 (first_match_index_<mode>):
3505 Calculate index using natural element order.
3506 (first_match_or_eos_index_<mode>):
3507 Calculate index using natural order.
3508 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
3509 for BE and LE modes.
3510 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
3511 P9V_BUILTIN_VCLZLSBB_V16QI.
3512 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
3515 2018-06-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3517 Backport from mainline
3518 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3521 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
3522 qualifier_void_pointer and qualifier_const_void_pointer.
3523 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
3524 (arm_init_builtins): Handle the above.
3525 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
3526 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
3527 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
3530 2018-06-07 Uros Bizjak <ubizjak@gmail.com>
3533 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3534 Generate SImode target register for null target.
3535 <case IX86_BUILTIN_XGETBV>: Ditto.
3536 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3537 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3539 2018-06-07 Peter Bergner <bergner@vnet.ibm.com>
3541 Backport from mainline
3542 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
3545 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
3546 Don't handle -mcpu=power8 if -mpower9-vector is also used.
3548 2018-06-06 Richard Biener <rguenther@suse.de>
3550 Backport from mainline
3551 2018-06-04 Richard Biener <rguenther@suse.de>
3553 PR tree-optimization/86038
3554 * tracer.c (find_best_successor): Check probability for
3555 being initialized, bail out if not.
3557 2018-05-30 Richard Biener <rguenther@suse.de>
3559 PR tree-optimization/85964
3560 * tracer.c (better_p): Drop initialized count check, we only
3561 call the function with initialized counts now.
3562 (find_best_successor): Do find a best edge if one
3563 has uninitialized count.
3564 (find_best_predecessor): Likewise. Do BB frequency check only
3565 if count is initialized.
3567 2018-05-28 Richard Biener <rguenther@suse.de>
3569 PR tree-optimization/85934
3570 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3571 vector boolean check before scalar optimization.
3573 2018-05-22 Richard Biener <rguenther@suse.de>
3575 PR tree-optimization/85863
3576 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3577 comparisons when vectype is specified.
3578 (vectorizable_condition): Do not specify vectype for
3579 vect_is_simple_cond when SLP vectorizing.
3581 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3583 Backport from mainline
3584 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3586 * config/s390/s390-builtin-types.def: Add void function type.
3587 * config/s390/s390-builtins.def: Use the function type for the
3590 2018-06-04 Martin Sebor <msebor@redhat.com>
3593 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3594 or ajust the presumed bound on an operation to avoid unnecessary
3597 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
3600 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
3601 of bits to ignore when comparing architectures.
3603 2018-06-04 Jakub Jelinek <jakub@redhat.com>
3606 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
3608 2018-06-01 Bill Schmidt <wschmidt@linux.ibm.com>
3610 PR tree-optimization/85712
3611 Backport from mainline:
3612 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3614 PR tree-optimization/85712
3615 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3617 (alloc_cand_and_find_basis): Initialize first_interp field.
3618 (slsr_process_mul): Modify first_interp field.
3619 (slsr_process_add): Likewise.
3620 (slsr_process_cast): Modify first_interp field for each new
3622 (slsr_process_copy): Likewise.
3623 (dump_candidate): Dump first_interp field.
3624 (replace_mult_candidate): Process all interpretations, not just
3626 (replace_rhs_if_not_dup): Likewise.
3627 (replace_one_candidate): Likewise.
3629 Backport from mainline:
3630 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3632 PR tree-optimization/85712
3633 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3634 this candidate has already been replaced in-situ by a copy.
3636 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
3639 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3640 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3642 (sse4_1_round<mode>2): Use nonimmediate_operand
3643 for operand 1 predicate.
3645 2018-05-31 Jakub Jelinek <jakub@redhat.com>
3648 * bb-reorder.c (pass_partition_blocks::gate): Return false for
3649 functions with naked attribute.
3651 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
3654 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
3655 and movx for Haswell.
3657 2018-05-30 Jakub Jelinek <jakub@redhat.com>
3659 Backported from mainline
3660 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3663 * config/i386/i386.md: Disallow non-commutative arithmetics in
3664 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3665 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3666 in the peephole2 before it.
3668 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3671 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
3672 after cmpelim optimization.
3674 2018-05-28 H.J. Lu <hongjiu.lu@intel.com>
3676 Backport from mainline
3677 2018-05-26 H.J. Lu <hongjiu.lu@intel.com>
3681 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3683 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3687 * varasm.c (assemble_alias): Check ifunc_resolver only on
3690 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3693 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3695 (cgraph_node::create_alias): Likewise.
3696 (cgraph_node::get_availability): Check ifunc_resolver instead
3697 of looking up ifunc attribute.
3698 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3699 * varasm.c (do_assemble_alias): Likewise.
3700 (assemble_alias): Likewise.
3701 (default_binds_local_p_3): Likewise.
3702 * cgraph.h (cgraph_node): Add ifunc_resolver.
3703 (cgraph_node::only_called_directly_or_aliased_p): Return false
3705 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3707 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3708 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3709 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3710 instead of looking up ifunc attribute.
3712 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3714 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3716 2018-05-25 Richard Biener <rguenther@suse.de>
3719 * tree-dump.c (dequeue_and_dump): Remove access to removed
3720 operand 2 of a SWITCH_EXPR.
3722 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3724 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3725 Add {q} suffix to insn mnemonic.
3727 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3730 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3731 when memory input operand is handled.
3733 2018-05-23 Bin Cheng <bin.cheng@arm.com>
3735 Backport from mainline
3736 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3737 Richard Biener <rguenther@suse.de>
3739 PR tree-optimization/85793
3740 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3741 for VMAT_ELEMENTWISE.
3743 2018-05-23 Richard Biener <rguenther@suse.de>
3746 * tree-data-ref.c (create_runtime_alias_checks): Defer
3747 and ignore overflow warnings.
3749 2018-05-23 Martin Liska <mliska@suse.cz>
3751 Backport from mainline
3752 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3754 PR tree-optimization/85822
3755 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3758 2018-05-23 Martin Liska <mliska@suse.cz>
3760 Backport from mainline
3761 2018-05-11 Martin Liska <mliska@suse.cz>
3764 * doc/extend.texi: Document LLVM style format for no_sanitize
3767 2018-05-22 Martin Sebor <msebor@redhat.com>
3770 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3772 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3774 PR tree-optimization/85814
3775 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3776 a null return from get_strinfo when unsharing the next
3777 strinfo in the chain.
3779 2018-05-21 Pat Haugen <pthaugen@us.ibm.com>
3781 Backport from mainline
3782 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3783 Segher Boessenkool <segher@kernel.crashing.org>
3786 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest operand.
3788 2018-05-18 Toon Moene <toon@moene.org>
3790 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3791 directly after that of -floop-interchange. Indicate that both
3792 options are enabled by default when specifying -O3.
3794 2018-05-11 Martin Jambor <mjambor@suse.cz>
3797 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3800 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3803 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3805 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3808 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3810 (cortex-m0): Use armv6s-m isa.
3811 (cortex-m0plus): Likewise.
3812 (cortex-m1): Likewise.
3813 (cortex-m0.small-multiply): Likewise.
3814 (cortex-m0plus.small-multiply): Likewise.
3815 (cortex-m1.small-multiply): Likewise.
3817 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
3819 PR rtl-optimization/85638
3820 * bb-reorder.c: Include common/common-target.h.
3821 (create_forwarder_block): New function extracted from...
3822 (fix_up_crossing_landing_pad): ...here. Rename into...
3823 (dw2_fix_up_crossing_landing_pad): ...this.
3824 (sjlj_fix_up_crossing_landing_pad): New function.
3825 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
3826 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
3827 from both partitions and exit the loop after one iteration.
3829 2018-05-08 Richard Biener <rguenther@suse.de>
3831 Backport from mainline
3832 2018-05-04 Richard Biener <rguenther@suse.de>
3835 * fold-const.c (negate_expr_p): Restrict negation of operand
3836 zero of a division to when we know that can happen without
3838 (fold_negate_expr_1): Likewise.
3840 2018-05-03 Richard Biener <rguenther@suse.de>
3842 PR tree-optimization/85615
3843 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3844 to loops not nested in BBs loop father to avoid creating multi-entry
3847 2018-05-02 Richard Biener <rguenther@suse.de>
3850 * gimplify.c (gimplify_save_expr): When in SSA form allow
3851 SAVE_EXPRs to compute to SSA vars.
3853 2018-05-02 Richard Biener <rguenther@suse.de>
3855 PR tree-optimization/85597
3856 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3857 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3860 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3863 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3864 (check_arch): Likewise.
3865 (check_fpu): Return the result rather than printing it.
3866 (end arch): Fix operator precedence.
3867 (end cpu): Likewise.
3868 (END): Print the result from check_fpu.
3870 2018-05-08 Richard Biener <rguenther@suse.de>
3872 Backport from mainline
3873 2018-05-04 Richard Biener <rguenther@suse.de>
3875 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3876 * gimple-ssa-store-merging.c
3877 (imm_store_chain_info::output_merged_store): Remove redundant create,
3878 release split_store vector contents on failure.
3879 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3880 scalar stmt vector on cache hit.
3882 2018-05-04 Richard Biener <rguenther@suse.de>
3885 * tree-complex.c (tree_lower_complex): Handle removed blocks.
3887 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
3889 Backport from mainline.
3890 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3892 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3893 GET_MODE_MASK before any checking.
3894 (nds32_can_use_bset_p): Likewise.
3895 (nds32_can_use_btgl_p): Likewise.
3897 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
3899 Backport from mainline.
3900 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3902 * config/nds32/nds32-doubleword.md: New define_split pattern for
3903 illegal register number.
3905 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
3907 Backport from mainline.
3908 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3912 2018-05-07 Chung-Ju Wu <jasonwucj@gmail.com>
3914 Backport from mainline.
3915 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3917 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3919 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3922 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3923 the type is addressable. Don't force op into register if it has
3926 2018-05-04 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3928 Backport from trunk.
3929 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3932 * config/aarch64/constraints.md (Usg): Limit to 31.
3935 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3937 * BASE-VER: Set to 8.1.1.
3939 2018-05-02 Release Manager
3941 * GCC 8.1.0 released.
3943 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3945 PR tree-optimization/85586
3946 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3947 exit early for statements in the same group if the accesses are
3950 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3953 * doc/install.texi2html: Replace _002d with - and _002a with * in
3954 generated html files using sed.
3956 2018-04-27 Martin Jambor <mjambor@suse.cz>
3959 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3960 the jump function allows for passing through aggregate values.
3962 2018-04-27 Jeff Law <law@redhat.com>
3964 * config.gcc: Mark tile* targets as deprecated/obsolete.
3966 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3968 PR tree-optimization/85529
3969 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3970 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3971 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3972 zero extension or masking of the MSB bit.
3973 (optimize_range_tests): Add FIRST_BB argument, pass it through
3974 to optimize_range_tests_var_bound.
3975 (maybe_optimize_range_tests, reassociate_bb): Adjust
3976 optimize_range_tests callers.
3978 2018-04-26 Richard Biener <rguenther@suse.de>
3979 Jakub Jelinek <jakub@redhat.com>
3981 * cgraph.h (symbol_table): Just declare debug method here.
3982 * symtab.c (symbol_table::debug): Define.
3984 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3986 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3988 2018-04-25 H.J. Lu <hongjiu.lu@intel.com>
3990 Backport from mainline
3991 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3994 * config/i386/i386.c (ix86_expand_builtin): Change memory
3995 operand to XI, extend p0 to Pmode.
3996 * config/i386/i386.md: Change unspec volatile and operand
3997 1 mode to XI, change operand 0 mode to P.
3999 2018-04-25 Jakub Jelinek <jakub@redhat.com>
4001 * DEV-PHASE: Set to prerelease.
4003 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
4005 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
4006 in __abskf2 and __powikf2.
4008 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4011 * config/aarch64/constraints.md (Usg, Usj): New constraints.
4012 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
4013 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
4014 Use the above on operand 2. Reindent.
4015 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
4017 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4020 * common/config/i386/i386-common.c (ix86_handle_option): Don't
4022 * config/i386/i386.opt (mcet): Removed.
4023 * doc/install.texi: Remove -mcet documentation.
4024 * doc/invoke.texi: Likewise.
4026 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
4029 * doc/install.texi: Remove -mcet from bootstrap-cet.
4031 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4034 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
4035 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4039 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4040 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4041 containing a CONST_VECTOR.
4043 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4045 * doc/install.texi: Update newlib dependency for nvptx.
4047 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4050 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4051 instead of INTVAL when shifting x left.
4053 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4055 PR tree-optimization/85478
4056 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4057 vect_grouped_store_supported for single element vectors.
4059 2018-04-24 Richard Biener <rguenther@suse.de>
4062 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4063 load cost increase to the case of non-constant step.
4065 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4068 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4069 destination if any_malformed_asm.
4071 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4074 * expr.c (store_field): In the bitfield case, if the value comes from
4075 a function call and is returned in registers by means of a PARALLEL,
4076 do not change the mode of the temporary unless BLKmode and VOIDmode.
4078 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4080 PR rtl-optimization/85423
4081 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4082 dependencies to debug insns when the previous insn is non-debug.
4084 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4086 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4087 enums into a single definition.
4088 (fls): Fix predicates and printing.
4091 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4093 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4094 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4095 and short u6 immediate.
4096 (check_if_valid_sleep_operand): Remove.
4097 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4099 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4101 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4102 flag_always_save_lp condition.
4103 * config/nds32/nds32.opt (malways-save-lp): New option.
4105 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4107 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4108 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4109 * config/nds32/nds32.h
4110 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4111 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4113 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4115 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4116 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4118 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4119 Chung-Ju Wu <jasonwucj@gmail.com>
4121 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4123 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4124 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4126 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4128 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4130 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4132 * config/nds32/nds32-protos.h (nds32_data_alignment,
4133 nds32_local_alignment): Declare.
4134 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4135 nds32_local_alignment): New functions.
4136 (TARGET_CONSTANT_ALIGNMENT): Define.
4137 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4139 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4141 * config/nds32/nds32.c
4142 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4143 (TARGET_MODES_TIEABLE_P): Likewise.
4145 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4147 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4150 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4151 Chung-Ju Wu <jasonwucj@gmail.com>
4153 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4154 for unaligned access.
4155 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4157 * config/nds32/nds32-intrinsic.md: Likewise.
4158 * config/nds32/nds32_intrinsic.h: Likewise.
4159 * config/nds32/nds32.h (nds32_builtins): Likewise.
4160 * config/nds32/nds32.opt (munaligned-access): New option.
4161 * config/nds32/nds32.c (nds32_asm_file_start): Display
4162 flag_unaligned_access status.
4164 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4166 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4167 -mno-relax is present.
4168 * config/riscv/linux.h (LINK_SPEC): Ditto.
4170 2018-04-20 Martin Sebor <msebor@redhat.com>
4173 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4175 (gimple_fold_builtin_stxcpy_chk): Same.
4176 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4178 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4181 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4182 __powikf2 when long double is IEEE 128-bit.
4184 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4186 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4187 step to make sure stack always aligned.
4189 2018-04-20 Carl Love <cel@us.ibm.com>
4192 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4193 size check for arg0.
4195 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4196 Tom de Vries <tom@codesourcery.com>
4199 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4200 Emit insns for calls too.
4201 (nvptx_find_par): Always look for worker-level predecessor insn.
4202 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4204 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4205 (nvptx_process_pars): Propagate frames for calls.
4207 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4210 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4212 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4213 (ix86_handle_option): Don't handle OPT_mibt.
4214 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4216 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4218 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4219 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4220 (ix86_target_macros): Define __CET__ with flag_cf_protection
4221 for -fcf-protection.
4222 * config/i386/i386.c (isa2_opts): Remove -mibt.
4223 * config/i386/i386.h (TARGET_IBT): Removed.
4224 (TARGET_IBT_P): Likewise.
4225 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4226 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4227 * config/i386/i386.opt (mcet): Update help message.
4230 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4231 -mcet as an alias for -mshstk.
4233 2018-04-20 Richard Biener <rguenther@suse.de>
4236 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4237 complexity by forcing a single use of the multiply operand.
4239 2018-04-20 Martin Jambor <mjambor@suse.cz>
4242 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4243 recursion dependency to only apply to non-clones.
4245 2018-04-20 Martin Jambor <mjambor@suse.cz>
4248 * ipa-cp.c (create_specialized_node): Check that clones of
4249 self-recursive edges exist during IPA-CP.
4251 2018-04-19 Toon Moene <toon@moene.org>
4253 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4256 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4258 PR tree-optimization/85467
4259 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4260 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4261 VECTOR_CST element to type.
4263 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4266 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4267 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4268 (builtin_longjmp): Likewise.
4269 (save_stack_nonlocal): New pattern.
4270 (restore_stack_nonlocal): Likewise.
4272 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4275 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4276 Replace ASM_OUTPUT_LABEL with fprintf.
4278 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4281 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4282 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4283 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4284 define __IBT__ and __SHSTK__ for -fcf-protection.
4285 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4287 (ix86_trampoline_init): Likewise.
4288 (x86_output_mi_thunk): Likewise.
4289 (ix86_notrack_prefixed_insn_p): Likewise.
4290 (ix86_option_override_internal): Don't disallow -fcf-protection.
4291 * config/i386/i386.md (rdssp<mode>): Also enable for
4293 (incssp<mode>): Likewise.
4294 (nop_endbr): Likewise.
4295 * config/i386/i386.opt (mcet): Change help message to built-in
4299 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4300 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4301 enable CET built-in functions.
4303 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4305 * common/config/i386/i386-common.c
4306 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4307 OPTION_MASK_ISA_MOVDIRI_UNSET,
4308 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4309 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4310 * config.gcc (movdirintrin.h): New header.
4311 * config/i386/cpuid.h (bit_MOVDIRI,
4312 bit_MOVDIR64B): New bits.
4313 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4315 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4316 (VOID, PVOID, PCVOID)): New function types.
4317 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4318 __builtin_ia32_directstoreu_u64,
4319 __builtin_ia32_movdir64b): New builtins.
4320 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4321 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4323 (ix86_valid_target_attribute_inner_p): Ditto.
4324 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4325 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4326 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4327 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4328 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4329 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4330 (movdiri<mode>, movdir64b_<mode>): New.
4331 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4332 * config/i386/immintrin.h: Include movdirintrin.h.
4333 * config/i386/movdirintrin.h: New file.
4334 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4336 2018-04-19 Richard Biener <rguenther@suse.de>
4339 * cfg.c (clear_bb_flags): When loop state says we have
4340 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4342 2018-04-19 Richard Biener <rguenther@suse.de>
4344 PR tree-optimization/84737
4345 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4346 copying restrict info.
4347 (vect_setup_realignment): Use it.
4348 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4349 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4350 the first DR to all generated stores.
4351 (vectorizable_load): Likewise for loads.
4353 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4355 PR tree-optimization/85446
4356 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4357 the integral and pointer types to have the same precision.
4359 * doc/install.texi: Document --disable-cet being the default and
4362 2018-04-18 Martin Liska <mliska@suse.cz>
4364 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4367 2018-04-18 Martin Liska <mliska@suse.cz>
4370 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4373 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4374 arguments if they are comparable.
4376 2018-04-18 Martin Liska <mliska@suse.cz>
4379 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4382 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4385 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4388 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4389 ENDBR after calling __morestack.
4391 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4394 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4395 by using gcc_base_ver to generate a gcc_driver_version, and use
4396 it when generating GCC_DRIVER_NAME.
4397 * configure: Regenerate.
4399 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4402 * config.gcc: Obsolete powerpc*-*-*spe*.
4404 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4407 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4408 (stabstr_D): Change type of unum from unsigned int to
4409 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4412 2018-04-17 Jim Wilson <jimw@sifive.com>
4415 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4416 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4417 Set arg_pointer_offset after using pretend_args_size.
4419 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4421 PR rtl-optimization/85431
4422 * dse.c (record_store): Ignore zero width stores.
4425 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4426 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4427 __builtin_stack_restore rather than after it.
4428 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4429 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4430 argument instead of virtual_dynamic_stack_rtx.
4432 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4434 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4436 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4437 Add note to error message to explain internal mapping of overloaded
4438 built-in function name to non-overloaded built-in function name.
4439 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4442 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4445 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4446 where the inputs overlap with the output.
4448 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4450 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4451 (=v, v) alternative and explicit "memory" attribute.
4452 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4453 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4455 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4456 "sselog1" type instead of "sselog".
4457 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4458 "sselog". Remove explicit "memory" attribute.
4459 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4460 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4462 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4463 "sselog1" type instead of "sselog". Remove explicit "memory"
4465 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4466 use "sselog1" type instead of "sselog". Remove explicit "memory"
4468 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4469 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4471 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4472 "sselog1" type instead of "sselog". Remove explicit "memory"
4474 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4475 use "sselog1" type instead of "sselog". Remove explicit "memory"
4479 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4482 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4485 2018-04-17 Martin Jambor <mjambor@suse.cz>
4488 * ipa-cp.c (create_specialized_node): Call
4489 expand_all_artificial_thunks if necessary.
4491 2018-04-17 Martin Liska <mliska@suse.cz>
4494 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4495 in message, remote space in between '_G' and '('.
4497 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4500 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4501 avx512f_vmcmp<mode>3<round_saeonly_name>,
4502 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4503 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4504 avx512f_rndscale<mode><round_saeonly_name>,
4505 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4506 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4507 Use %<iptr>2 instead of %2 for -masm=intel.
4508 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4509 avx512f_vcvttss2usi<round_saeonly_name>,
4510 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4512 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4513 avx512f_vcvttsd2usi<round_saeonly_name>,
4514 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4515 Use %q1 instead of %1 for -masm=intel.
4516 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4517 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4518 of %3 for -masm=intel.
4519 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4521 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4523 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4525 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4526 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4528 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4530 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4531 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4532 %g1 and one with %0 and %1.
4533 (avx512er_vmrcp28<mode><round_saeonly_name>,
4534 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4536 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4537 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4538 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4539 of %0 and %{%4%} for -masm=intel.
4540 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4541 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4542 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4543 order of %0 and %{%5%}%{z%} for -masm=intel.
4545 2018-04-17 Jan Hubicka <jh@suse.cz>
4548 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4550 2018-04-17 Martin Liska <mliska@suse.cz>
4553 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4554 for target_clone error message. Make default implementation
4555 clone to be a local declaration.
4556 (separate_attrs): Add new argument and check for an empty
4558 (expand_target_clones): Handle it.
4559 (ipa_target_clone): Make redirection just for target_clones
4562 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4563 Tom de Vries <tom@codesourcery.com>
4566 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4567 tiled basic blocks without omp continue statements.
4569 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4572 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4573 vec_extract expression as having side effects to make sure it gets
4576 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4579 * config/i386/i386.c (get_builtin_code_for_version): Check
4582 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4585 * gcc/config.gcc: Support "skylake".
4586 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4588 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4589 (processor_target_table): Add "skylake".
4590 (ix86_option_override_internal): Add "skylake".
4591 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4592 PROCESSOR_CANNONLAKE.
4593 (get_builtin_code_for_version): Fix priority for
4594 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4595 PROCESSOR_SKYLAKE-AVX512.
4596 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4597 (processor_type): Add PROCESSOR_SKYLAKE.
4599 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4600 Jason Merrill <jason@redhat.com>
4603 * convert.c (convert_to_integer_1): Use direct recursion for
4604 enumeral types and types with a precision less than the number
4605 of bits in their mode.
4607 2018-04-16 Julia Koval <julia.koval@intel.com>
4610 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4611 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4613 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4616 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4617 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4618 and -mno-direct-move.
4620 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4623 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4624 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4625 Ensure negative shifts result in {0}.
4627 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4629 PR rtl-optimization/79916
4630 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4631 regs (if any) to define how to gnerate SD moves when LRA is in
4634 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4636 PR rtl-optimization/85393
4637 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4638 * except.c (expand_dw2_landing_pad_for_region): Make static.
4639 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4640 a label and unconditional jump to old_bb, rather than
4641 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4644 PR rtl-optimization/85376
4645 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4646 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4647 instead of a specific value.
4649 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4650 Bin Cheng <bin.cheng@arm.com>
4652 PR tree-optimization/82965
4653 PR tree-optimization/83991
4654 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4655 by_profile_only parameter.
4656 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4657 information if the loop was predicted to iterate too many times.
4658 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4660 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4663 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4666 2018-04-13 Martin Liska <mliska@suse.cz>
4667 Jakub Jelinek <jakub@redhat.com>
4670 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4671 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4672 * builtins.c (expand_builtin_memory_copy_args): Use
4673 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4674 handle dest_addr == pc_rtx.
4676 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4679 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4680 asked to not generate direct moves.
4681 (fix_trunc<mode>si2_stfiwx): Similar.
4682 (fix_trunc<mode>si2_internal): Similar.
4684 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4687 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4688 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4689 lookup if dest in some wider mode is known to be const0_rtx and
4690 if so, record permanent equivalence for it to be ZERO_EXTEND of
4691 the narrower mode destination.
4693 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4695 * lto-streamer-out.c (output_function): Revert 259346.
4696 * omp-expand.c (expand_oacc_for): Likewise.
4698 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4700 PR rtl-optimization/85354
4701 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4702 * sel-sched.c (sel_global_init): ... here.
4704 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4707 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4708 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4709 mode for PE-COFF targets.
4710 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4711 (i386_pe_asm_lto_end): Likewise.
4712 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4713 (TARGET_ASM_LTO_END): Likewise.
4714 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4715 (i386_pe_asm_lto_start): New function.
4716 (i386_pe_asm_lto_end): Likewise.
4718 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4719 Richard Biener <rguenther@suse.de>
4722 * lto-streamer-out.c (output_function): Fix CFG loop state before
4724 * omp-expand.c (expand_oacc_for): Handle calls to internal
4725 functions like regular functions.
4727 2018-04-12 Richard Biener <rguenther@suse.de>
4730 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4731 for the early LTO debug to properly generate references to it
4732 during DIE emission. Do not re-use that for the skeleton for
4734 (dwarf2out_early_finish): Likewise.
4736 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4739 * config/i386/sse.md
4740 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4741 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4742 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4743 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4744 and output is a reg, avoid creating invalid lowpart subreg, but
4745 instead split into a 512-bit move. Don't split if not AVX512VL,
4746 input is xmm16+ reg and output is a mem.
4747 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4748 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4749 xmm16+ reg and output is a mem.
4751 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4753 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4754 also for flag_dwarf2_cfi_asm.
4756 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4758 PR rtl-optimization/85342
4759 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4760 a bool scalar var inside of the loop instead. Don't try to update
4761 recog_data.operand after failed apply_change_group.
4763 2018-04-12 Tom de Vries <tom@codesourcery.com>
4766 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4767 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4768 array with flexible array member as array without given dimension.
4769 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4770 argument for undefined param to true.
4772 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4775 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4776 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4777 from PowerPC section.
4778 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4779 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4780 help text that is too long.
4781 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4782 help text that is too long.
4783 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4784 help text that is too long.
4786 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4788 * config/alpha/alpha.md (stack_probe_internal): Rename
4789 from "probe_stack". Update all callers.
4791 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4793 PR rtl-optimization/84566
4794 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4795 sched_macro_fuse_insns.
4797 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4800 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4801 (compute_block_dependences): ... from here.
4803 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4805 PR tree-optimization/85331
4806 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4807 from int to HOST_WIDE_INT.
4809 2018-04-11 Martin Jambor <mjambor@suse.cz>
4812 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4813 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4814 not the same as the source val.
4815 (cgraph_edge_brings_value_p): New parameter.
4816 (gather_edges_for_value): Pass destination value to
4817 cgraph_edge_brings_value_p.
4818 (perhaps_add_new_callers): Likewise.
4819 (get_info_about_necessary_edges): Likewise and exclude values brought
4820 only by self-recursive edges.
4821 (create_specialized_node): Redirect only clones of self-calling edges.
4822 (+self_recursive_pass_through_p): New function.
4823 (find_more_scalar_values_for_callers_subset): Use it.
4824 (find_aggregate_values_for_callers_subset): Likewise.
4825 (known_aggs_to_agg_replacement_list): Removed.
4826 (decide_whether_version_node): Re-calculate known constants for all
4827 remaining context clones.
4829 2018-04-11 Richard Biener <rguenther@suse.de>
4832 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4833 from early DWARF output.
4834 (dwarf2out_early_finish): Output line info unconditionally into
4835 early DWARF and add reference to it.
4837 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4840 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4841 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4842 other than V2DFmode using iptr mode attribute.
4843 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4845 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4847 PR rtl-optimization/84659
4848 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4850 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4853 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4855 (output_loc_list): Pass address of a dummy size variable even in the
4856 locview handling loop.
4857 (index_location_lists): Add comment on why skip_loc_list_entry can't
4860 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4863 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4866 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4869 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4870 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4871 and -mstring-compare-inline-limit.
4873 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4876 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4877 for stack clash protection in a register whenever we need it to be in
4880 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4882 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4883 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4885 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4888 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4890 (mlong-double-): Ditto.
4891 * config/rs6000/sysv4.opt (msdata=): Ditto.
4892 (mtls-size=): Ditto.
4894 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4896 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4897 erroneous entries for
4898 "vector int vec_ldl (int, long int *)", and
4899 "vector unsigned int vec_ldl (int, unsigned long int *)".
4900 Add comments and entries for
4901 "vector bool char vec_ldl (int, bool char *)",
4902 "vector bool short vec_ldl (int, bool short *)",
4903 "vector bool int vec_ldl (int, bool int *)",
4904 "vector bool long long vec_ldl (int, bool long long *)",
4905 "vector pixel vec_ldl (int, pixel *)",
4906 "vector long long vec_ldl (int, long long *)",
4907 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4908 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4909 type tree bool_long_long_type_node and correct definition of
4910 bool_V2DI_type_node to make reference to this new type tree.
4911 (rs6000_mangle_type): Replace erroneous reference to
4912 bool_long_type_node with bool_long_long_type_node.
4913 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4914 comments to emphasize sign distinctions for char and int types and
4915 replace RS6000_BTI_bool_long constant with
4916 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4917 use of RS6000_BTI_pixel.
4918 (bool_long_type_node): Remove this macro definition.
4919 (bool_long_long_type_node): New macro definition
4921 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4923 PR rtl-optimization/85300
4924 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4925 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4926 simplify_unary_operation fails.
4928 2018-04-10 Martin Liska <mliska@suse.cz>
4930 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4931 cgraph_edge and ipa_ref.
4933 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4937 * config/i386/sse.md
4938 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4939 computation of the VEC_MERGE selector from mask.
4940 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4941 Fix decoding of the VEC_MERGE selector into mask.
4943 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4945 PR tree-optimization/85286
4946 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4948 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4950 * final.c (final_1): Set insn_last_address as well as
4951 insn_current_address.
4953 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4956 * explow.c (emit_stack_probe): Call validize_mem on memory location
4957 before passing it to gen_probe_stack. Create address operand and
4958 legitimize it for the probe_stack_address case.
4960 2018-04-09 Jan Hubicka <jh@suse.cz>
4963 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4964 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4965 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4967 (free_lang_data): Rebuild type inheritance graph.
4969 2018-04-09 Martin Sebor <msebor@redhat.com>
4971 * invoke.texi (-finline-small-functions): Mention other optimization
4973 (-findirect-inlining, -fpartial-inlining): Same.
4974 (-finline-functions-called-once): Same.
4975 (-freorder-blocks-and-partition): Same.
4977 2018-04-09 Jan Hubicka <jh@suse.cz>
4980 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4981 jumps; choose last target that matches the criteria (i.e.
4982 no partition changes for non-crossing jumps).
4983 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4984 support for redirecting crossing jumps to non-crossing.
4986 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4988 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4989 also for naked functions.
4991 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4993 * config/arc/arc.md (add_shift): New pattern.
4994 (add_shift2): Likewise.
4995 (sub_shift): Likewise.
4996 (sub_shift_cmp0_noout): Likewise.
4997 (compare_si_ashiftsi): Likewise.
4998 (xbfu_cmp0_noout): New combine pattern.
4999 (xbfu_cmp0"): Likewise.
5000 (movsi_set_cc_insn): Place the predicable variant first.
5001 (commutative_binary_cmp0_noout): Remove clobber.
5002 (commutative_binary_cmp0): New pattern.
5003 (noncommutative_binary_cmp0): Likewise.
5004 (noncommutative_binary_cmp0_noout): Likewise.
5005 (noncommutative_binary_comparison_result_used): Removed.
5006 (rsub_cmp0): New pattern.
5007 (rsub_cmp0_noout): Likewise.
5008 (extzvsi): Changed, keep only meaningful variants.
5009 (SQH, SEZ): New iterators.
5010 (SQH_postfix): New mode attribute.
5011 (SEZ_prefix): New code attribute.
5012 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
5013 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
5014 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
5016 (noncommutative_operator): Check the availability of barrel
5019 2018-04-09 Richard Biener <rguenther@suse.de>
5021 PR tree-optimization/85284
5022 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
5023 Only use the niter constraining form of simple_iv when the exit
5026 2018-04-09 Tom de Vries <tom@codesourcery.com>
5029 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5030 (define_expand "*memory_barrier"): New define_expand.
5031 (define_insn "memory_barrier"): New insn.
5033 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5035 PR rtl-optimization/80463
5036 PR rtl-optimization/83972
5037 PR rtl-optimization/83480
5039 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5040 correct producer for the insn.
5041 (tidy_control_flow): Fixup seqnos in case of debug insns.
5043 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5045 PR rtl-optimization/83913
5047 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5048 different sched-times when merging exprs.
5050 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5052 PR rtl-optimization/83962
5054 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5055 tidy_fallthru_edge and tidy_control_flow.
5057 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5059 PR rtl-optimization/83530
5061 * sel-sched.c (force_next_insn): New global variable.
5062 (remove_insn_for_debug): When force_next_insn is true, also leave only
5063 next insn in the ready list.
5064 (sel_sched_region): When the region wasn't scheduled, make another pass
5065 over it with force_next_insn set to 1.
5067 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5069 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5071 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5072 for interrupt control.
5073 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5074 functions for interrupt control.
5075 * config/nds32/nds32-intrinsic.md: Likewise.
5076 * config/nds32/nds32_intrinsic.h: Likewise.
5077 * config/nds32/nds32.h (nds32_builtins): Likewise.
5079 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5081 * config/nds32/nds32.c (nds32_init_machine_status,
5082 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5083 strict_aligned_p field.
5084 (nds32_expand_to_rtl_hook): New function.
5085 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5086 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5088 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5089 Chung-Ju Wu <jasonwucj@gmail.com>
5091 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5092 * config/nds32/nds32-n7.md: New file.
5093 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5094 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5096 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5097 * config/nds32/nds32.md (pipeline_model): Add n7.
5098 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5099 * config/nds32/pipelines.md: Include n7 settings.
5101 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5102 Chung-Ju Wu <jasonwucj@gmail.com>
5104 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5105 * config/nds32/nds32-e8.md: New file.
5106 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5107 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5109 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5110 * config/nds32/nds32.md (pipeline_model): Add e8.
5111 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5112 * config/nds32/pipelines.md: Include e8 settings.
5114 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5115 Chung-Ju Wu <jasonwucj@gmail.com>
5117 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5118 * config/nds32/nds32-n8.md: New file.
5119 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5120 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5122 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5123 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5124 * config/nds32/nds32.md (pipeline_model): Add n8.
5125 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5126 * config/nds32/pipelines.md: Include n8 settings.
5128 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5129 Chung-Ju Wu <jasonwucj@gmail.com>
5131 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5132 * config/nds32/nds32-n9-2r1w.md: New file.
5133 * config/nds32/nds32-n9-3r2w.md: New file.
5134 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5135 nds32_register_ports): New or modify for cpu n9.
5136 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5138 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5139 * config/nds32/nds32-utils.c: New file.
5140 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5141 TARGET_MUL_SLOW): Define.
5142 * config/nds32/nds32.md (pipeline_model): New attribute.
5143 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5144 New options that support cpu n9.
5145 * config/nds32/pipelines.md: Include n9 settings.
5146 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5148 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5150 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5151 information if necessary.
5152 (output_cond_branch_compare_zero): Likewise.
5153 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5154 (nds32_target_alignment): Refine for alignment.
5155 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5156 (FUNCTION_BOUNDARY): Modify.
5157 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5159 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5161 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5163 * config/nds32/constants.md (unspec_volatile_element): Add values for
5164 TLB operation and data prefetch.
5165 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5166 functions for TLB operation and data prefetch.
5167 * config/nds32/nds32-intrinsic.md: Likewise.
5168 * config/nds32/nds32_intrinsic.h: Likewise.
5169 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5170 (nds32_print_operand): Likewise.
5171 * config/nds32/nds32.h (nds32_builtins): Likewise.
5173 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5174 Andrew Pinski <pinsika@gcc.gnu.org>
5177 * match.pd: Use constant_boolean_node of correct type instead of
5178 boolean_true_node or boolean_false_node for simplifying
5179 pointer comparisons to zero.
5181 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5183 PR tree-optimization/80021
5184 * tree.c (verify_type_variant): Make error call in verify_variant_match
5185 translatable and remove final full stop.
5187 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5189 * config/nds32/constants.md (unspec_volatile_element): Add
5190 UNSPEC_VOLATILE_EH_RETURN.
5191 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5192 nds32_output_stack_pop): Support dwarf exception handling process.
5193 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5194 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5195 exception handling process.
5196 (nds32_compute_stack_frame): Likewise.
5197 (nds32_return_addr_rtx): Likewise.
5198 (nds32_initial_elimination_offset): Likewise.
5199 (nds32_expand_prologue): Likewise.
5200 (nds32_expand_epilogue): Likewise.
5201 (nds32_dynamic_chain_address): New function.
5202 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5204 (DYNAMIC_CHAIN_ADDRESS): Define.
5205 (EH_RETURN_DATA_REGNO): Define.
5206 (EH_RETURN_STACKADJ_RTX): Define.
5207 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5208 patterns for dwarf exception handling.
5210 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5212 * config/nds32/nds32.h: Clean up obsolete macros.
5214 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5216 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5217 Add enum values for particular instructions.
5218 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5219 particular intrinsic functions.
5220 * config/nds32/nds32-intrinsic.md: Likewise.
5221 * config/nds32/nds32_intrinsic.h: Likewise.
5222 * config/nds32/nds32.h (nds32_builtins): Likewise.
5223 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5224 (btst, ave): New patterns for particular instructions.
5226 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5228 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5229 Add enum values for atomic load/store and memory sync.
5230 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5232 * config/nds32/nds32-intrinsic.md: Likewise.
5233 * config/nds32/nds32_intrinsic.h: Likewise.
5234 * config/nds32/nds32.h (nds32_builtins): Likewise.
5236 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5238 PR tree-optimization/85257
5239 * fold-const.c (native_encode_vector): If not all elts could fit
5240 and off is -1, return 0 rather than offset.
5241 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5242 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5243 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5244 adjust buffer in native_interpret_expr call.
5246 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5248 * config/nds32/constants.md (unspec_volatile_element): Add cache
5249 control enum values.
5250 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5251 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5252 * config/nds32/nds32.c (nds32_cctl_names): New.
5253 (nds32_print_operand): Handle cache control register names.
5254 * config/nds32/nds32.h (nds32_builtins): New enum values.
5255 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5257 * config/nds32/nds32.md (type): Add mmu.
5258 * config/nds32/pipelines.md (simple_insn): Add mmu.
5260 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5262 * config/nds32/nds32.md (type): Remove call.
5263 * config/nds32/pipelines.md (simple_insn): Likewise.
5265 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5267 * config/nds32/constants.md (unspec_volatile_element): Add
5268 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5269 UNSPEC_VOLATILE_FMFCFG.
5270 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5271 description for fmfcfg and fmfcsr.
5272 (bdesc_1arg): Add fmtcsr.
5273 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5274 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5275 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5276 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5277 unspec_fmfcfg): New patterns.
5278 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5279 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5280 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5281 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5282 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5283 __nds32__fmfcfg): Define.
5285 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5287 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5288 intrinsic register names.
5289 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5290 intrinsic register enum values and macros.
5292 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5294 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5295 for load/store addressing form.
5296 (nds32_print_operand_address): Likewise.
5298 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5301 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5302 based on LABEL_REF. Remove useless assertion.
5303 (pic_address_needs_scratch): Fix formatting.
5304 (sparc_legitimize_pic_address): Minor tweaks.
5305 (sparc_delegitimize_address): Adjust assertion accordingly.
5306 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5307 into symbolic_operand.
5308 (movsi_high_pic_label_ref): Likewise.
5309 (movsi_lo_sum_pic_label_ref): Likewise.
5310 (movdi_pic_label_ref): Likewise.
5311 (movdi_high_pic_label_ref): Likewise.
5312 (movdi_lo_sum_pic_label_ref): Likewise.
5314 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5316 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5317 custom LIB_SPEC setup.
5319 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5320 Kito Cheng <kito.cheng@gmail.com>
5322 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5323 * config/riscv/freebsd.h: New.
5325 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5327 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5328 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5331 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5332 Kito Cheng <kito.cheng@gmail.com>
5334 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5335 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5336 * config/nds32/nds32-protos.h (nds32_output_call,
5337 nds32_output_return): Declare.
5338 * config/nds32/nds32.md: Refine all the call and return patterns.
5340 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5343 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5344 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5346 PR rtl-optimization/84872
5347 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5348 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5351 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5353 * expr.c (copy_blkmode_to_reg): Revert 254862.
5354 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5356 2018-04-06 Richard Biener <rguenther@suse.de>
5359 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5360 after seeing a component reference with an adjacent field. Treat
5361 refs to arrays at struct end of external decls similar to
5362 refs to unconstrained commons.
5364 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5367 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5368 look through SAVE_EXPRs with non-side-effects argument. Adjust
5370 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5373 2018-04-06 Richard Biener <rguenther@suse.de>
5376 * alias.c (find_base_term): New wrapper around find_base_term
5377 unwinding CSELIB_VAL_PTR changes.
5378 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5381 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5383 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5385 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5386 constant definitions.
5387 ("nop"): lr 0,0 -> nopr r0
5388 ("nop_lr0", "nop_lr1"): New insn definitions.
5390 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5392 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5393 NDS32_V3PUSH_AVAILABLE_P macro.
5395 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5396 Chung-Ju Wu <jasonwucj@gmail.com>
5398 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5399 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5400 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5401 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5402 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5403 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5404 * config/nds32/constraints.md: New constraints and checking for hard
5405 float configuration.
5406 * config/nds32/iterators.md: New mode iterator and attribute for hard
5407 float configuration.
5408 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5410 * config/nds32/nds32-fpu.md: New file.
5411 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5412 deal with hard float code generation.
5413 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5415 (abi_type, float_reg_number): New enum type.
5416 * config/nds32/nds32-predicates.c: New predicates for hard float.
5417 * config/nds32/nds32-protos.h: Declare functions for hard float.
5418 * config/nds32/nds32.c: Implementation for hard float configuration.
5419 * config/nds32/nds32.h: Definitions for hard float configuration.
5420 * config/nds32/nds32.md: Include hard float machine description and
5421 modify patterns for hard float configuration.
5422 * config/nds32/nds32.opt: New options for hard float configuration.
5423 * config/nds32/predicates.md: New predicates for hard float
5426 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5428 * common/config/nds32/nds32-common.c
5429 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5431 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5434 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5435 CONSTRUCTOR_ELT (ctor, ...)->value.
5437 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5440 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5442 2018-04-05 Tom de Vries <tom@codesourcery.com>
5445 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5448 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5449 Kito Cheng <kito.cheng@gmail.com>
5451 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5452 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5453 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5454 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5456 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5457 Kito Cheng <kito.cheng@gmail.com>
5459 * config/nds32/constraints.md (Ufe): New memory constraint.
5460 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5461 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5462 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5464 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5465 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5467 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5469 * config/nds32/nds32.md: Use optimize_size in the condition for
5470 alu-shift instructions.
5472 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5474 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5476 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5478 * config/nds32/nds32.md (negsi2): Refine pattern.
5480 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5481 Chung-Ju Wu <jasonwucj@gmail.com>
5483 * config/nds32/iterators.md (shift_rotate): New code iterator.
5484 (shift): New code attribute.
5485 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5486 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5487 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5488 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5489 bit-wise operations.
5490 (andsi3, *andsi3): Ditto.
5491 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5492 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5493 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5494 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5495 nds32_ior_operand, nds32_xor_operand): New predicates.
5497 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5499 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5500 (addsi3, subsi3): ... this.
5502 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5504 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5506 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5508 * config/nds32/nds32.md: Adjust indention.
5510 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5512 * config/nds32/nds32.md (feature): New attribute.
5514 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5516 * config/nds32/nds32.md (subtype): New attribute.
5518 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5521 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5522 expansion to perform a bitwise AND of the argument followed by a
5523 boolean negation of the result.
5525 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5527 PR rtl-optimization/84878
5528 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5529 the basic block. Assert the use reference is not artificial and that
5530 it has an associated insn.
5532 2018-04-04 Michael Matz <matz@suse.de>
5534 * builtins.c (compute_objsize): Pass correct operand
5535 to array_at_struct_end_p.
5537 2018-04-04 Richard Biener <rguenther@suse.de>
5540 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5541 from contexts for DINFO_LEVEL_TERSE and below.
5543 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5545 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5546 resiter_operand condition.
5547 * config/nds32/nds32.md (*move<mode>): Ditto.
5549 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5550 Monk Chiang <sh.chiang04@gmail.com>
5552 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5554 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5556 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5558 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5559 Kito Cheng <kito.cheng@gmail.com>
5561 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5562 nds32_cond_code_str, output_cond_branch,
5563 output_cond_branch_compare_zero, nds32_expand_cbranch,
5564 nds32_expand_cstore, nds32_expand_movcc,
5565 nds32_output_cbranchsi4_equality_zero,
5566 nds32_output_cbranchsi4_equality_reg,
5567 nds32_output_cbranchsi4_equality_reg_or_const_int,
5568 nds32_output_cbranchsi4_greater_less_zero: New functions.
5569 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5570 nds32_expand_cstore, nds32_expand_movcc,
5571 nds32_output_cbranchsi4_equality_zero,
5572 nds32_output_cbranchsi4_equality_reg,
5573 nds32_output_cbranchsi4_equality_reg_or_const_int,
5574 nds32_output_cbranchsi4_greater_less_zero): Declare.
5575 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5576 nds32_rimm11s_operand): New predicates.
5577 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5578 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5581 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5583 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5584 * config/nds32/nds32.md: Ditto.
5585 * config/nds32/pipelines.md: Ditto.
5587 2018-04-04 Richard Biener <rguenther@suse.de>
5589 PR tree-optimization/85168
5590 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5591 propagating abnormals.
5593 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5595 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5597 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5598 Kito Cheng <kito.cheng@gmail.com>
5600 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5601 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5602 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5603 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5604 * config/nds32/nds32.md (sibcall_internal): New.
5605 (sibcall_register): Remove.
5606 (sibcall_immediate): Remove.
5607 (sibcall_value_internal): New.
5608 (sibcall_value_register): Remove.
5609 (sibcall_value_immediate): Remove.
5610 * config/nds32/predicates.md (nds32_general_register_operand): New.
5611 (nds32_call_address_operand): New.
5613 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5615 PR rtl-optimization/85167
5616 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5617 bb_defs if *split_p, instead preinitialize it to NULL.
5619 PR tree-optimization/85156
5620 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5621 evaluating the argument multiple times.
5623 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5625 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5627 (_mm_cvtpd_ps): Likewise.
5628 (_mm_cvttpd_epi32): Likewise.
5629 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5630 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5631 vector, pixel, and bool following altivec.h include.
5633 2018-04-03 Martin Sebor <msebor@redhat.com>
5635 * doc/extend.texi (Common Function Attributes): Clarify.
5636 (const attribute): Likewise.
5637 (pure attribute): Likewise.
5639 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5642 * config/i386/i386.c (ix86_expand_vector_set): Use
5643 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5645 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5647 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5648 instructions when changing rounding bits to preserve precision bits
5649 in the x87 control word.
5651 2018-04-03 Martin Liska <mliska@suse.cz>
5653 PR tree-optimization/82491
5654 * rtl.h (strip_offset_and_add): Replace += suboffset with
5655 poly_uint64 () + suboffset.
5657 2018-03-29 Martin Liska <mliska@suse.cz>
5658 Martin Jambor <mjambor@suse.cz>
5661 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5662 param_type is not an integral or pointer type.
5664 2018-04-03 Richard Biener <rguenther@suse.de>
5666 * sese.h (recompute_all_dominators): Remove.
5668 2018-04-02 Martin Sebor <msebor@redhat.com>
5670 * doc/invoke.texi (-Wrestrict): Fix typos.
5672 2018-04-02 Jim Wilson <jimw@sifive.com>
5674 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5675 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5676 (<optab>di3, <optab>si3_extend): Likewise.
5677 (<optab>si3_mask, <optab>si3_mask_1): New.
5678 (<optab>di3_mask, <optab>di3_mask_1): New.
5679 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5680 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5681 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5683 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5685 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5688 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5690 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5691 (nds32_canonicalize_comparison): New function.
5693 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5694 Kito Cheng <kito.cheng@gmail.com>
5695 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5697 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5698 * config/nds32/constants.md (unspec_volatile_element): Add
5699 UNSPEC_VOLATILE_RELAX_GROUP.
5700 * config/nds32/nds32-relax-opt.c: New file.
5701 * config/nds32/nds32-predicates.c
5702 (nds32_symbol_load_store_p): New function.
5703 * config/nds32/nds32-protos.h
5704 (nds32_symbol_load_store_p): Declare function.
5705 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5706 * config/nds32/nds32.c
5707 (nds32_register_pass): New function to register pass.
5708 (nds32_register_passes): New function to register passes.
5709 * config/nds32/nds32.md (relax_group): New pattern.
5710 * config/nds32/nds32.opt (mrelax-hint): New option.
5711 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5713 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5715 * config/nds32/t-nds32: Modify files dependency.
5717 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5719 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5720 (PROFILE_HOOK): Define its implementation.
5722 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5724 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5725 type and 32-bit size.
5727 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5730 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5731 (V_128_256): New mode iterator.
5732 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5733 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5734 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5736 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5737 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5739 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5742 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5743 NaN inputs correctly.
5745 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5748 * config/rs6000/vsx.md (??r): New mode attribute.
5749 (*vsx_mov<mode>_64bit): Use it.
5750 (*vsx_mov<mode>_32bit): Likewise.
5752 2018-03-30 Martin Sebor <msebor@redhat.com>
5754 PR tree-optimization/84818
5755 * builtins.c (check_access): Use warning_n.
5757 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5760 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5762 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5765 2018-03-30 Julia Koval <julia.koval@intel.com>
5768 * x86-tune.def (movx, partial_reg_dependency): Enable for
5771 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5774 * lra-constraints.c (process_alt_operands): Move setting
5775 this_alternative_matches below.
5777 2018-03-29 Martin Liska <mliska@suse.cz>
5780 * doc/invoke.texi: Document how LTO works with debug info.
5781 Describe auto-load support of binutils. Mention 'x86-64'
5782 as valid option value of -march option.
5784 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5786 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5789 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5790 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5791 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5794 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5797 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5798 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5799 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5800 for RS6000_BTM_POWERPC64.
5801 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5802 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5803 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5808 2018-03-28 Carl Love <cel@us.ibm.com>
5811 2017-09-27 Carl Love <cel@us.ibm.com>
5813 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5814 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5815 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5818 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5820 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5821 instead of __vector bool.
5822 (_mm_max_pu8): Likewise.
5823 (_mm_min_pi16): Likewise.
5825 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5828 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5829 (DIVWEUO): Likewise.
5831 (DIVDEUO): Likewise.
5832 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5833 DIVWEUO and DIVDEUO.
5834 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5835 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5836 (div_extend): Likewise.
5837 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5839 (__builtin_divweuo): Likewise.
5840 (__builtin_divdeo): Likewise.
5841 (__builtin_divdeuo): Likewise.
5843 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5846 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5847 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5849 PR tree-optimization/82004
5850 * gimple-match-head.c (optimize_pow_to_exp): New function.
5851 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5852 Don't fold to exp if optimize_pow_to_exp is false.
5854 2018-03-28 Martin Liska <mliska@suse.cz>
5857 * calls.c (initialize_argument_information): Fix trailing space.
5858 * common.opt: Fix typo and provide better explanation for
5859 -fsanitize-coverage option.
5860 * config/i386/i386.opt: Fix typo.
5862 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5863 Martin Liska <mliska@suse.cz>
5866 * gimplify.c (asan_poison_variable): Don't do the check for
5867 gimplify_omp_ctxp here.
5868 (gimplify_decl_expr): Do it here.
5869 (gimplify_target_expr): Likewise.
5871 2018-03-28 Martin Liska <mliska@suse.cz>
5874 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5875 chkp_type_bounds_count if MPX is not enabled.
5877 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5879 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5881 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5884 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5885 function to create the function decl for complex long double
5886 multiply and divide for -mabi=ieeelongdouble.
5887 (init_float128_ieee): Call it.
5889 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5892 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5893 -fcf-protection=branch -mibt.
5894 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5896 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5899 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5901 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5904 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5905 extern array declarations.
5907 2018-03-27 Richard Biener <rguenther@suse.de>
5910 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5911 explicit single_use checks.
5913 2018-03-27 Richard Biener <rguenther@suse.de>
5915 PR tree-optimization/85082
5916 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5919 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5921 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5922 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5923 Turn on fasynchronous-unwind-tables and funwind-tables.
5925 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5928 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5929 (*bmi_blsr_<mode>_ccz): Ditto.
5931 2018-03-26 Tom de Vries <tom@codesourcery.com>
5933 PR tree-optimization/85063
5934 * omp-general.c (offloading_function_p): New function. Factor out
5936 * omp-offload.c (pass_omp_target_link::gate): ... here.
5937 * omp-general.h (offloading_function_p): Declare.
5938 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5939 with attribute omp declare target for offloading functions.
5941 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5943 PR tree-optimization/84005
5944 * tree-data-ref.h (get_base_for_alignment): Declare.
5945 * tree-data-ref.c (get_base_for_alignment_1): New function.
5946 (get_base_for_alignment): Likewise.
5947 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5948 get_base_for_alignment to find a suitable base object, instead
5949 of always using drb->base_address.
5951 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5954 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5955 known size by default.
5957 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5960 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5961 and non BLKmode operands.
5963 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5966 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5967 Clean up attributes.
5969 2018-03-23 Richard Biener <rguenther@suse.de>
5972 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5973 we are going to emit early debug for LTO.
5975 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5978 * function.c (match_asm_constraints_1): Don't optimize if input
5979 doesn't satisfy general_operand predicate for output's mode.
5982 * alias.c (write_dependence_p): Don't require for x_canonicalized
5983 non-VOIDmode if x has VOIDmode.
5986 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5987 just don't try to optimize it rather than assert it never happens.
5989 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5991 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5992 macro expansions for definition of ST_INTERNAL_<mode> and
5993 LD_INTERNAL_<mode> builtins.
5994 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5996 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5998 (altivec_expand_st_builtin): Likewise.
5999 (altivec_expand_builtin): Remove calls to deleted functions.
6000 (rs6000_address_for_altivec): Delete this function.
6001 * config/rs6000/vector.md: Remove expands for
6002 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
6004 2018-03-22 Sudakshina Das <sudi.das@arm.com>
6007 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
6008 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
6009 re-computing once computed.
6010 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
6011 (arm_init_machine_status): Initialize
6012 machine->static_chain_stack_bytes.
6014 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6017 * doc/extend.texi: Add four new prototypes for vec_ld.
6018 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
6019 definitions for more logical presentation.
6020 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
6021 entries for V1TI variants of __builtin_altivec_ld builtin.
6022 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
6023 handling of V1TI variant of LVX icode pattern.
6024 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
6025 (rs6000_gimple_fold_builtin): Likewise.
6026 (altivec_init_builtins): Add code to define
6027 __builtin_altivec_lvx_v1ti function.
6029 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6032 * function.c (match_asm_constraints_1): Don't do the optimization
6033 if input isn't a REG, SUBREG, MEM or constant.
6035 2018-03-22 Tom de Vries <tom@codesourcery.com>
6037 PR tree-optimization/84956
6038 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6039 bb_has_abnormal_pred.
6041 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6044 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6045 DECL_INITIAL (decl) to decl at the end.
6046 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6049 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6051 * doc/extend.texi (__builtin_tgmath): Document when complex
6052 integer types are treated as _Complex _Float64.
6054 2018-03-21 Tom de Vries <tom@codesourcery.com>
6056 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6058 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6060 PR tree-optimization/84960
6061 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6062 if it is ENTRY block, move them into single succ of ENTRY in that case.
6064 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6066 PR tree-optimization/84811
6067 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6068 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6069 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6070 (known_subrange_p): Update accordingly. Cast each value involved
6071 in the size comparison, rather than casting the result of the
6074 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6076 PR tree-optimization/84982
6077 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6078 by flipping the least significant bit rather than all bits from
6079 bitpos to bitpos + bitsize - 1.
6081 2018-03-21 Nathan Sidwell <nathan@acm.org>
6083 * doc/extend.texi (Deprecated Features): Remove mention of
6084 long-deleted deprecations.
6086 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6089 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6090 * configure: Regenerate.
6092 2018-03-21 Tom de Vries <tom@codesourcery.com>
6094 PR tree-optimization/83126
6095 * tree-parloops.c (num_phis): New function.
6096 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6098 2018-03-21 Nathan Sidwell <nathan@acm.org>
6100 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6101 mention anon-struct/union members and trailing attributes.
6103 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6105 PR tree-optimization/84969
6106 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6107 builtin memset partitions if they set different rhs values.
6109 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6111 PR rtl-optimization/84989
6112 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6113 VEC_DUPLICATE with scalar result mode.
6115 2018-03-21 Martin Liska <mliska@suse.cz>
6118 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6119 not intended return statement.
6121 2018-03-21 Martin Liska <mliska@suse.cz>
6124 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6125 (chkp_find_bound_slots_1): Limit number of iterations.
6127 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6130 * Minor grammar fixes for x86 options.
6132 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6135 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6136 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6138 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6141 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6142 (altivec_lvx_<mode>_1op): Likewise.
6143 (altivec_stvx_<mode>_2op): Likewise.
6144 (altivec_stvx_<mode>_1op): Likewise.
6145 (altivec_lvx_<VM2:mode>): New define_expand.
6146 (altivec_stvx_<VM2:mode>): Likewise.
6147 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6148 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6149 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6150 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6151 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6152 (rs6000_gen_lvx): Likewise.
6153 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6154 (altivec_expand_stv_builtin): Likewise.
6155 (altivec_expand_builtin): Likewise.
6156 * config/rs6000/vector.md: Likewise.
6158 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6161 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6164 2018-03-20 Richard Biener <rguenther@suse.de>
6167 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6168 sign-conversions as zero, fall back to standard scalar_stmt
6171 2018-03-20 Martin Liska <mliska@suse.cz>
6174 * predict.c (rebuild_frequencies): Handle case when we have
6175 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6177 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6180 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6181 flag_section_anchors.
6182 * varasm.c (use_blocks_for_decl_p): Remove hack for
6183 dw2_force_const_mem.
6186 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6188 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6189 be created, use lowpart_subreg of operands[0] rather than operands[0]
6191 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6192 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6193 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6194 and n constraint instead of aarch64_shift_imm_di and Usd.
6195 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6196 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6198 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6201 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6202 to favor GPR over NEON registers.
6203 (<shift>di3_neon): Likewise.
6205 2018-03-20 Tom de Vries <tom@codesourcery.com>
6208 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6209 (nvptx_process_pars): Emit bar.sync asap and alap.
6211 2018-03-20 Tom de Vries <tom@codesourcery.com>
6214 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6215 seen_label if seen_label is already set.
6217 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6220 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6221 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6222 Use 1U instead of 1. Formatting fixes.
6225 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6226 instead of TREE_TYPE (s1) for the return value.
6228 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6230 PR tree-optimization/84946
6231 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6232 bitsize + bitsize in poly_uint64 rather than poly_int64.
6235 * dwarf2asm.c: Include fold-const.c.
6236 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6237 of decl rather than decl itself.
6239 PR rtl-optimization/84643
6240 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6242 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6245 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6246 calling assemble_variable.
6248 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6251 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6252 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6254 2018-03-19 Jim Wilson <jimw@sifive.com>
6257 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6258 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6259 (riscv_first_stack_step): Likewise.
6260 (riscv_option_override): Use STACK_BOUNDARY instead of
6262 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6264 (BIGGEST_ALIGNMENT): Set to 128.
6265 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6266 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6269 2018-03-19 Richard Biener <rguenther@suse.de>
6271 PR tree-optimization/84933
6272 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6273 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6275 2018-03-19 Richard Biener <rguenther@suse.de>
6277 PR tree-optimization/84859
6278 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6279 (cond_if_else_store_replacement): Perform sinking operation on
6280 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6281 Generalize what a BB with a single eligible store is.
6283 2018-03-19 Richard Biener <rguenther@suse.de>
6285 PR tree-optimization/84929
6286 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6287 chrec_is_positive against non-chrec arg.
6289 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6292 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6294 2018-03-18 Martin Liska <mliska@suse.cz>
6296 PR rtl-optimization/84635
6297 * regrename.c (build_def_use): Use matches_mode only when
6300 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6302 PR tree-optimization/84913
6303 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6304 vectorize chains of COND_EXPRs.
6306 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6308 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6310 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6312 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6314 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6316 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6318 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6319 Kito Cheng <kito.cheng@gmail.com>
6321 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6322 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6323 (nds32_adjust_reg_alloc_order): New function.
6324 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6326 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6328 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6329 nds32_print_operand, nds32_print_operand_address): Use
6330 HOST_WIDE_INT_PRINT_DEC instead.
6332 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6334 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6336 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6339 * config/i386/i386.c (initial_ix86_tune_features,
6340 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6342 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6343 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6344 rather than 1u << ix86_tune. Formatting fix.
6345 (ix86_option_override_internal): Change ix86_arch_mask from
6346 unsigned int to unsigned HOST_WIDE_INT, initialize to
6347 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6348 (ix86_function_specific_restore): Likewise.
6350 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6353 * postreload.c (reload_combine_recognize_pattern): Perform
6354 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6355 truncate_int_for_mode the result for the destination's mode.
6358 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6359 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6361 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6364 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6367 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6368 regno_allocno_class_array and sorted_pseudos.
6369 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6370 insns where regno is used.
6372 2018-03-16 Martin Liska <mliska@suse.cz>
6375 * multiple_target.c (create_dispatcher_calls): Redirect
6376 reference in the symbol table.
6378 2018-03-16 Martin Liska <mliska@suse.cz>
6381 * multiple_target.c (create_dispatcher_calls): Redirect also
6384 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6388 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6389 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6390 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6392 2018-03-16 Julia Koval <julia.koval@intel.com>
6394 * doc/invoke.texi (Skylake Server): Add CLWB.
6395 Cannonlake): Remove CLWB.
6397 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6399 PR tree-optimization/84841
6400 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6402 (FLOAT_ONE_CONST_TYPE): Define.
6403 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6404 (sort_by_operand_rank): Put entries with higher constant_type last
6405 rather than first to match comments.
6407 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6409 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6412 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6415 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6416 check and use address of target always.
6418 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6421 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6422 (indirect_thunk_bnd_needed): Likewise.
6423 (indirect_thunks_used): Likewise.
6424 (indirect_thunks_bnd_used): Likewise.
6425 (indirect_return_needed): New.
6426 (indirect_return_bnd_needed): Likewise.
6427 (output_indirect_thunk_function): Add a bool argument for
6429 (output_indirect_thunk_function): Don't generate alias for
6430 function return thunk.
6431 (ix86_code_end): Call output_indirect_thunk_function to generate
6432 function return thunks.
6433 (ix86_output_function_return): Set indirect_return_bnd_needed
6434 and indirect_return_needed instead of indirect_thunk_bnd_needed
6435 and indirect_thunk_needed.
6437 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6439 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6440 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6441 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6443 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6444 Paul Hua <paul.hua.gm@gmail.com>
6447 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6449 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6451 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6452 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6455 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6458 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6459 instead of GET_MODE_SIZE when comparing Units.
6461 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6464 * varasm.c (hash_section): Return an unchangeble hash value
6465 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6466 Return !aarch64_can_use_per_function_literal_pools_p ().
6468 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6471 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6472 rather than address of cmode as last argument to prepare_cmp_insn.
6474 2018-03-15 Julia Koval <julia.koval@intel.com>
6476 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6477 F_AVX512VNNI, F_AVX512BITALG): New.
6479 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6482 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6483 insn for floating-point loads and stores.
6485 2018-03-14 Carl Love <cel@us.ibm.com>
6487 * config/rs6000/rs6000-c.c: Add macro definitions for
6488 ALTIVEC_BUILTIN_VEC_PERMXOR.
6489 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6490 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6491 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6492 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6494 * config/doc/extend.texi: Add prototypes for vec_permxor.
6496 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6499 * diagnostic-show-locus.c (class layout_point): Convert m_line
6500 from int to linenum_type.
6501 (line_span::comparator): Use linenum "compare" function when
6502 comparing line numbers.
6503 (test_line_span): New function.
6504 (layout_range::contains_point): Convert param "row" from int to
6506 (layout_range::intersects_line_p): Likewise.
6507 (layout::will_show_line_p): Likewise.
6508 (layout::print_source_line): Likewise.
6509 (layout::should_print_annotation_line_p): Likewise.
6510 (layout::print_annotation_line): Likewise.
6511 (layout::print_leading_fixits): Likewise.
6512 (layout::annotation_line_showed_range_p): Likewise.
6513 (struct line_corrections): Likewise for field m_row.
6514 (line_corrections::line_corrections): Likewise for param "row".
6515 (layout::print_trailing_fixits): Likewise.
6516 (layout::get_state_at_point): Likewise.
6517 (layout::get_x_bound_for_row): Likewise.
6518 (layout::print_line): Likewise.
6519 (diagnostic_show_locus): Likewise for locals "last_line" and
6521 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6522 * input.c (selftest::test_linenum_comparisons): New function.
6523 (selftest::input_c_tests): Call it.
6524 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6525 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6526 * selftest.h (ASSERT_GT): New macro.
6527 (ASSERT_GT_AT): New macro.
6528 (ASSERT_LT): New macro.
6529 (ASSERT_LT_AT): New macro.
6531 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6533 PR rtl-optimization/84780
6534 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6536 2018-03-14 Martin Liska <mliska@suse.cz>
6538 * tree.c (record_node_allocation_statistics): Use
6539 get_stats_node_kind.
6540 (get_stats_node_kind): New function extracted from
6541 record_node_allocation_statistics.
6542 (free_node): Use get_stats_node_kind.
6544 2018-03-14 Richard Biener <rguenther@suse.de>
6546 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6547 that the value-set of ANTIC_IN doesn't grow.
6550 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6552 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6553 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6555 2018-03-14 Julia Koval <julia.koval@intel.com>
6557 * config.gcc (icelake-client, icelake-server): New.
6559 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6560 (initial_ix86_arch_features): Ditto.
6561 (PTA_SKYLAKE): Add SGX.
6562 (PTA_ICELAKE): Remove.
6563 (PTA_ICELAKE_CLIENT): New.
6564 (PTA_ICELAKE_SERVER): New.
6565 (ix86_option_override_internal): Split up icelake on icelake client and
6567 (get_builtin_code_for_version): Ditto.
6568 (fold_builtin_cpu): Ditto.
6569 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6570 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6571 * config/i386/i386.h (processor_type): Ditto.
6572 * doc/invoke.texi: Ditto.
6574 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6577 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6578 INTEGER_CST offset, add it together with bitpos / 8 and
6579 sign extend based on POINTER_SIZE.
6583 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6586 * config/i386/constraints.md (Yc): New register constraint.
6587 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6588 Use Yc constraint for alternative 2 of operand 0. Remove
6589 preferred_for_speed attribute.
6591 2018-03-14 Richard Biener <rguenther@suse.de>
6593 PR tree-optimization/84830
6594 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6595 with the old one to avoid oscillations.
6597 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6600 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6602 (assign_by_spills): Return a flag of reload assignment failure.
6603 Do not process the reload assignment failures. Do not spill other
6604 reload pseudos if they has the same reg class. Update n if
6606 (lra_assign): Add a return arg. Set up from the result of
6607 assign_by_spills call.
6608 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6609 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6610 usage_insns if it is not NULL.
6611 (spill_hard_reg_in_range): New function.
6612 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6613 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6614 function prototypes.
6615 (lra_assign): Change prototype.
6616 * lra.c (lra): Add code to deal with fails by splitting hard reg
6619 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6621 * config/riscv/riscv.opt (mrelax): New option.
6622 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6623 "norelax" when riscv_mrelax is disabled.
6624 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6626 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6629 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6630 reassociation for int modes.
6632 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6634 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6635 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6637 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6638 * config/aarch64/aarch64-sve.md
6639 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6640 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6641 (*extend<mode><Vwide>2): Rename to...
6642 (aarch64_sve_extend<mode><Vwide>2): ...this.
6643 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6644 renaming the old pattern to...
6645 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6647 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6648 define_expand, renaming the old pattern to...
6649 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6650 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6651 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6652 account when deciding which SVE instruction the optab should use.
6653 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6655 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6657 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6658 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6659 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6660 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6661 (tlsdesc_small_<mode>): Turn a define_expand and use
6662 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6663 (tlsdesc_small_advsimd_<mode>): ...this.
6664 (tlsdesc_small_sve_<mode>): New pattern.
6666 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6668 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6669 (UNSPEC_UMUL_HIGHPART): New constants.
6670 (MUL_HIGHPART): New int iteraor.
6671 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6672 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6674 (*<su>mul<mode>3_highpart): New define_insn.
6676 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6679 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6682 2018-03-13 Martin Liska <mliska@suse.cz>
6685 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6687 (sem_item_optimizer::~sem_item_optimizer): Release it.
6688 (sem_item_optimizer::merge_classes): Register variable aliases.
6689 (sem_item_optimizer::fixup_pt_set): New function.
6690 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6691 * ipa-icf.h: Declare new variables and functions.
6693 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6696 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6697 integer_pow2p@2 and test integer_pow2p in condition.
6698 (A < 0 ? C : 0): Similarly for @1.
6701 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6702 characters starting at p contain '\0' character, don't look beyond
6706 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6707 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6710 * reg-stack.c (change_stack): Change update_end var from int to
6711 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6712 also call set_block_for_insn on the newly added insns and rescan.
6715 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6716 on the last operand.
6719 * tree.c (stabilize_reference_1): Return save_expr (e) for
6720 STATEMENT_LIST even if it doesn't have side-effects.
6722 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6724 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6726 2018-03-12 Renlin Li <renlin.li@arm.com>
6728 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6729 aarch64_output_scalar_simd_mov_immediate.
6731 2018-03-12 Martin Sebor <msebor@redhat.com>
6733 PR tree-optimization/83456
6734 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6735 for perfectly overlapping calls to memcpy.
6736 (gimple_fold_builtin_memory_chk): Same.
6737 (gimple_fold_builtin_strcpy): Handle no-warning.
6738 (gimple_fold_builtin_stxcpy_chk): Same.
6739 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6741 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6743 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6744 parameter. Use it for SFmode.
6745 (rs6000_function_arg_advance_1): Adjust.
6746 (rs6000_function_arg): Adjust.
6747 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6749 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6751 PR rtl-optimization/84169
6752 PR rtl-optimization/84780
6753 * combine.c (can_combine_p): Check for a 2-insn combination whether
6754 the destination register is used between the two insns, too.
6756 2018-03-12 Richard Biener <rguenther@suse.de>
6758 PR tree-optimization/84803
6759 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6760 for refs DR analysis didn't process.
6762 2018-03-12 Richard Biener <rguenther@suse.de>
6764 PR tree-optimization/84777
6765 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6766 force-vectorize loops ignore whether we are optimizing for size.
6768 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6770 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6771 (TARGET_MD_ASM_ADJUST): Define.
6773 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6774 Kito Cheng <kito.cheng@gmail.com>
6775 Chung-Ju Wu <jasonwucj@gmail.com>
6777 * config/nds32/nds32.c (nds32_compute_stack_frame,
6778 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6779 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6780 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6781 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6782 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6783 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6784 * config/nds32/nds32.md (prologue, epilogue): Use macro
6785 NDS32_V3PUSH_AVAILABLE_P to do checking.
6787 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6790 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6791 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6792 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6793 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6794 addition of most attributes on !orig_type_die or the attribute not
6795 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6797 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6798 Chung-Ju Wu <jasonwucj@gmail.com>
6800 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6802 * config/nds32/nds32.opt (mvh): New option.
6804 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6805 Chung-Ju Wu <jasonwucj@gmail.com>
6807 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6809 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6810 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6813 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6814 Chung-Ju Wu <jasonwucj@gmail.com>
6816 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6818 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6819 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6821 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6822 Kito Cheng <kito.cheng@gmail.com>
6823 Chung-Ju Wu <jasonwucj@gmail.com>
6825 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6826 UNSPEC_FFMISM and UNSPEC_FLMISM.
6827 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6828 for ffb, ffmism and flmism.
6829 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6830 (unspec_ffmism): Ditto.
6831 (unspec_flmism): Ditto.
6832 (nds32_expand_builtin_impl): Check if string extension is available.
6833 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6834 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6836 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6839 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6842 * lra-assigns.c (assign_by_spills): Return a flag of reload
6843 assignment failure. Do not process the reload assignment
6844 failures. Do not spill other reload pseudos if they has the same
6846 (lra_assign): Add a return arg. Set up from the result of
6847 assign_by_spills call.
6848 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6849 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6850 usage_insns if it is not NULL.
6851 (spill_hard_reg_in_range): New function.
6852 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6853 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6854 function prototypes.
6855 (lra_assign): Change prototype.
6856 * lra.c (lra): Add code to deal with fails by splitting hard reg
6859 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6862 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6864 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6867 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6868 (dw_val_node): Add val_symbolic_view.
6869 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6870 (symview_upper_bound): New.
6871 (new_line_info_table): Initialize symviews_since_reset.
6872 (dwarf2out_source_line): Count symviews_since_reset and set
6873 symview_upper_bound.
6874 (dw_val_equal_p): Handle symview.
6875 (add_AT_symview): New.
6876 (print_dw_val): Handle symview.
6877 (attr_checksum, attr_checksum_ordered): Likewise.
6878 (same_dw_val_p, size_of_die): Likewise.
6879 (value_format, output_die): Likewise.
6880 (add_high_low_attributes): Use add_AT_symview for entry_view.
6881 (dwarf2out_finish): Reset symview_upper_bound, clear
6884 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6887 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6888 Add strict argument and use it.
6889 (rs6000_split_multireg_move): Update for new strict argument.
6890 (mem_operand_gpr): Disallow all non-offsettable addresses.
6891 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6893 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6896 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6897 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6898 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6901 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6902 decl, use remap_type if we want to use the type.
6904 2018-03-09 Martin Sebor <msebor@redhat.com>
6906 PR tree-optimization/84526
6907 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6909 (builtin_access::generic_overlap): Be prepared to handle non-array
6912 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6914 PR rtl-optimization/84682
6915 * lra-constraints.c (process_address_1): Check is_address flag
6916 for address constraints.
6917 (process_alt_operands): Likewise.
6918 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6919 preprocess_constraints.
6920 * recog.h (preprocess_constraints): Add oploc parameter.
6922 * recog.c (preprocess_constraints): Test address_operand for
6923 CT_ADDRESS constraints.
6925 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6928 * lra-assigns.c (assign_by_spills): Return a flag of reload
6929 assignment failure. Do not process the reload assignment
6930 failures. Do not spill other reload pseudos if they has the same
6932 (lra_assign): Add a return arg. Set up from the result of
6933 assign_by_spills call.
6934 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6935 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6936 usage_insns if it is not NULL.
6937 (spill_hard_reg_in_range): New function.
6938 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6939 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6940 function prototypes.
6941 (lra_assign): Change prototype.
6942 * lra.c (lra): Add code to deal with fails by splitting hard reg
6945 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6948 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6949 Accept complain bool parameter. Only emit errors if it is true.
6950 (arm_parse_cpu_option_name): Likewise.
6951 (arm_target_thumb_only): Adjust callers of the above.
6952 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6953 prototype to take a default true bool parameter.
6954 (arm_parse_arch_option_name): Likewise.
6956 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6957 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6961 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6962 * configure: Regenerate.
6963 * configure.ac ("linker --version-script option"): New.
6964 ("linker soname option"): New.
6966 2018-03-09 Richard Biener <rguenther@suse.de>
6968 PR tree-optimization/84775
6969 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6970 immediate uses of predicate stmts and mark them modified.
6973 PR tree-optimization/84178
6974 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6976 (version_loop_for_if_conversion): Delay update_ssa call.
6977 (tree_if_conversion): Delay update_ssa until after predicate
6980 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6983 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6984 when the function accesses prior frames.
6986 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6989 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6990 gen_llsym, otherwise call maybe_gen_llsym.
6993 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6994 has ',' character inside of it.
6996 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6999 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
7000 as clobbering CC_REGNUM.
7002 2018-03-08 Richard Biener <rguenther@suse.de>
7005 * tree-scalar-evolution.c: Include tree-into-ssa.h.
7006 (follow_copies_to_constant): Do not follow SSA names registered
7009 2018-03-08 Richard Biener <rguenther@suse.de>
7011 PR tree-optimization/84178
7012 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
7014 (version_loop_for_if_conversion): Delay update_ssa call.
7015 (tree_if_conversion): Delay update_ssa until after predicate
7018 2018-03-08 David Malcolm <dmalcolm@redhat.com>
7020 PR tree-optimization/84178
7021 * tree-if-conv.c (release_bb_predicate): Remove the
7022 the assertion that the stmts have NULL use_ops.
7023 Discard the statements, asserting that they haven't
7024 yet been added to a BB.
7026 2018-03-08 Richard Biener <rguenther@suse.de>
7028 PR tree-optimization/84746
7029 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
7030 (phi_translate): Pass in destination ANTIC_OUT set.
7031 (phi_translate_1): Likewise. For a simplified result lookup
7032 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
7033 (phi_translate_set): Adjust.
7034 (do_pre_regular_insertion): Likewise.
7035 (do_pre_partial_partial_insertion): Likewise.
7037 2018-03-08 Martin Liska <mliska@suse.cz>
7039 PR gcov-profile/84735
7040 * doc/gcov.texi: Document usage of profile files.
7041 * gcov-io.h: Document changes in the format.
7043 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7047 * dwarf2out.c (struct dw_line_info_table): Update comments for
7049 (FORCE_RESET_NEXT_VIEW): New.
7050 (FORCE_RESETTING_VIEW_P): New.
7051 (RESETTING_VIEW_P): Check for -1 too.
7052 (ZERO_VIEW_P): Likewise.
7053 (new_line_info_table): Force-reset next view.
7054 (dwarf2out_begin_function): Likewise.
7055 (dwarf2out_source_line): Simplify zero_view_p initialization.
7056 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7057 view directly. Omit view when omitting .loc at line 0.
7059 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7061 PR tree-optimization/84740
7062 * tree-switch-conversion.c (process_switch): Call build_constructors
7063 only if info.phi_count is non-zero.
7065 PR tree-optimization/84739
7066 * tree-tailcall.c (find_tail_calls): Check call arguments against
7067 DECL_ARGUMENTS (current_function_decl) rather than
7068 DECL_ARGUMENTS (func) when checking for tail recursion.
7070 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7072 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7073 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7074 Volker Reichelt's entry and add entries for people that perform
7075 GCC fuzzy testing and report numerous bugs.
7077 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7080 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7081 readonly data in sdata, if that is disabled.
7082 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7083 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7084 -mreadonly-in-sdata option.
7086 2018-03-07 Martin Sebor <msebor@redhat.com>
7088 PR tree-optimization/84468
7089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7090 basic block when looking for nul assignment.
7092 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7095 * except.h (output_function_exception_table): Adjust prototype.
7096 * except.c (output_function_exception_table): Remove FNNAME parameter
7097 and add SECTION parameter. Ouput one part of the table at a time.
7098 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7099 the first part of the exception table and emit unwind directives.
7100 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7101 (i386_pe_seh_cold_init): Likewise.
7102 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7103 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7104 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7105 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7106 * config/i386/winnt.c: Include except.h.
7107 (struct seh_frame_state): Add reg_offset, after_prologue and
7108 in_cold_section fields.
7109 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7110 (i386_pe_seh_cold_init): New function.
7111 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7112 to seh->in_cold_section.
7113 (seh_emit_push): Record the offset of the push.
7114 (seh_emit_save): Record the offet of the save.
7115 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7116 Test seh->after_prologue to disregard the epilogue.
7117 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7118 (i386_pe_end_cold_function): New function.
7120 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7123 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7124 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7127 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7129 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7130 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7133 * multiple_target.c: Include tree-inline.h and intl.h.
7134 (expand_target_clones): Diagnose and fail if node->definition and
7135 !tree_versionable_function_p (node->decl).
7137 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7139 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7141 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7142 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7143 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7145 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7148 * combine.c (try_combine): Use reg_or_subregno instead of handling
7149 just paradoxical SUBREGs and REGs.
7151 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7153 * config/arc/arc.c (arc_finalize_pic): Remove function.
7154 (arc_must_save_register): We use single base PIC register, remove
7155 checks to save/restore the PIC register.
7156 (arc_expand_prologue): Likewise.
7157 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7159 (arc_verify_short): Likewise.
7160 (arc_attr_type): Likewise.
7161 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7162 (walk_stores): Likewise.
7163 (arc_address_cost): Make it static.
7164 (arc_verify_short): Likewise.
7165 (branch_dest): Likewise.
7166 (arc_attr_type): Likewise.
7167 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7168 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7169 (arc_final_prescan_insn): Remove inserting the nops due to
7170 hardware hazards. It is done in reorg step.
7171 (insn_length_variant_t): Remove.
7172 (insn_length_parameters_t): Likewise.
7173 (arc_insn_length_parameters): Likewise.
7174 (arc_get_insn_variants): Likewise.
7175 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7177 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7180 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7183 PR tree-optimization/84687
7184 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7186 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7188 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7190 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7191 Rename to ppc_speculation_barrier.
7192 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7193 __builtin_ppc_speculation_barrier.
7195 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7198 * combine.c (combine_simplify_rtx): Don't try to simplify if
7199 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7202 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7204 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7205 to 32 bytes when compiling for POWER9.
7207 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7210 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7211 regparm >= 3 with no arg reg available also for calls with
7212 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7215 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7217 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7219 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7222 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7224 2018-03-05 Richard Biener <rguenther@suse.de>
7226 PR tree-optimization/84486
7227 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7228 When inserting a __builtin_assume_aligned call set the LHS
7229 SSA name alignment info accordingly.
7231 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7233 PR tree-optimization/84114
7234 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7235 Avoid reassociation of FLOAT_MODE addition.
7237 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7239 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7240 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7241 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7242 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7243 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7244 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7245 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7247 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7248 __builtin_ia32_wbinvd): New builtins.
7249 (SPECIAL_ARGS2): New.
7250 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7251 (SPECIAL_ARGS2): New.
7252 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7253 (ix86_valid_target_attribute_inner_p): Ditto.
7254 (ix86_init_mmx_sse_builtins): Add special_args2.
7255 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7256 TARGET_WBNOINVD_P): New.
7257 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7258 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7259 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7260 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7261 * config/i386/pconfigintrin.h: New file.
7262 * config/i386/wbnoinvdintrin.h: Ditto.
7263 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7264 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7266 2018-03-05 Richard Biener <rguenther@suse.de>
7268 PR tree-optimization/84670
7269 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7271 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7272 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7273 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7274 doesn't grow if all successors (recursively) were visited at least
7277 2018-03-05 Richard Biener <rguenther@suse.de>
7279 PR tree-optimization/84650
7280 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7281 if executed in the loop pipeline.
7283 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7285 * doc/configfiles.texi (Configuration Files): Move info about
7286 conditionalizing $target-protos.h to...
7287 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7288 differs from $target-protos.h.
7290 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7291 Chung-Ju Wu <jasonwucj@gmail.com>
7293 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7294 * config/nds32/nds32-multiple.md (setmemsi): Define.
7295 * config/nds32/nds32-memory-manipulation.c
7296 (nds32_gen_dup_4_byte_to_word_value): New.
7297 (emit_setmem_word_loop): New.
7298 (emit_setmem_byte_loop): New.
7299 (nds32_expand_setmem_loop): New.
7300 (nds32_expand_setmem_loop_v3m): New.
7301 (nds32_expand_setmem_unroll): New.
7302 (nds32_expand_setmem): New.
7304 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7305 Chung-Ju Wu <jasonwucj@gmail.com>
7307 * config/nds32/nds32-memory-manipulation.c
7308 (nds32_emit_load_store): New.
7309 (nds32_emit_post_inc_load_store): New.
7310 (nds32_emit_mem_move): New.
7311 (nds32_emit_mem_move_block): New.
7312 (nds32_expand_movmemsi_loop_unknown_size): New.
7313 (nds32_expand_movmemsi_loop_known_size): New.
7314 (nds32_expand_movmemsi_loop): New.
7315 (nds32_expand_movmemsi_unroll): New.
7316 (nds32_expand_movmemqi): Rename ...
7317 (nds32_expand_movmemsi): ... to this.
7318 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7319 (movmemsi): ... to this.
7320 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7321 (nds32_expand_movmemsi): ... to this.
7323 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7324 Monk Chiang <sh.chiang04@gmail.com>
7325 Chung-Ju Wu <jasonwucj@gmail.com>
7327 * config/nds32/nds32-protos.h
7328 (nds32_expand_load_multiple): New arguments.
7329 (nds32_expand_store_multiple): Ditto.
7330 (nds32_valid_multiple_load_store): Rename ...
7331 (nds32_valid_multiple_load_store_p): ... to this.
7332 * config/nds32/nds32-memory-manipulation.c
7333 (nds32_expand_load_multiple): Refine implementation.
7334 (nds32_expand_store_multiple): Ditto.
7335 * config/nds32/nds32-multiple.md
7336 (load_multiple): Update nds32_expand_load_multiple interface.
7337 (store_multiple): Update nds32_expand_store_multiple interface.
7338 * config/nds32/nds32-predicates.c
7339 (nds32_valid_multiple_load_store): Rename ...
7340 (nds32_valid_multiple_load_store_p): ... to this and refine
7342 * config/nds32/predicates.md
7343 (nds32_load_multiple_and_update_address_operation): New predicate.
7344 (nds32_store_multiple_and_update_address_operation): New predicate.
7346 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7347 Chung-Ju Wu <jasonwucj@gmail.com>
7349 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7350 (combo): New attribute.
7351 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7353 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7355 * config/nds32/nds32.opt: Change -mcmodel= default value.
7357 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7358 Monk Chiang <sh.chiang04@gmail.com>
7359 Chung-Ju Wu <jasonwucj@gmail.com>
7361 * config/nds32/constants.md (unspec_element): New enum.
7362 * config/nds32/constraints.md (Umw): New constraint.
7363 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7364 * config/nds32/nds32-intrinsic.md: Likewise.
7365 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7366 (nds32_valid_smw_lwm_base_p): New.
7367 (nds32_output_smw_single_word): New.
7368 (nds32_output_lmw_single_word): New.
7369 (nds32_expand_unaligned_load): New.
7370 (nds32_expand_unaligned_store): New.
7371 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7372 (nds32_output_smw_single_word): Declare.
7373 (nds32_output_lmw_single_word): Declare.
7374 (nds32_expand_unaligned_load): Declare.
7375 (nds32_expand_unaligned_store): Declare.
7376 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7377 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7378 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7379 NDS32_BUILTIN_UASTORE_DW.
7380 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7383 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7384 Kito Cheng <kito.cheng@gmail.com>
7385 Chung-Ju Wu <jasonwucj@gmail.com>
7387 * config/nds32/nds32-intrinsic.c
7388 (nds32_expand_builtin_null_ftype_reg): Delete.
7389 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7390 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7391 (nds32_read_argument): New.
7392 (nds32_legitimize_target): Ditto.
7393 (nds32_legitimize_argument): Ditto.
7394 (nds32_check_constant_argument): Ditto.
7395 (nds32_expand_unop_builtin): Ditto.
7396 (nds32_expand_unopimm_builtin): Ditto.
7397 (nds32_expand_binop_builtin): Ditto.
7398 (nds32_builtin_decl_impl): Ditto.
7399 (builtin_description): Ditto.
7400 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7401 (nds32_init_builtins_impl): Ditto.
7402 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7403 (nds32_builtin_decl): New.
7404 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7405 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7407 2018-03-02 Jeff Law <law@redhat.com>
7409 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7410 (redundant_insn, fill_simple_delay_slots): Likewise.
7411 (fill_slots_from_thread): Likewise.
7412 * resource.c (mark_referenced_resources): Likewise.
7413 (mark_set_resources, find_dead_or_set_registers): Likewise.
7415 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7417 * substring-locations.h (format_warning_va): Formatting fix for
7419 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7421 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7422 * substring-locations.c: Include intl.h.
7423 (format_warning_va): Turned into small wrapper around
7424 format_warning_n_va, renamed to ...
7425 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7426 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7428 (format_warning_at_substring_n): New function.
7429 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7430 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7431 format_warning_at_substring with just a shorter name instead of
7432 const function pointer.
7433 (fmtwarn_n): New function.
7434 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7435 appropriate, get rid of all the fmtstr temporaries, move conditionals
7436 with G_() wrapped string literals directly into fmtwarn arguments,
7437 cast dir.len to (int), formatting fixes.
7439 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7441 * doc/invoke.texi: Remove "Cilk Plus" references.
7443 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7444 Richard Biener <rguenther@suse.de>
7447 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7448 for error or warning attributes if CALL_FROM_THUNK_P is set.
7451 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7454 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7455 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7458 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7459 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7461 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7462 instead of -1U in last predictors element's probability member.
7464 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7467 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7468 arguments if they are comparable.
7470 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7472 PR tree-optimization/84634
7473 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7474 masks and masked_loop_p with a single loop_masks, making sure it's
7475 null for bb vectorization.
7477 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7479 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7480 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7481 loop->force_vectorize to check whether there is no alias.
7483 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7486 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7488 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7490 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7491 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7492 instead of a loop around prev_real_insn.
7493 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7497 * config/i386/i386.c (ix86_print_operand): Use conditional
7498 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7501 2018-03-02 Richard Biener <rguenther@suse.de>
7503 PR tree-optimization/84427
7504 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7505 (bitmap_set_subtract_values): Rewrite to handle multiple
7508 (prune_clobbered_mems): Likewise.
7509 (phi_translate): Take edge instead of pred/phiblock.
7510 (phi_translate_1): Likewise.
7511 (phi_translate_set): Likewise. Insert all translated
7512 exprs for a value into the set, keeping possibly multiple
7513 expressions per value.
7514 (compute_antic_aux): Adjust for phi_translate changes.
7515 When intersecting union the expressions and prune those
7516 not in the final value set, keeping possibly multiple
7517 expressions per value. Do not use value-insertion
7518 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7519 all expressions. Add verification that the value-sets
7520 only shrink during iteration.
7521 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7522 (do_pre_regular_insertion): Likewise.
7523 (do_pre_partial_partial_insertion): Likewise.
7525 2018-03-02 Richard Biener <rguenther@suse.de>
7528 * config/darwin.c (saved_debug_info_level): New static global.
7529 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7530 (darwin_asm_lto_end): Restore debug info generation settings.
7532 2018-03-01 Martin Liska <mliska@suse.cz>
7535 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7538 2018-03-01 Richard Biener <rguenther@suse.de>
7541 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7542 type annotation with LTO when debug was disabled at compile-time.
7544 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7546 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7548 (mips_final_postscan_insn): Likewise.
7550 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7552 PR rtl-optimization/84528
7553 * alias.c (init_alias_target): Add commentary.
7554 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7555 a unique base value if the frame pointer is not eliminated
7556 to the stack pointer.
7558 2018-03-01 Tom de Vries <tom@codesourcery.com>
7560 PR rtl-optimization/83327
7561 * lra-int.h (hard_regs_spilled_into): Declare.
7562 * lra.c (hard_regs_spilled_into): Define.
7563 (init_reg_info): Init hard_regs_spilled_into.
7564 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7565 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7566 (process_bb_lives): Handle hard_regs_spilled_into.
7567 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7568 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7570 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7572 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7573 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7574 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7575 * config/rs6000/aix72.h: New file.
7577 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7579 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7580 instead of warning_at with conditional singular and plural messages
7584 * stor-layout.c (update_alignment_for_field): For
7585 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7586 && !DECL_PACKED (field), do the alignment update, just use
7587 only desired_align instead of MAX (type_align, desired_align)
7589 (place_field): Don't do known_align < desired_align handling
7590 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7591 is non-NULL, instead do it after rli->prev_field handling and
7592 only if not within a bitfield word. For DECL_PACKED (field)
7593 use type_align of BITS_PER_UNIT.
7595 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7597 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7598 superfluous parentheses and trailing spaces.
7600 2018-02-28 Richard Biener <rguenther@suse.de>
7602 PR tree-optimization/84584
7603 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7604 SCoPs with fake exit edge.
7606 2018-02-28 Martin Liska <mliska@suse.cz>
7609 * timevar.c (timer::print): Fix format to properly print 100%
7612 2018-02-28 Richard Biener <rguenther@suse.de>
7615 * genmatch.c (capture_info::walk_match): Do not mark
7616 captured expressions without operands as expr_p given
7617 they act more like predicates and should be subject to
7618 "lost tail" side-effect preserving.
7620 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7622 PR rtl-optimization/81611
7623 * auto-inc-dec.c (attempt_change): Move dead note from
7624 mem_insn if it's the next use of regno
7625 (find_address): Take address use of reg holding
7626 non-incremented value. Add parm to limit search to the named
7628 (merge_in_block): Attempt to use a mem insn that is the next
7629 use of the original regno.
7631 2018-02-27 Martin Sebor <msebor@redhat.com>
7634 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7635 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7637 2018-02-27 Martin Sebor <msebor@redhat.com>
7639 PR translation/84207
7640 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7641 n argument to unsigned HOST_WIDE_INT.
7642 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7643 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7644 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7645 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7647 2018-02-27 Richard Biener <rguenther@suse.de>
7649 PR tree-optimization/84512
7650 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7651 Do not use the estimate returned from record_stmt_cost for
7652 the scalar iteration cost but sum properly using add_stmt_cost.
7654 2018-02-27 Richard Biener <rguenther@suse.de>
7656 PR tree-optimization/84466
7657 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7658 Adjust last change to less strictly validate use operands.
7660 2018-02-27 Martin Liska <mliska@suse.cz>
7662 PR gcov-profile/84548
7663 * gcov.c (process_file): Allow partial overlap and consider it
7664 also as group functions.
7665 (output_lines): Properly calculate range of lines for a group.
7667 2018-02-27 Martin Liska <mliska@suse.cz>
7669 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7670 'ggc' suffixes. Change first column width.
7671 (timer::print): Fix formatting of the column.
7673 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7675 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7676 preserve inline entry blocks for the sake of debug inline
7677 entry point markers alone.
7678 (remove_unused_locals): Suggest in comments a better place to
7679 force the preservation of inline entry blocks that are
7680 otherwise unused, but do not preserve them.
7682 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7684 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7686 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7689 * config/i386/constraints.md (Bs): Replace
7690 ix86_indirect_branch_register with
7691 TARGET_INDIRECT_BRANCH_REGISTER.
7693 * config/i386/i386.md (indirect_jump): Likewise.
7694 (tablejump): Likewise.
7695 (*sibcall_memory): Likewise.
7696 (*sibcall_value_memory): Likewise.
7697 Peepholes of indirect call and jump via memory: Likewise.
7698 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7699 (*sibcall_value_GOT_32): Likewise.
7700 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7701 (GOT_memory_operand): Likewise.
7702 (call_insn_operand): Likewise.
7703 (sibcall_insn_operand): Likewise.
7704 (GOT32_symbol_operand): Likewise.
7705 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7707 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7709 PR rtl-optimization/83496
7710 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7711 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7712 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7713 redundant insn, if any.
7714 (relax_delay_slots): Likewise.
7715 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7717 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7719 PR tree-optimization/83965
7720 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7721 that grouped statements are part of a reduction chain. Return
7722 true if the statement is not marked as a reduction itself but
7724 (vect_recog_dot_prod_pattern): Don't check whether the statement
7725 is part of a group here.
7726 (vect_recog_sad_pattern): Likewise.
7727 (vect_recog_widen_sum_pattern): Likewise.
7729 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7732 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7734 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7737 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7739 (ix86_output_indirect_function_return): New prototype.
7740 (ix86_split_simple_return_pop_internal): Likewise.
7741 * config/i386/i386.c (indirect_return_via_cx): New.
7742 (indirect_return_via_cx_bnd): Likewise.
7743 (indirect_thunk_name): Handle return va CX_REG.
7744 (output_indirect_thunk_function): Create alias for
7745 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7746 (ix86_output_indirect_jmp): Remove the bool argument.
7747 (ix86_output_indirect_function_return): New function.
7748 (ix86_split_simple_return_pop_internal): Likewise.
7749 * config/i386/i386.md (*indirect_jump): Don't pass false
7750 to ix86_output_indirect_jmp.
7751 (*tablejump_1): Likewise.
7752 (simple_return_pop_internal): Change it to define_insn_and_split.
7753 Call ix86_split_simple_return_pop_internal to split it for
7755 (simple_return_indirect_internal): Call
7756 ix86_output_indirect_function_return instead of
7757 ix86_output_indirect_jmp.
7759 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7762 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7763 memset and value initialization afterwards.
7765 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7767 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7769 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7772 * common/config/aarch64/aarch64-common.c
7773 (aarch_option_optimization_table[]): Switch
7774 off fomit-frame-pointer
7776 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7777 Chung-Ju Wu <jasonwucj@gmail.com>
7779 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7781 (store_multiple): Ditto.
7783 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7785 * config.gcc: Add --with-cpu support for nds32 target.
7786 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7787 * config/nds32/nds32.opt: Add -mcpu= option.
7789 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7791 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7792 isel=yes): Warn for these deprecated options.
7794 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7796 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7797 ISA_2_5_MASKS_EMBEDDED.
7799 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7801 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7802 p->max as pointers rather than using iterative_hash_expr.
7804 2018-02-23 Carl Love <cel@us.ibm.com>
7806 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7807 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7809 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7810 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7811 P8V_BUILTIN_VEC_VUNSIGNED2.
7813 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7816 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7817 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7819 * lra-constraints.c (curr_insn_transform): Set up
7820 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7822 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7823 * lra-eliminations.c (spill_pseudos): Ditto.
7824 (process_insn_for_elimination): Ditto.
7825 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7826 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7827 LRA_NON_CLOBBERED_ALT.
7829 2018-02-22 Martin Sebor <msebor@redhat.com>
7831 PR tree-optimization/84480
7832 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7833 to maybe_diag_stxncpy_trunc. Call it.
7834 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7835 from gimple_fold_builtin_strcpy. Print inlining stack.
7836 (handle_builtin_stxncpy): Print inlining stack.
7837 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7839 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7842 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7843 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7844 and -fcheck-pointer-bounds are used together.
7845 (indirect_thunk_prefix): New enum.
7846 (indirect_thunk_need_prefix): New function.
7847 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7848 "_nt" instead of "_bnd" for NOTRACK prefix.
7849 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7850 (output_indirect_thunk_function): Likewise.
7852 (ix86_code_end): Update output_indirect_thunk_function calls.
7853 (ix86_output_indirect_branch_via_reg): Replace
7854 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7855 (ix86_output_indirect_branch_via_push): Likewise.
7856 (ix86_output_function_return): Likewise.
7857 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7858 incompatible with -fcf-protection=branch and
7859 -fcheck-pointer-bounds.
7861 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7864 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7865 Change gcc_assert call to output_operand_lossage.
7867 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7869 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7871 2018-02-22 DJ Delorie <dj@redhat.com>
7872 Sebastian Perta <sebastian.perta@renesas.com>
7873 Oleg Endo <olegendo@gcc.gnu.org>
7875 * config/rx/rx.c (rx_rtx_costs): New function.
7876 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7878 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7880 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7882 2018-02-22 Martin Liska <mliska@suse.cz>
7885 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7886 Add "native" as a possible value.
7888 2018-02-22 Martin Liska <mliska@suse.cz>
7891 * config/i386/i386.c (ix86_option_override_internal):
7892 Add "native" as a possible value for -march and -mtune.
7894 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7897 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7898 to all type variants.
7900 PR tree-optimization/84503
7901 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7902 width as info->bitpos + info->bitsize - start.
7903 (merged_store_group::merge_overlapping): Simplify width computation.
7904 (check_no_overlap): New function.
7905 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7906 start + width and last_order of the group, fail if check_no_overlap
7908 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7909 to group if check_no_overlap fails.
7911 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7913 * config/rs6000/altivec.md: Delete contraint arguments to
7914 define_expand, define_split, and define_peephole2, and in
7915 define_insn_and_split if always unused.
7916 * config/rs6000/darwin.md: Ditto.
7917 * config/rs6000/dfp.md: Ditto.
7918 * config/rs6000/rs6000.md: Ditto.
7919 * config/rs6000/sync.md: Ditto.
7920 * config/rs6000/vector.md: Ditto.
7921 * config/rs6000/vsx.md: Ditto.
7923 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7925 * config/rs6000/altivec.md: Write output control strings as braced
7926 blocks instead of double-quoted strings.
7927 * config/rs6000/darwin.md: Ditto.
7928 * config/rs6000/rs6000.md: Ditto.
7929 * config/rs6000/vector.md: Ditto.
7930 * config/rs6000/vsx.md: Ditto.
7932 2018-02-21 Jason Merrill <jason@redhat.com>
7934 PR c++/84314 - ICE with templates and fastcall attribute.
7935 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7937 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7939 * ipa-cp.c (determine_versionability): Fix comment typos.
7941 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7944 * ipa-cp.c (determine_versionability): Do not version functions caling
7947 2018-02-21 Martin Liska <mliska@suse.cz>
7950 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7951 Add "native" as a possible value.
7952 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7953 the macro when native cpu detection is available.
7955 2018-02-21 Martin Liska <mliska@suse.cz>
7958 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7959 Add "native" as a possible value.
7960 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7961 when native cpu detection is available.
7963 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7964 Martin Sebor <msebor@redhat.com>
7966 PR tree-optimization/84478
7967 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7969 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7970 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7971 support which is conservatively correct, for 2 only stay conservative
7972 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7973 argument to the 2 argument get_range_strlen, adjust 6 arg
7974 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7976 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7977 (gimple_fold_builtin_strlen): Pass true as last argument to
7980 2018-02-20 Martin Sebor <msebor@redhat.com>
7983 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7984 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7985 (builtin_memref::builtin_memref): Factor out parts into
7986 set_base_and_offset and call it.
7988 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7991 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7992 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7993 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7994 search at the associated MODE_INT.
7996 2018-02-20 Jeff Law <law@redhat.com>
7999 PR tree-optimization/81592
8001 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
8002 for range data rather than using global data.
8003 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
8004 range data rather than using global data.
8005 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
8006 pass it to children as needed.
8007 (struct directive::fmtresult): Similarly.
8008 (struct directive::set_width): Similarly.
8009 (struct directive::set_precision): Similarly.
8010 (format_integer, format_directive, parse_directive): Similarly.
8011 (format_none): Accept unnamed vr_values parameter.
8012 (format_percent, format_floating, format_character): Similarly.
8013 (format_string, format_plain): Similarly.
8014 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
8015 the EVRP range analyzer for range data rather than using global data.
8016 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
8017 gimple-ssa-evrp-analyze.h
8018 (class sprintf_dom_walker): Add after_dom_children member function.
8019 Add evrp_range_analyzer member.
8020 (sprintf_dom_walker::before_dom_children): Call into the EVRP
8021 range analyzer as needed.
8022 (sprintf_dom_walker::after_dom_children): New member function.
8023 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
8025 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
8026 (evrp_range_analyzer::pop_to_marker): Likewise.
8028 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
8030 PR tree-optimization/84419
8031 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
8032 with the required type if its current type is compatible but
8035 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8038 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8039 after vectorization.
8041 2018-02-20 Martin Liska <mliska@suse.cz>
8044 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8045 possible values if we don't have a hint.
8047 2018-02-20 Martin Liska <mliska@suse.cz>
8051 * final.c (shorten_branches): Build align_tab array with one
8053 * opts.c (finish_options): Add alignment option limit check.
8054 (MAX_CODE_ALIGN): Likewise.
8055 (MAX_CODE_ALIGN_VALUE): Likewise.
8056 * doc/invoke.texi: Document maximum allowed option value for
8057 all -falign-* options.
8059 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8062 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8063 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8064 * var-tracking.c (emit_note_insn_var_location): Remove all references
8065 to NOTE_INSN_CALL_ARG_LOCATION.
8066 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8067 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8068 Use copy_rtx_if_shared.
8069 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8070 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8071 (dwarf2out_var_location): Remove handling of
8072 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8074 * final.c (final_scan_insn): Remove all references to
8075 NOTE_INSN_CALL_ARG_LOCATION.
8076 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8077 before dumping final insns.
8078 * except.c (emit_note_eh_region_end): Remove all references to
8079 NOTE_INSN_CALL_ARG_LOCATION.
8080 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8081 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8082 * config/arc/arc.c (hwloop_optimize): Likewise.
8083 * config/arm/arm.c (create_fix_barrier): Likewise.
8084 * config/s390/s390.c (s390_chunkify_start): Likewise.
8085 * config/sh/sh.c (find_barrier): Likewise.
8086 * config/i386/i386.c (rest_of_insert_endbranch,
8087 ix86_seh_fixup_eh_fallthru): Likewise.
8088 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8089 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8090 * config/frv/frv.c (frv_function_prologue): Likewise.
8091 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8093 (note_outside_basic_block_p): Remove all references to
8094 NOTE_INSN_CALL_ARG_LOCATION.
8095 * gengtype.c (adjust_field_rtx_def): Likewise.
8096 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8098 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8099 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8102 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8105 PR tree-optimization/84452
8106 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8107 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8110 2018-02-19 Martin Liska <mliska@suse.cz>
8113 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8115 2018-02-19 Martin Liska <mliska@suse.cz>
8116 Richard Sandiford <richard.sandiford@linaro.org>
8118 PR tree-optimization/82491
8119 * gimple-fold.c (get_base_constructor): Make earlier bail out
8122 2018-02-19 Carl Love <cel@us.ibm.com>
8124 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8125 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8127 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8128 P8V_BUILTIN_VEC_NEG.
8130 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8132 * config/rl78/rl78.md (movdf): New define expand.
8134 2018-02-19 Martin Liska <mliska@suse.cz>
8137 * doc/invoke.texi: Fix typo.
8138 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8140 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8142 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8143 handle rs6000_single_float and rs6000_double_float specially for
8146 2018-02-16 Jeff Law <law@redhat.com>
8148 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8151 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8154 * ipa-inline.c (inline_small_functions): Fix a typo.
8156 2018-02-16 Nathan Sidwell <nathan@acm.org>
8158 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8160 2018-02-16 Carl Love <cel@us.ibm.com>
8162 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8163 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8164 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8165 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8166 expansion to P8V_BUILTIN_VEC_FLOAT2.
8168 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8170 PR rtl-optimization/70023
8171 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8172 src_regno into account.
8174 2018-02-16 Carl Love <cel@us.ibm.com>
8176 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8177 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8178 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8179 * config/rs6000/rs6000.c: Remove case statements for
8180 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8181 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8182 and P9V_BUILTIN_VEC_VINSERT4B.
8183 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8184 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8185 * config/rs6000/vsx.md:
8186 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8189 2018-02-16 Carl Love <cel@us.ibm.com>
8191 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8193 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8195 * config/rs6000/rs6000-c.c: Add the definitions for
8196 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8197 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8198 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8199 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8200 definition for insert4b and define insn *insert3b_internal.
8201 * doc/extend.texi: Add documentation for vec_extract4b.
8203 2018-02-16 Nathan Sidwell <nathan@acm.org>
8205 * doc/extend.texi (Backwards Compatibility): Mention friend
8206 injection. Note for-scope is deprecated.
8207 * doc/invoke.texi (-ffriend-injection): Deprecate.
8209 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8211 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8212 that moved to I2, also allow destinations that are a paradoxical
8213 subreg (instead of a normal reg).
8215 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8218 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8221 2018-02-16 Richard Biener <rguenther@suse.de>
8223 PR tree-optimization/84037
8224 PR tree-optimization/84016
8226 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8227 Adjust vec_construct for the fact we need additional higher latency
8228 128bit inserts for AVX256 and AVX512 vector builds.
8229 (ix86_add_stmt_cost): Scale vector construction cost for
8232 2018-02-16 Richard Biener <rguenther@suse.de>
8234 PR tree-optimization/84417
8235 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8236 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8237 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8239 2018-02-16 Martin Liska <mliska@suse.cz>
8242 * internal-fn.def (ASAN_CHECK): Set proper flags.
8243 (ASAN_MARK): Likewise.
8245 2018-02-16 Julia Koval <julia.koval@intel.com>
8247 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8248 from PTA_CANNONLAKE.
8250 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8253 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8254 Use ++iter rather than iter++ for std::list iterators.
8255 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8256 defer deleting them until all nodes in the forest are processed. Do
8257 free even leaf nodes. Change to_process into auto_vec.
8260 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8261 * vec.h (vec_default_construct): Use memset instead of placement new
8262 if BROKEN_VALUE_INITIALIZATION is defined.
8263 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8264 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8267 PR rtl-optimization/83723
8268 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8269 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8270 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8272 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8274 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8276 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8278 PR rtl-optimization/81443
8279 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8280 from inner REGs to paradoxical SUBREGs.
8282 2018-02-16 Richard Biener <rguenther@suse.de>
8284 PR tree-optimization/84399
8285 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8286 For operands we can analyze at their definition make sure we can
8287 analyze them at each use as well.
8289 2018-02-16 Richard Biener <rguenther@suse.de>
8291 PR tree-optimization/84190
8292 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8293 volatile accesses if the decl isn't volatile.
8295 2018-02-15 Jason Merrill <jason@redhat.com>
8297 PR c++/84314 - ICE with templates and fastcall attribute.
8298 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8299 TYPE_CANONICAL on an existing type.
8301 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8303 PR tree-optimization/84383
8304 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8305 dstoff nor call operand_equal_p if dstbase is NULL.
8307 PR tree-optimization/84334
8308 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8309 also a CONSTANT_CLASS_P, punt.
8311 2018-02-14 Jim Wilson <jimw@sifive.com>
8313 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8314 first SMALL_OPERAND check. New local min_second_step. Move assert
8315 to where locals are set. Add TARGET_RVC support.
8316 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8318 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8320 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8322 2018-02-14 Martin Sebor <msebor@redhat.com>
8324 PR tree-optimization/83698
8325 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8326 arrays constrain the offset range to their bounds.
8327 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8328 (builtin_access::overlap): Avoid setting the size of overlap if it's
8330 (maybe_diag_overlap): Also consider arrays when deciding what values
8331 of offsets to include in diagnostics.
8333 2018-02-14 Martin Sebor <msebor@redhat.com>
8336 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8337 that correspond to the kind of a declaration.
8339 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8342 * config/pa/pa.md: Load address of PIC label using the linkage table
8343 if the label is nonlocal.
8345 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8347 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8348 warning message if user requests -maltivec=be.
8349 * doc/invoke.texi: Document deprecation of -maltivec=be.
8351 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8354 * config/rs6000/rs6000-c.c: Update definitions for
8355 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8356 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8358 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8361 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8362 add _get_ssp intrinsics. Remove argument from
8363 __builtin_ia32_rdssp[d|q].
8364 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8365 * config/i386/i386-builtin.def: Remove argument from
8366 __builtin_ia32_rdssp[d|q].
8367 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8368 ix86_expand_special_args_builtin for _rdssp[d|q].
8369 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8370 Clear register before usage.
8371 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8372 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8374 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8376 PR tree-optimization/84357
8377 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8378 operand 1 of an ARRAY_REF too.
8380 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8383 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8384 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8386 (set_of_reg): New struct.
8387 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8388 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8389 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8391 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8392 Split into bitclr, bitset, bitinvert patterns if appropriate.
8393 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8394 use rx_fuse_in_memory_bitop.
8395 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8396 to named insn, correct maximum insn length.
8398 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8401 * machmode.def: Define a complex mode for PARTIAL_INT.
8402 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8404 * doc/rtl.texi: Document CSPImode.
8405 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8407 (msp430_hard_regno_nregs_with_padding): Likewise.
8409 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8412 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8414 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8416 PR rtl-optimization/84169
8417 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8418 we generated a parallel as new i3 and we split that to new i2 and i3
8419 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8420 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8421 those to i2, not i1. Partially rewrite this scan code.
8423 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8426 * stor-layout.c (place_field): For variable length fields, adjust
8427 offset_align afterwards not just based on the field's alignment,
8428 but also on the size.
8431 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8432 of exps and logs in the use_exp2 case.
8434 2018-02-13 Jeff Law <law@redhat.com>
8436 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8439 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8442 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8446 * common.opt (gas-loc-support, gas-locview-support): New.
8447 (ginline-points, ginternal-reset-location-views): New.
8448 * doc/invoke.texi: Document them. Use @itemx where intended.
8449 (gvariable-location-views): Adjust.
8450 * target.def (reset_location_view): New.
8451 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8452 (TARGET_RESET_LOCATION_VIEW): New.
8453 * doc/tm.texi: Rebuilt.
8454 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8455 (dwarf2out_default_as_locview_support): New.
8456 (output_asm_line_debug_info): Use option variables.
8457 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8458 (output_loc_list): Likewise.
8459 (add_high_low_attributes): Check option variables.
8460 Don't output entry view attribute in strict mode.
8461 (gen_inlined_subroutine_die): Check option variables.
8462 (dwarf2out_inline_entry): Likewise.
8463 (init_sections_and_labels): Likewise.
8464 (dwarf2out_early_finish): Likewise.
8465 (maybe_reset_location_view): New, from...
8466 (dwarf2out_var_location): ... here. Call it.
8467 * debug.h (dwarf2out_default_as_loc_support): Declare.
8468 (dwarf2out_default_as_locview_support): Declare.
8469 * hooks.c (hook_int_rtx_insn_0): New.
8470 * hooks.h (hook_int_rtx_insn_0): Declare.
8471 * toplev.c (process_options): Take -gas-loc-support and
8472 -gas-locview-support from dwarf2out. Enable
8473 -gvariable-location-views by default only with locview
8474 assembler support. Enable -ginternal-reset-location-views by
8475 default only if the target defines the corresponding hook.
8476 Enable -ginline-points by default if location views are
8477 enabled; force it disabled if statement frontiers are
8479 * tree-inline.c (expand_call_inline): Check option variables.
8480 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8482 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8484 PR tree-optimization/84321
8485 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8486 handling. Also check whether the anti-range contains any values
8487 that satisfy the mask; switch to a VR_RANGE if not.
8489 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8492 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8494 2018-02-13 Martin Jambor <mjambor@suse.cz>
8497 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8498 of call statements, also set location of a load to a temporary.
8500 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8502 * config/rl78/rl78.c (add_vector_labels): New function.
8503 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8504 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8505 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8506 which checks that no arguments are passed.
8507 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8508 * doc/extend.texi: Documentation for the new attribute.
8510 2018-02-13 Andreas Schwab <schwab@suse.de>
8512 * config/riscv/linux.h (CPP_SPEC): Define.
8514 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8517 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8518 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8519 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8520 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8521 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8522 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8523 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8524 temporarily for AES and PCLMUL builtins.
8526 PR tree-optimization/84339
8527 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8528 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8532 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8533 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8534 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8536 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8538 * omp-simd-clone.h: New file.
8539 * omp-simd-clone.c: Include omp-simd-clone.h.
8540 (expand_simd_clones): No longer static.
8541 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8542 cgraph.h and omp-simd-clone.h.
8543 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8544 (vect_recog_widen_shift_pattern): Formatting fix.
8545 (vect_pattern_recog_1): Don't check optab for calls.
8548 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8549 operands[2] into a REG before using gen_lowpart on it.
8551 2018-02-12 Jeff Law <law@redhat.com>
8554 * config/sh/sh.c (find_barrier): Consider a sibling call
8557 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8558 successfully back substituting a reg.
8560 2018-02-12 Richard Biener <rguenther@suse.de>
8562 PR tree-optimization/84037
8563 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8564 parameter, move visited init to caller.
8565 (vect_slp_analyze_operations): Separate cost from validity
8566 check, initialize visited once for all instances.
8567 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8569 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8571 (vect_model_promotion_demotion_cost): Likewise.
8572 (vectorizable_bswap): Guard cost modeling with !slp_node
8573 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8575 (vectorizable_call): Likewise.
8576 (vectorizable_conversion): Likewise.
8577 (vectorizable_assignment): Likewise.
8578 (vectorizable_shift): Likewise.
8579 (vectorizable_operation): Likewise.
8580 (vectorizable_store): Likewise.
8581 (vectorizable_load): Likewise.
8582 (vectorizable_condition): Likewise.
8583 (vectorizable_comparison): Likewise.
8585 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8588 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8589 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8590 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8592 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8595 * params.def (inline-min-speedup): Increase from 8 to 15.
8596 (max-inline-insns-auto): Decrease from 40 to 30.
8597 * ipa-split.c (consider_split): Add some buffer for function to
8598 be considered inlining candidate.
8599 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8602 2018-02-12 Richard Biener <rguenther@suse.de>
8604 PR tree-optimization/84037
8605 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8606 matched stmts if we cannot swap the non-matched ones.
8608 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8610 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8611 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8612 _mm_maskz_scalef_round_ss): New intrinsics.
8613 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8614 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8615 __builtin_ia32_scalefss_round): Remove.
8616 (__builtin_ia32_scalefsd_mask_round,
8617 __builtin_ia32_scalefss_mask_round): New intrinsics.
8618 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8619 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8620 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8621 "<round_constraint>")): Changed to ...
8622 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8623 "<round_scalar_constraint>")): ... this.
8624 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8625 %0, %1, %2<round_op3>}"): Changed to ...
8626 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8627 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8628 %2<round_scalar_mask_op3>}"): ... this.
8629 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8631 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8633 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8634 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8635 (_mm_maskz_sqrt_round_ss): New intrinsics.
8636 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8637 (__builtin_ia32_sqrtsd_mask_round)
8638 (__builtin_ia32_sqrtss_mask_round): New builtins.
8639 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8640 (__builtin_ia32_sqrtss_round): Remove.
8641 (__builtin_ia32_sqrtsd_mask_round)
8642 (__builtin_ia32_sqrtss_mask_round): New builtins.
8643 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8644 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8645 ((match_operand:VF_128 1 "vector_operand"
8646 "xBm,<round_constraint>")): Changed to ...
8647 ((match_operand:VF_128 1 "vector_operand"
8648 "xBm,<round_scalar_constraint>")): ... this.
8649 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8650 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8651 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8652 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8653 %<iptr>1<round_scalar_mask_op3>}): ... this.
8654 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8655 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8657 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8660 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8661 Cast vec_cmpeq result to correct type.
8662 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8663 Cast vec_cmpgt result to correct type.
8665 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8667 * final.c (final_scan_insn_1): Renamed from...
8668 (final_scan_insn): ... this. New wrapper, to recover
8669 seen from the outermost call in recursive ones.
8670 * config/sparc/sparc.c (output_return): Drop seen from call.
8671 (output_sibcall): Likewise.
8672 * config/visium/visium.c (output_branch): Likewise.
8674 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8676 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8679 2018-02-10 Alan Modra <amodra@gmail.com>
8682 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8683 Specify LR as an input.
8685 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8688 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8689 remove_member_access_dummy_vars): New functions.
8690 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8691 lower_omp_1, execute_lower_omp): Use them.
8693 PR rtl-optimization/84308
8694 * shrink-wrap.c (spread_components): Release todo vector.
8696 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8698 PR rtl-optimization/57193
8699 * ira-color.c (struct allocno_color_data): Add member
8700 conflict_allocno_hard_prefs.
8701 (update_conflict_allocno_hard_prefs): New.
8702 (bucket_allocno_compare_func): Add a preference based on
8703 conflict_allocno_hard_prefs.
8704 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8705 (color_allocnos): Remove a dead code. Initiate
8706 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8708 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8711 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8712 constraint from =wa to wa. Avoid a subreg on the output operand,
8713 instead use a pseudo and subreg it in a move.
8714 (p9_xxbrd_<mode>): Changed to ...
8715 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8716 (p9_xxbrd_v2df): New expander.
8717 (p9_xxbrw_<mode>): Changed to ...
8718 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8719 (p9_xxbrw_v4sf): New expander.
8721 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8723 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8725 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8728 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8729 multiply in 32-bit mode.
8730 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8731 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8734 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8736 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8737 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8738 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8739 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8741 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8744 * dwarf2out.c (is_trivial_indirect_ref): New function.
8745 (dwarf2out_late_global_decl): Do not generate a location
8746 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8747 and that are not defined in the current unit.
8749 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8751 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8752 instead of a libcall for UNORDERED.
8754 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8757 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8758 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8760 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8763 * config/s390/s390.c (s390_set_current_function): Invoke
8764 s390_indirect_branch_settings also if fndecl didn't change.
8766 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8768 * config/rs6000/rs6000.md (blockage): Set length to zero.
8770 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8772 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8774 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8777 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8778 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8782 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8783 PARALLEL incoming that failed vt_get_decl_and_offset check.
8786 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8787 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8789 (get_variable_section): For decls in named .bss* sections pass true as
8790 second argument to bss_initializer_p.
8792 2018-02-09 Marek Polacek <polacek@redhat.com>
8793 Jakub Jelinek <jakub@redhat.com>
8796 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8797 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8798 Sync some changes from cxx_fold_indirect_ref.
8800 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8802 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8804 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8805 (BLOCK_INLINE_ENTRY_LABEL): New.
8806 (dwarf2out_var_location): Disregard inline entry markers.
8807 (inline_entry_data): New struct.
8808 (inline_entry_data_hasher): New hashtable type.
8809 (inline_entry_data_hasher::hash): New.
8810 (inline_entry_data_hasher::equal): New.
8811 (inline_entry_data_table): New variable.
8812 (add_high_low_attributes): Add DW_AT_entry_pc and
8813 DW_AT_GNU_entry_view attributes if a pending entry is found
8814 in inline_entry_data_table. Add old entry_pc attribute only
8815 if debug nonbinding markers are disabled.
8816 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8817 markers are enabled.
8818 (block_within_block_p, dwarf2out_inline_entry): New.
8819 (dwarf2out_finish): Check that no entries remained in
8820 inline_entry_data_table.
8821 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8822 (final_scan_insn, notice_source_line): Likewise.
8823 (rest_of_clean_state): Skip inline entry markers.
8824 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8826 * gimple.c (gimple_build_debug_inline_entry): New.
8827 * gimple.h (enum gimple_debug_subcode): Add
8828 GIMPLE_DEBUG_INLINE_ENTRY.
8829 (gimple_build_debug_inline_entry): Declare.
8830 (gimple_debug_inline_entry_p): New.
8831 (gimple_debug_nonbind_marker_p): Adjust.
8832 * insn-notes.def (INLINE_ENTRY): New.
8833 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8834 inline entry marker notes.
8835 (print_insn): Likewise.
8836 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8837 (INSN_DEBUG_MARKER_KIND): Likewise.
8838 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8839 * tree-inline.c (expand_call_inline): Build and insert
8840 debug_inline_entry stmt.
8841 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8842 inline entry blocks early, if nonbind markers are enabled.
8843 (dump_scope_block): Dump fragment info.
8844 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8845 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8846 (gimple_build_debug_inline_entry): New.
8847 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8848 Enable/disable inline entry points too.
8849 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8850 (DEBUG_INSN): Describe inline entry markers.
8852 * common.opt (gvariable-location-views): New.
8853 (gvariable-location-views=incompat5): New.
8854 * config.in: Rebuilt.
8855 * configure: Rebuilt.
8856 * configure.ac: Test assembler for view support.
8857 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8858 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8859 * dwarf2out.c (var_loc_view): New typedef.
8860 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8861 (dwarf2out_locviews_in_attribute): New.
8862 (dwarf2out_locviews_in_loclist): New.
8863 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8864 (enum dw_line_info_opcode): Add LI_adv_address.
8865 (struct dw_line_info_table): Add view.
8866 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8867 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8868 (zero_view_p): New variable.
8869 (ZERO_VIEW_P): New macro.
8870 (output_asm_line_debug_info): New.
8871 (struct var_loc_node): Add view.
8872 (add_AT_view_list, AT_loc_list): New.
8873 (add_var_loc_to_decl): Add view param. Test it against last.
8874 (new_loc_list): Add view params. Record them.
8875 (AT_loc_list_ptr): Handle loc and view lists.
8876 (view_list_to_loc_list_val_node): New.
8877 (print_dw_val): Handle dw_val_class_view_list.
8878 (size_of_die): Likewise.
8879 (value_format): Likewise.
8880 (loc_list_has_views): New.
8881 (gen_llsym): Set vl_symbol too.
8882 (maybe_gen_llsym, skip_loc_list_entry): New.
8883 (dwarf2out_maybe_output_loclist_view_pair): New.
8884 (output_loc_list): Output view list or entries too.
8885 (output_view_list_offset): New.
8886 (output_die): Handle dw_val_class_view_list.
8887 (output_dwarf_version): New.
8888 (output_compilation_unit_header): Use it.
8889 (output_skeleton_debug_sections): Likewise.
8890 (output_rnglists, output_line_info): Likewise.
8891 (output_pubnames, output_aranges): Update version comments.
8892 (output_one_line_info_table): Output view numbers in asm comments.
8893 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8894 Call maybe_gen_llsym.
8895 (loc_list_from_tree_1): Adjust.
8896 (add_AT_location_description): Create view list attribute if
8897 needed, check it's absent otherwise.
8898 (convert_cfa_to_fb_loc_list): Adjust.
8899 (maybe_emit_file): Call output_asm_line_debug_info for test.
8900 (dwarf2out_var_location): Reset views as needed. Precompute
8901 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8902 attribute. Set view.
8903 (new_line_info_table): Reset next view.
8904 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8905 (dwarf2out_source_line): Likewise. Output view resets and labels to
8906 the assembler, or select appropriate line info opcodes.
8907 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8908 (optimize_string_length): Catch it. Adjust.
8909 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8910 dw_val_class_view_list, and remove it if no longer needed.
8911 (hash_loc_list): Hash view numbers.
8912 (loc_list_hasher::equal): Compare them.
8913 (optimize_location_lists): Check whether a view list symbol is
8914 needed, and whether the locview attribute is present, and
8915 whether they match. Remove the locview attribute if no longer
8917 (index_location_lists): Call skip_loc_list_entry for test.
8918 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8919 Use output_dwarf_version.
8920 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8921 (struct dw_val_node): Add val_view_list.
8922 * final.c (SEEN_NEXT_VIEW): New.
8923 (set_next_view_needed): New.
8924 (clear_next_view_needed): New.
8925 (maybe_output_next_view): New.
8926 (final_start_function): Rename to...
8927 (final_start_function_1): ... this. Take pointer to FIRST,
8928 add SEEN parameter. Emit param bindings in the initial view.
8929 (final_start_function): Reintroduce SEEN-less interface.
8930 (final): Rename to...
8931 (final_1): ... this. Take SEEN parameter. Output final pending
8932 next view at the end.
8933 (final): Reintroduce seen-less interface.
8934 (final_scan_insn): Output pending next view before switching
8935 sections or ending a block. Mark the next view as needed when
8936 outputting variable locations. Notify debug backend of section
8937 changes, and of location view changes.
8938 (rest_of_handle_final): Adjust.
8939 * toplev.c (process_options): Autodetect value for debug variable
8940 location views option. Warn on incompat5 without -gdwarf-5.
8941 * doc/invoke.texi (gvariable-location-views): New.
8942 (gvariable-location-views=incompat5): New.
8943 (gno-variable-location-views): New.
8945 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8947 PR tree-optimization/84136
8948 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8949 that the result of find_edge is non-NULL.
8951 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8954 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8955 storing integer register in SImode. Fix cost of 256 and 512
8956 byte aligned SSE register store.
8958 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8960 * config/i386/i386.c (ix86_multiplication_cost): Fix
8961 multiplication cost for TARGET_AVX512DQ.
8963 2018-02-08 Marek Polacek <polacek@redhat.com>
8965 PR tree-optimization/84238
8966 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8969 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8971 PR tree-optimization/84265
8972 * tree-vect-stmts.c (vectorizable_store): Don't treat
8973 VMAT_CONTIGUOUS accesses as grouped.
8974 (vectorizable_load): Likewise.
8976 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8978 PR tree-optimization/81635
8979 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8980 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8981 (test_round_for_mask): New functions.
8982 (wide_int_cc_tests): Call test_round_for_mask.
8983 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8984 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8985 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8986 range returned by get_range_info.
8988 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8991 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8992 * symtab.c: Include builtins.h
8993 (symtab_node::output_to_lto_symbol_table_p): Move here
8994 from lto-streamer-out.c:output_symbol_p.
8995 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8996 (output_symbol_p): Move all logic to symtab.c
8997 (produce_symtab): Update.
8999 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9001 * config/s390/s390-opts.h (enum indirect_branch): Define.
9002 * config/s390/s390-protos.h (s390_return_addr_from_memory)
9003 (s390_indirect_branch_via_thunk)
9004 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
9005 (enum s390_indirect_branch_type): Define.
9006 * config/s390/s390.c (struct s390_frame_layout, struct
9007 machine_function): Remove.
9008 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
9009 (indirect_branch_table_label_no, indirect_branch_table_name):
9011 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
9012 (enum s390_indirect_branch_option): Define.
9013 (s390_return_addr_from_memory): New function.
9014 (s390_handle_string_attribute): New function.
9015 (s390_attribute_table): Add new attribute handler.
9016 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
9017 (s390_indirect_branch_via_thunk): New function.
9018 (s390_indirect_branch_via_inline_thunk): New function.
9019 (s390_function_ok_for_sibcall): When jumping via thunk disallow
9020 sibling call optimization for non z10 compiles.
9021 (s390_emit_call): Force indirect branch target to be a single
9022 register. Add r1 clobber for non-z10 compiles.
9023 (s390_emit_epilogue): Emit return jump via return_use expander.
9024 (s390_reorg): Handle JUMP_INSNs as execute targets.
9025 (s390_option_override_internal): Perform validity checks for the
9026 new command line options.
9027 (s390_indirect_branch_attrvalue): New function.
9028 (s390_indirect_branch_settings): New function.
9029 (s390_set_current_function): Invoke s390_indirect_branch_settings.
9030 (s390_output_indirect_thunk_function): New function.
9031 (s390_code_end): Implement target hook.
9032 (s390_case_values_threshold): Implement target hook.
9033 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
9035 * config/s390/s390.h (struct s390_frame_layout)
9036 (struct machine_function): Move here from s390.c.
9037 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9038 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9039 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9040 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9041 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9042 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9043 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9044 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9045 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9046 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9047 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9048 (mnemonic attribute): Add values which aren't recognized
9050 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9051 pattern for branch conversion. Fix mnemonic attribute.
9052 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9053 indirect branch via thunk if requested.
9054 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9055 ("*indirect_jump"): Disable for branch conversion using out of
9057 ("indirect_jump_via_thunk<mode>_z10")
9058 ("indirect_jump_via_thunk<mode>")
9059 ("indirect_jump_via_inlinethunk<mode>_z10")
9060 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9061 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9062 ("casesi_jump_via_inlinethunk<mode>_z10")
9063 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9064 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9065 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9066 ("*indirect2_jump"): Disable for branch conversion.
9067 ("casesi_jump"): Turn into expander and expand patterns for branch
9069 ("return_use"): New expander.
9070 ("*return"): Emit return via thunk and rename it to ...
9071 ("*return<mode>"): ... this one.
9072 * config/s390/s390.opt: Add new options and and enum for the
9075 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9077 * lra-constraints.c (match_reload): Unconditionally use
9078 gen_lowpart_SUBREG, rather than selecting between that
9079 and equivalent gen_rtx_SUBREG code.
9081 2018-02-08 Richard Biener <rguenther@suse.de>
9083 PR tree-optimization/84233
9084 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9085 changed flag instead of boguously re-using phi_inserted.
9087 2018-02-08 Martin Jambor <mjambor@suse.cz>
9089 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9090 static local variables.
9092 2018-02-08 Richard Biener <rguenther@suse.de>
9094 PR tree-optimization/84278
9095 * tree-vect-stmts.c (vectorizable_store): When looking for
9096 smaller vector types to perform grouped strided loads/stores
9097 make sure the mode is supported by the target.
9098 (vectorizable_load): Likewise.
9100 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9102 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9103 Increase LDP/STP opportunities by adding adjacent callee-saves.
9105 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9107 PR rtl-optimization/84068
9108 PR rtl-optimization/83459
9109 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9111 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9113 PR tree-optimization/84224
9114 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9115 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9118 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9121 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9122 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9123 count, remove one USE.
9125 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9127 * doc/install.texi (Configuration): Document the
9128 --with-long-double-format={ibm,ieee} PowerPC configuration
9132 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9133 Convert from define_expand to be define_insn_and_split. Rework
9134 float/double/_Float128 conversions to QI/HI/SImode to work with
9135 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9136 conversions to QI/HImode types did a store and then a load to
9137 truncate the value. For conversions to VSX registers, don't split
9138 the insn, instead emit the code directly. Use the code iterator
9139 any_fix to combine signed and unsigned conversions.
9140 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9141 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9142 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9143 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9144 (fix_<mode>di2_hw): Likewise.
9145 (fixuns_<mode>di2_hw): Likewise.
9146 (fix_<mode>si2_hw): Likewise.
9147 (fixuns_<mode>si2_hw): Likewise.
9148 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9149 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9150 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9151 fix<uns>_trunc<SFDF:mode>si2_p8.
9152 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9154 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9155 (fix<uns>_<mode>_mem): Likewise.
9156 (fctiw<u>z_<mode>_mem): Likewise.
9157 (fix<uns>_<mode>_mem): Likewise.
9158 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9159 the register allocator from doing a direct move to the GPRs to do
9160 a store, and instead use the ISA 3.0 store byte/half-word from
9161 vector register instruction. For IEEE 128-bit floating point,
9162 also optimize stores of 32-bit ints.
9163 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9165 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9167 * genextract.c (push_pathstr_operand): New function to support
9169 (walk_rtx): Call push_pathstr_operand.
9170 (print_path): Support [a-zA-Z].
9172 2018-02-07 Richard Biener <rguenther@suse.de>
9174 PR tree-optimization/84037
9175 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9176 (cse_and_gimplify_to_preheader): Declare.
9177 (vect_get_place_in_interleaving_chain): Likewise.
9178 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9180 (_loop_vec_info::~_loop_vec_info): Delete it.
9181 (cse_and_gimplify_to_preheader): New function.
9182 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9183 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9184 (vectorizable_load): Likewise. For grouped stores always base
9185 the IV on the first element.
9186 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9187 condition before gimplifying.
9189 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9191 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9192 *DIV_EXPR and *MOD_EXPR.
9194 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9197 * config/i386/i386.c (ix86_option_override_internal): Mask out
9198 the CF_SET bit when checking -fcf-protection.
9200 2018-02-07 Tom de Vries <tom@codesourcery.com>
9203 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9206 2018-02-07 Richard Biener <rguenther@suse.de>
9208 PR tree-optimization/84204
9209 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9212 PR tree-optimization/84205
9213 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9214 special-case isl_ast_op_zdiv_r.
9216 PR tree-optimization/84223
9217 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9218 Only add conditions from within the region.
9219 (gather_bbs::after_dom_children): Adjust.
9221 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9224 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9225 * config/avr/avr.md: Only post-reload split REG-REG moves if
9226 either register is GENERAL_REG_P.
9228 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9230 PR tree-optimization/84235
9231 * tree-ssa-scopedtables.c
9232 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9233 if the subtraction is performed in floating point type where NaNs are
9234 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9235 build 1. Formatting fix.
9237 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9240 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9241 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9242 and skip it regardless of bb boundaries. Use CALL_P macro,
9243 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9244 unnecessarily, formatting fix.
9246 2018-02-06 Michael Collison <michael.collison@arm.com>
9248 * config/arm/thumb2.md:
9249 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9250 (*thumb_mov_notscc): Ditto.
9252 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9255 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9256 unsigned_fix, not "s".
9258 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9260 * configure.ac (gcc_fn_eh_frame_ro): New function.
9261 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9262 correct .eh_frame permissions.
9263 * configure: Regenerate.
9265 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9267 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9270 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9272 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9273 Display warning message for -mno-speculate-indirect-jumps.
9275 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9277 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9279 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9281 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9283 PR tree-optimization/84225
9284 * tree-eh.c (find_trapping_overflow): Only call
9285 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9287 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9290 * config/i386/i386.c: Reimplement the check of possible options
9291 -mibt/-mshstk conbination. Change error messages.
9292 * doc/invoke.texi: Fix a typo: remove extra '='.
9294 2018-02-06 Marek Polacek <polacek@redhat.com>
9296 PR tree-optimization/84228
9297 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9299 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9302 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9303 emitted arch directives.
9304 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9305 __ARM_FEATURE_COPROC before changing architectures.
9307 2018-02-06 Richard Biener <rguenther@suse.de>
9309 * config/i386/i386.c (print_reg): Fix typo.
9310 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9312 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9314 * configure: Regenerate.
9316 2018-02-05 Martin Sebor <msebor@redhat.com>
9318 PR tree-optimization/83369
9319 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9322 2018-02-05 Martin Liska <mliska@suse.cz>
9324 * doc/invoke.texi: Cherry-pick upstream r323995.
9326 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9328 * ira.c (ira_init_register_move_cost): Adjust comment.
9330 2018-02-05 Martin Liska <mliska@suse.cz>
9332 PR gcov-profile/84137
9333 * doc/gcov.texi: Fix typo in documentation.
9335 2018-02-05 Martin Liska <mliska@suse.cz>
9337 PR gcov-profile/83879
9338 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9339 using dlopen functionality.
9341 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9343 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9344 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9345 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9346 _mm_maskz_range_round_ss): New intrinsics.
9347 (__builtin_ia32_rangesd128_round)
9348 (__builtin_ia32_rangess128_round): Remove.
9349 (__builtin_ia32_rangesd128_mask_round,
9350 __builtin_ia32_rangess128_mask_round): New builtins.
9351 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9352 __builtin_ia32_rangess128_round): Remove.
9353 (__builtin_ia32_rangesd128_mask_round,
9354 __builtin_ia32_rangess128_mask_round): New builtins.
9355 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9356 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9357 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9358 "<round_saeonly_constraint>")): Changed to ...
9359 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9360 "<round_saeonly_scalar_constraint>")): ... this.
9361 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9362 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9363 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9364 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9365 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9367 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9369 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9371 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9372 Remove all values except native, 8540 and 8548.
9374 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9376 * config/i386/i386.c (ix86_output_function_return): Pass
9377 INVALID_REGNUM, instead of -1, as invalid register number to
9378 indirect_thunk_name and output_indirect_thunk.
9380 2018-02-02 Julia Koval <julia.koval@intel.com>
9382 * config.gcc: Add -march=icelake.
9383 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9384 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9385 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9386 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9387 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9388 (processor_target_table): Add icelake.
9389 (ix86_option_override_internal): Handle new PTAs.
9390 (get_builtin_code_for_version): Handle icelake.
9391 (M_INTEL_COREI7_ICELAKE): New.
9392 (fold_builtin_cpu): Handle icelake.
9393 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9394 * doc/invoke.texi: Add -march=icelake.
9396 2018-02-02 Julia Koval <julia.koval@intel.com>
9398 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9399 to wide_int_bitmask.
9400 * wide-int-bitmask.h: New.
9402 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9405 * config/i386/i386.md: Replace Pmode with word_mode in
9406 builtin_setjmp_setup and builtin_longjmp to support x32.
9408 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9412 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9414 (rs6000_supported_cpu_names): New static variable.
9415 (linux_cpu_translation_table): Likewise.
9416 (elf_platform) <cpu>: Define new static variable and use it.
9417 Translate kernel AT_PLATFORM name to canonical name if needed.
9418 Error if platform name is unknown.
9420 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9423 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9425 2018-02-01 Jeff Law <law@redhat.com>
9428 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9429 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9430 the scratch if RELEASE_VIA_POP is false.
9431 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9432 If we have to save a temporary register, decrement SIZE appropriately.
9433 Pass new arguments to release_scratch_register_on_entry.
9434 (ix86_adjust_stack_and_probe): Likewise.
9435 (ix86_emit_probe_stack_range): Pass new arguments to
9436 release_scratch_register_on_entry.
9438 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9440 PR rtl-optimization/84157
9441 * combine.c (change_zero_ext): Use REG_P predicate in
9442 front of HARD_REGISTER_P predicate.
9444 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9446 * config/avr/avr.c (avr_option_override): Move disabling of
9447 -fdelete-null-pointer-checks to...
9448 * common/config/avr/avr-common.c (avr_option_optimization_table):
9451 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9453 PR tree-optimization/81635
9454 * tree-data-ref.c (split_constant_offset_1): For types that
9455 wrap on overflow, try to use range info to prove that wrapping
9458 2018-02-01 Renlin Li <renlin.li@arm.com>
9461 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9463 (aarch64_register_move_cost): Likewise.
9464 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9466 (REG_CLASS_NAMES): Likewise.
9467 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9468 TAILCALL_ADDR_REGS. Remove IP registers.
9469 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9471 2018-02-01 Richard Biener <rguenther@suse.de>
9473 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9474 for specifying RPO order and allow NULL for that.
9475 * domwalk.c (dom_walker::dom_walker): Likewise.
9476 (dom_walker::walk): Handle NULL RPO order.
9477 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9479 (rewrite_update_dom_walker): Likewise.
9480 (mark_def_dom_walker): Likewise.
9482 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9484 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9485 (aarch64_maybe_expand_sve_subreg_move): Declare.
9486 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9487 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9489 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9490 that are semantically a reverse operation.
9491 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9492 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9493 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9495 (aarch64_can_change_mode_class): For big-endian, forbid changes
9496 between two SVE modes if they have different element sizes.
9498 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9500 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9501 the TImode handling for big-endian targets.
9503 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9505 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9506 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9508 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9509 Remove BSWAP handing for big-endian targets and use the form of
9510 LD1RQ appropariate for the mode.
9512 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9514 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9515 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9518 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9521 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9522 check for operands that need to go through aarch64_sve_reload_be.
9524 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9526 PR tree-optimization/81661
9527 PR tree-optimization/84117
9528 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9529 * tree-eh.c: Include gimplify.h.
9530 (find_trapping_overflow, replace_trapping_overflow,
9531 rewrite_to_non_trapping_overflow): New functions.
9532 * tree-vect-loop.c: Include tree-eh.h.
9533 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9534 * tree-data-ref.c: Include tree-eh.h.
9535 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9537 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9539 PR rtl-optimization/84123
9540 * combine.c (change_zero_ext): Check if hard register satisfies
9541 can_change_dest_mode before calling gen_lowpart_SUBREG.
9543 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9546 * ira.c (ira_init_register_move_cost): Remove assert.
9548 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9550 PR rtl-optimization/84071
9551 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9552 * doc/tm.texi: Regenerate.
9554 2018-01-31 Richard Biener <rguenther@suse.de>
9556 PR tree-optimization/84132
9557 * tree-data-ref.c (analyze_miv_subscript): Properly
9558 check whether evolution_function_is_affine_multivariate_p
9559 before calling gcd_of_steps_may_divide_p.
9561 2018-01-31 Julia Koval <julia.koval@intel.com>
9564 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9565 * config/i386/i386.md (rdpid_rex64) New.
9566 (rdpid): Make 32bit only.
9568 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9571 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9572 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9574 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9577 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9579 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9581 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9583 PR rtl-optimization/84071
9584 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9585 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9587 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9589 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9590 (arc_attribute_table): Add 'aux' attribute.
9591 (arc_in_small_data_p): Consider aux like variables.
9592 (arc_is_aux_reg_p): New function.
9593 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9594 (arc_get_aux_arg): New function.
9595 (prepare_move_operands): Handle aux-register access.
9596 (arc_handle_aux_attribute): New function.
9597 * doc/extend.texi (ARC Variable attributes): Add subsection.
9599 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9601 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9602 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9603 (arc_attribute_table): Add 'uncached' attribute.
9604 (arc_print_operand): Print '.di' flag for uncached memory
9606 (arc_in_small_data_p): Do not consider for small data the uncached
9608 (arc_is_uncached_mem_p): New function.
9609 * config/arc/predicates.md (compact_store_memory_operand): Check
9610 for uncached memory accesses.
9611 (nonvol_nonimm_operand): Likewise.
9612 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9614 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9617 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9618 falign-loops=): Add Optimization flag.
9620 2018-01-30 Jeff Law <law@redhat.com>
9623 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9624 INT_REGISTERS_SAVED. Check it prior to calling
9625 get_scratch_register_on_entry.
9626 (ix86_adjust_stack_and_probe): Similarly.
9627 (ix86_emit_probe_stack_range): Similarly.
9628 (ix86_expand_prologue): Corresponding changes.
9630 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9633 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9634 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9636 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9639 * lra-constraints.c (curr_insn_transform): Process AND in the
9642 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9644 PR rtl-optimization/83986
9645 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9646 dependence against last_pending_memory_flush in addition to
9649 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9651 PR tree-optimization/81611
9652 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9655 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9658 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9661 2018-01-30 Richard Biener <rguenther@suse.de>
9662 Jakub Jelinek <jakub@redhat.com>
9664 PR tree-optimization/84111
9665 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9666 inner loops added during recursion, as they don't have up-to-date
9669 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9672 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9673 (can_inline_edge_by_limits_p): ... here.
9674 (can_early_inline_edge_p, check_callers,
9675 update_caller_keys, update_callee_keys, recursive_inlining,
9676 add_new_edges_to_heap, speculation_useful_p,
9677 inline_small_functions,
9678 inline_small_functions, flatten_function,
9679 inline_to_all_callers_1): Update.
9681 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9683 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9686 2018-01-30 Richard Biener <rguenther@suse.de>
9688 PR tree-optimization/83008
9689 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9690 invariant and constant vector uses in stmts when they need
9693 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9696 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9697 * configure: Regenerate.
9699 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9701 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9703 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9704 Use gen_rtx_REG rather than gen_lowpart.
9706 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9708 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9709 rather than 0 when creating partial subregs.
9711 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9713 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9716 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9719 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9720 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9721 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9722 flags. This restores the settings used before the 2017-07-24.
9723 Turning off pre increment/decrement/modify allows IVOPTS to
9724 optimize DF/SF loops where the index is an int.
9726 2018-01-29 Richard Biener <rguenther@suse.de>
9727 Kelvin Nilsen <kelvin@gcc.gnu.org>
9730 * tree-vect-stmts.c (vectorizable_call): Don't call
9731 targetm.vectorize_builtin_md_vectorized_function if callee is
9734 2018-01-22 Carl Love <cel@us.ibm.com>
9736 * doc/extend.tex: Fix typo in second arg in
9737 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9739 2018-01-29 Richard Biener <rguenther@suse.de>
9741 PR tree-optimization/84086
9742 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9743 (flush_ssaname_freelist): When SSA names were released reset
9744 the SCEV hash table.
9746 2018-01-29 Richard Biener <rguenther@suse.de>
9748 PR tree-optimization/84057
9749 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9750 removed paths when removing edges.
9752 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9754 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9755 -mfunction-return=@var{choice}.
9757 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9760 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9762 (layout::print_source_line): Likewise.
9763 (test_get_line_width_without_trailing_whitespace): Add test cases.
9765 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9768 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9771 2018-01-26 Jim Wilson <jimw@sifive.com>
9773 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9775 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9778 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9780 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9781 and CMP + SUB-immediate -> SUBS.
9783 2018-01-26 Martin Sebor <msebor@redhat.com>
9785 PR tree-optimization/83896
9786 * tree-ssa-strlen.c (get_string_len): Rename...
9787 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9788 Avoid assuming length is constant.
9789 (handle_char_store): Use HOST_WIDE_INT for string length.
9791 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9794 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9795 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9797 2018-01-26 Richard Biener <rguenther@suse.de>
9799 PR rtl-optimization/84003
9800 * dse.c (record_store): Only record redundant stores when
9801 the earlier store aliases at least all accesses the later one does.
9803 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9805 PR rtl-optimization/83985
9806 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9807 REG_CFA_RESTORE insns.
9808 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9809 REG_CFA_RESTORE insns here.
9812 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9813 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9815 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9817 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9818 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9819 (arc_init): Likewise.
9820 (arc_override_options): Likewise.
9821 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9823 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9825 * config/arc/arc.h (TARGET_DBNZ): Define.
9826 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9827 properly set the tune attribute.
9828 (dbnz): Use TARGET_DBNZ guard.
9829 * config/arc/arc.opt (mtune): Add core3 option.
9831 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9833 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9834 recognize new pic like addresses.
9835 (arc_delegitimize_address): Clean up.
9837 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9839 * config/arc/arc-arches.def: Option mrf16 valid for all
9841 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9842 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9843 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9844 * config/arc/arc-tables.opt: Regenerate.
9845 * config/arc/arc.c (arc_conditional_register_usage): Handle
9846 reduced register file case.
9847 (arc_file_start): Set must have build attributes.
9848 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9850 * config/arc/arc.opt (mrf16): Add new option.
9851 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9852 * config/arc/genmultilib.awk: Handle new mrf16 option.
9853 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9854 * config/arc/t-multilib: Regenerate.
9855 * doc/invoke.texi (ARC Options): Document mrf16 option.
9857 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9859 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9860 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9861 (arc_attribute_table): Add 'secure_call' attribute.
9862 (arc_print_operand): Print secure call operand.
9863 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9865 (arc_is_secure_call_p): New function. * config/arc/arc.md
9866 (call_i): Add support for sjli instruction.
9867 (call_value_i): Likewise.
9868 * config/arc/constraints.md (Csc): New constraint.
9870 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9871 John Eric Martin <John.Martin@emmicro-us.com>
9873 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9874 * config/arc/arc.c (_arc_jli_section): New struct.
9875 (arc_jli_section): New type.
9876 (rc_jli_sections): New static variable.
9877 (arc_handle_jli_attribute): New function.
9878 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9879 (arc_file_end): New function.
9880 (TARGET_ASM_FILE_END): Define.
9881 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9882 (arc_add_jli_section): New function.
9883 (jli_call_scan): Likewise.
9884 (arc_reorg): Call jli_call_scan.
9885 (arc_output_addsi): Remove 'S' from printing asm operand.
9886 (arc_is_jli_call_p): New function.
9887 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9889 (movhi_insn): Likewise.
9890 (movsi_insn): Likewise.
9891 (movsi_set_cc_insn): Likewise.
9892 (loadqi_update): Likewise.
9893 (load_zeroextendqisi_update): Likewise.
9894 (load_signextendqisi_update): Likewise.
9895 (loadhi_update): Likewise.
9896 (load_zeroextendhisi_update): Likewise.
9897 (load_signextendhisi_update): Likewise.
9898 (loadsi_update): Likewise.
9899 (loadsf_update): Likewise.
9900 (movsicc_insn): Likewise.
9901 (bset_insn): Likewise.
9902 (bxor_insn): Likewise.
9903 (bclr_insn): Likewise.
9904 (bmsk_insn): Likewise.
9905 (bicsi3_insn): Likewise.
9906 (cmpsi_cc_c_insn): Likewise.
9907 (movsi_ne): Likewise.
9908 (movsi_cond_exec): Likewise.
9909 (clrsbsi2): Likewise.
9920 (call_i): Remove 'S' asm letter, add jli instruction.
9921 (call_value_i): Likewise.
9922 * config/arc/arc.op (mjli-always): New option.
9923 * config/arc/constraints.md (Cji): New constraint.
9924 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9926 (subsf3_fpx): Likewise.
9927 (mulsf3_fpx): Likewise.
9928 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9930 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9932 * doc/invoke.texi (ARC): Document mjli-always option.
9934 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9936 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9937 avoid addition with 0 and use incw and decw where possible.
9939 2018-01-26 Richard Biener <rguenther@suse.de>
9941 PR tree-optimization/81082
9942 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9943 association if it requires casting to unsigned.
9944 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9945 from fold_plusminus_mult_expr to catch important cases late when
9946 range info is available.
9948 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9950 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9951 * configure.ac (hidden_linkonce): New test.
9952 * configure: Regenerate.
9953 * config.in: Regenerate.
9955 2018-01-26 Julia Koval <julia.koval@intel.com>
9957 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9958 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9959 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9960 _mm_mask_bitshuffle_epi64_mask): Fix type.
9961 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9962 USI_FTYPE_V4DI_V4DI_USI): Remove.
9963 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9964 __builtin_ia32_vpshufbitqmb256_mask,
9965 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9966 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9967 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9969 2018-01-26 Alan Modra <amodra@gmail.com>
9972 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9973 UNSPEC_VBPERMQ. Sort other unspecs.
9975 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9977 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9979 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9982 * predict.c (drop_profile): Do not push/pop cfun; update also
9984 (handle_missing_profiles): Fix logic looking for zero profiles.
9986 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9989 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9990 on functions with #pragma omp declare simd or functions with simd
9992 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9993 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9994 Remove trailing \n from warning_at calls.
9996 2018-01-25 Tom de Vries <tom@codesourcery.com>
9999 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10000 for neutered workers.
10002 2018-01-24 Joseph Myers <joseph@codesourcery.com>
10005 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
10006 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
10008 2018-01-24 Jeff Law <law@redhat.com>
10011 * i386.c (get_probe_interval): Move to earlier point.
10012 (ix86_compute_frame_layout): If -fstack-clash-protection and
10013 the frame is larger than the probe interval, then use pushes
10014 to save registers rather than reg->mem moves.
10015 (ix86_expand_prologue): Remove conditional for int_registers_saved
10018 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
10021 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
10022 min/max for never referenced object.
10024 2018-01-24 Jakub Jelinek <jakub@redhat.com>
10026 PR middle-end/83977
10027 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
10029 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
10030 attributes from DECL_ATTRIBUTES (decl) without affecting
10031 DECL_ATTRIBUTES (current_function_decl).
10032 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
10033 functions with non-NULL DECL_ABSTRACT_ORIGIN.
10035 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
10037 PR tree-optimization/83979
10038 * fold-const.c (fold_comparison): Use constant_boolean_node
10039 instead of boolean_{true,false}_node.
10041 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10043 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10046 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10048 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10049 Simplify the clause that sets the length attribute.
10050 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10051 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10052 clause that sets the length attribute.
10053 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10055 2018-01-24 Tom de Vries <tom@codesourcery.com>
10058 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10059 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10060 Add strict parameter.
10061 (prevent_branch_around_nothing): Insert dummy insn between branch to
10062 label and label with no ptx insn inbetween.
10063 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10065 2018-01-24 Tom de Vries <tom@codesourcery.com>
10068 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10069 for neutered threads in warp.
10070 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10072 2018-01-24 Richard Biener <rguenther@suse.de>
10074 PR tree-optimization/83176
10075 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10078 2018-01-24 Richard Biener <rguenther@suse.de>
10080 PR tree-optimization/82819
10081 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10082 code generating pluses that are no-ops in the target precision.
10084 2018-01-24 Richard Biener <rguenther@suse.de>
10086 PR middle-end/84000
10087 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10089 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10091 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10092 to merge probabilities.
10093 * predict.c (probably_never_executed): Also mark as cold functions
10094 with global 0 profile and guessed local profile.
10095 * profile-count.c (profile_probability::combine_with_count): New
10097 * profile-count.h (profile_probability::operator*,
10098 profile_probability::operator*=, profile_probability::operator/,
10099 profile_probability::operator/=): Reduce precision to adjusted
10100 and set value to guessed on contradictory divisions.
10101 (profile_probability::combine_with_freq): Remove.
10102 (profile_probability::combine_wiht_count): Declare.
10103 (profile_count::force_nonzero):: Set to adjusted.
10104 (profile_count::probability_in):: Set quality to adjusted.
10105 * tree-ssa-tail-merge.c (replace_block_by): Use
10106 combine_with_count.
10108 2018-01-23 Andrew Waterman <andrew@sifive.com>
10109 Jim Wilson <jimw@sifive.com>
10111 * config/riscv/riscv.c (riscv_stack_boundary): New.
10112 (riscv_option_override): Set riscv_stack_boundary. Handle
10113 riscv_preferred_stack_boundary_arg.
10114 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10115 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10116 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10117 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10118 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10119 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10121 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10124 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10125 of struct ix86_frame.
10126 (ix86_expand_epilogue): Likewise. Add a local variable for
10127 the reg_save_offset field in struct ix86_frame.
10129 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10131 PR tree-optimization/82604
10132 * tree-loop-distribution.c (enum partition_kind): New enum item
10133 PKIND_PARTIAL_MEMSET.
10134 (partition_builtin_p): Support above new enum item.
10135 (generate_code_for_partition): Ditto.
10136 (compute_access_range): Differentiate cases that equality can be
10137 proven at all loops, the innermost loops or no loops.
10138 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10139 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10140 (finalize_partitions, distribute_loop): Don't fuse partition of
10141 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10142 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10143 parloop is enabled.
10145 2018-01-23 Martin Liska <mliska@suse.cz>
10147 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10148 order to ignore the predictor.
10149 (PRED_POLYMORPHIC_CALL): Likewise.
10150 (PRED_RECURSIVE_CALL): Likewise.
10152 2018-01-23 Martin Liska <mliska@suse.cz>
10154 * tree-profile.c (tree_profiling): Print function header to
10155 aware reader which function we are working on.
10156 * value-prof.c (gimple_find_values_to_profile): Do not print
10157 not interesting value histograms.
10159 2018-01-23 Martin Liska <mliska@suse.cz>
10161 * profile-count.h (enum profile_quality): Add
10162 profile_uninitialized as the first value. Do not number values
10163 as they are zero based.
10164 (profile_count::verify): Update sanity check.
10165 (profile_probability::verify): Likewise.
10167 2018-01-23 Nathan Sidwell <nathan@acm.org>
10169 * doc/invoke.texi (ffor-scope): Deprecate.
10171 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10173 PR tree-optimization/83510
10174 * domwalk.c (set_all_edges_as_executable): New function.
10175 (dom_walker::dom_walker): Convert bool param
10176 "skip_unreachable_blocks" to enum reachability. Move setup of
10177 edge flags to set_all_edges_as_executable and only do it when
10178 reachability is REACHABLE_BLOCKS.
10179 * domwalk.h (enum dom_walker::reachability): New enum.
10180 (dom_walker::dom_walker): Convert bool param
10181 "skip_unreachable_blocks" to enum reachability.
10182 (set_all_edges_as_executable): New decl.
10183 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10184 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10186 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10187 but converting true to REACHABLE_BLOCKS.
10188 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10190 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10191 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10192 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10194 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10195 if check_all_array_refs will be called.
10197 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10199 * tree.c (selftest::test_location_wrappers): Add more test
10202 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10204 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10205 (selftest::test_bit_in_range): Likewise.
10207 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10210 * doc/sourcebuild.texi (vect_float): Say that the selector
10211 only describes the situation when -funsafe-math-optimizations is on.
10212 (vect_float_strict): Document.
10214 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10216 PR tree-optimization/83965
10217 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10218 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10219 instead of checking only for a reduction.
10220 (vect_recog_widen_sum_pattern): Likewise.
10222 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10224 * predict.c (probably_never_executed): Only use precise profile info.
10225 (compute_function_frequency): Skip after inlining hack since we now
10226 have quality checking.
10228 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10230 * profile-count.h (profile_probability::very_unlikely,
10231 profile_probability::unlikely, profile_probability::even): Set
10232 precision to guessed.
10234 2018-01-23 Richard Biener <rguenther@suse.de>
10236 PR tree-optimization/83963
10237 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10238 Properly terminate dominator walk when crossing the exit edge not
10239 when visiting its source block.
10241 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10244 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10245 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10247 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10249 PR tree-optimization/83957
10250 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10251 semicolon after for body surrounded by braces.
10253 PR tree-optimization/83081
10254 * profile-count.h (profile_probability::split): New method.
10255 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10256 Use profile_probability::split.
10257 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10258 when splitting a single conditional jump into 2.
10260 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10262 PR tree-optimization/69452
10263 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10266 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10268 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10269 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10270 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10272 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10274 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10275 * config/rl78/rl78.md (movdi): New define_expand.
10276 * config/rl78/rl78.c (rl78_split_movdi): New function.
10278 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10281 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10283 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10284 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10285 128-bit to produce an UNSPEC move to get the double word with the
10286 signbit and then a shift directly to do signbit.
10287 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10288 implementation with a new version that just does either a direct
10289 move or a regular move. Move memory interface to separate insns.
10290 Move insns so they are next to the expander.
10291 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10292 with signbit move. Split big and little endian case.
10293 (signbit<mode>2_dm_mem_le): Likewise.
10294 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10295 (signbit<mode>2_dm2): Likewise.
10297 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10299 * config/rl78/rl78.md (anddi3): New define_expand.
10301 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10303 * config/rl78/rl78.md (umindi3): New define_expand.
10305 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10307 * config/rl78/rl78.md (smindi3): New define_expand.
10309 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10311 * config/rl78/rl78.md (smaxdi3): New define_expand.
10313 2018-01-22 Carl Love <cel@us.ibm.com>
10315 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10316 LVX_V1TI): Add macro expansion.
10317 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10318 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10319 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10320 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10321 Change check to determine if the instruction is a byte reversing
10322 entry. Fix typo in comment.
10323 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10324 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10325 Add def_builtin calls for new builtins.
10326 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10327 Add define_insn expansion.
10329 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10331 * config/rl78/rl78.md (umaxdi3): New define_expand.
10333 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10335 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10336 for non-QImode registers.
10338 2018-01-22 Richard Biener <rguenther@suse.de>
10340 PR tree-optimization/83963
10341 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10342 including the loop exit block.
10343 (scop_detection::merge_sese): Likewise.
10344 (scop_detection::add_scop): Do it here instead.
10346 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10348 * doc/sourcebuild.texi (arm_softfloat): Document.
10350 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10353 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10354 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10355 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10357 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10358 David Edelsohn <dje.gcc@gmail.com>
10361 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10362 Change "crset eq" to "crset 2".
10363 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10364 (*call_indirect_aix<mode>_nospec): Likewise.
10365 (*call_value_indirect_aix<mode>_nospec): Likewise.
10366 (*call_indirect_elfv2<mode>_nospec): Likewise.
10367 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10368 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10369 change assembly output from . to $.
10370 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10371 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10372 (*tablejump<mode>_internal1_nospec): Likewise.
10374 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10377 * config/sh/sh_optimize_sett_clrt.cc:
10378 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10380 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10382 PR tree-optimization/83940
10383 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10384 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10385 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10386 use it to pass back the definition type.
10387 (vect_check_store_rhs): Likewise rhs_dt_out.
10388 (vect_build_gather_load_calls): Add a mask_dt argument and use
10389 it instead of a call to vect_is_simple_use.
10390 (vectorizable_store): Update calls to vect_check_load_store_mask
10391 and vect_check_store_rhs. Use the dt returned by the latter instead
10392 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10393 instead of calls to vect_is_simple_use. Pass the scalar rather
10394 than the vector operand to vect_is_simple_use when handling
10395 second and subsequent copies of an rhs value.
10396 (vectorizable_load): Update calls to vect_check_load_store_mask
10397 and vect_build_gather_load_calls. Use the cached mask_dt and
10398 gs_info.offset_dt instead of calls to vect_is_simple_use.
10400 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10402 PR middle-end/83945
10403 * tree-emutls.c: Include gimplify.h.
10404 (lower_emutls_2): New function.
10405 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10406 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10407 it before further processing.
10410 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10411 UINTVAL (trueop1) instead of INTVAL (op1).
10413 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10417 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10418 INCOMING_FRAME_SP_OFFSET if not defined.
10419 (scan_trace): Add ENTRY argument. If true and
10420 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10421 emit a note to adjust the CFA offset.
10422 (create_cfi_notes): Adjust scan_trace callers.
10423 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10424 INCOMING_FRAME_SP_OFFSET in the CIE.
10425 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10426 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10428 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10429 * doc/tm.texi: Regenerated.
10431 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10433 PR rtl-optimization/83147
10434 * lra-constraints.c (remove_inheritance_pseudos): Use
10435 lra_substitute_pseudo_within_insn.
10437 2018-01-19 Tom de Vries <tom@codesourcery.com>
10438 Cesar Philippidis <cesar@codesourcery.com>
10441 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10443 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10446 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10447 spaces for function labels.
10449 2018-01-19 Martin Liska <mliska@suse.cz>
10451 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10452 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10453 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10454 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10455 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10456 (PRED_CONST_RETURN): Change from 69 to 65.
10457 (PRED_NULL_RETURN): Change from 91 to 71.
10458 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10459 (PRED_LOOP_GUARD): Change from 66 to 73.
10461 2018-01-19 Martin Liska <mliska@suse.cz>
10463 * predict.c (predict_insn_def): Add new assert.
10464 (struct branch_predictor): Change type to signed integer.
10465 (test_prediction_value_range): Amend test to cover
10466 PROB_UNINITIALIZED.
10467 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10468 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10469 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10470 (PRED_LOOP_IV_COMPARE): Likewise.
10471 * predict.h (PROB_UNINITIALIZED): Define new constant.
10473 2018-01-19 Martin Liska <mliska@suse.cz>
10475 * predict.c (dump_prediction): Add new format for
10476 analyze_brprob.py script which is enabled with -details
10478 * profile-count.h (precise_p): New function.
10480 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10482 PR tree-optimization/83922
10483 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10484 there are no statements that need masking.
10485 (vect_active_double_reduction_p): New function.
10486 (vect_analyze_loop_operations): Use it when handling phis that
10487 are not in the loop header.
10489 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10491 PR tree-optimization/83914
10492 * tree-vect-loop.c (vectorizable_induction): Don't convert
10493 init_expr or apply the peeling adjustment for inductions
10494 that are nested within the vectorized loop.
10496 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10498 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10501 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10505 * function.h (gimplify_parameters): Add gimple_seq * argument.
10506 * function.c: Include gimple.h and options.h.
10507 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10508 for the added local temporaries if needed.
10509 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10510 if there are any parameter cleanups, wrap whole body into a
10511 try/finally with the cleanups.
10513 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10516 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10517 Use GET_MODE_CLASS for scalar floating point.
10519 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10523 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10524 Fix call of call_cgraph_insertion_hooks.
10526 2018-01-18 Martin Sebor <msebor@redhat.com>
10528 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10530 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10533 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10536 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10539 * common.opt: (-ffile-prefix-map): New option.
10540 * opts.c (common_handle_option): Defer it.
10541 * opts-global.c (handle_common_deferred_options): Handle it.
10542 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10543 * file-prefix-map.h: New file.
10544 (remap_debug_filename, add_debug_prefix_map): ...here.
10545 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10546 * final.c (debug_prefix_map, add_debug_prefix_map
10547 remap_debug_filename): Move to...
10548 * file-prefix-map.c: New file.
10549 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10550 generalize, get rid of alloca(), use strrchr() instead of strchr().
10551 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10552 Implement in terms of add_prefix_map().
10553 (remap_macro_filename, remap_debug_filename): Implement in term of
10555 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10556 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10557 * dbxout.c: Include file-prefix-map.h.
10558 * varasm.c: Likewise.
10559 * vmsdbgout.c: Likewise.
10560 * xcoffout.c: Likewise.
10561 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10562 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10563 * doc/invoke.texi (-ffile-prefix-map): Document.
10564 (-fdebug-prefix-map): Update description.
10566 2018-01-18 Martin Liska <mliska@suse.cz>
10568 * config/i386/i386.c (indirect_thunk_name): Document that also
10570 (output_indirect_thunk): Document why both instructions
10571 (pause and lfence) are generated.
10573 2018-01-18 Richard Biener <rguenther@suse.de>
10575 PR tree-optimization/83887
10576 * graphite-scop-detection.c
10577 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10578 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10579 (scop_detection::merge_sese): Re-implement with a flood-fill
10580 algorithm that properly finds a SESE region if it exists.
10582 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10585 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10586 pointer_diff optimizations use view_convert instead of convert.
10588 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10590 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10591 Generate different code for -mno-speculate-indirect-jumps.
10592 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10593 (*call_indirect_aix<mode>): Disable for
10594 -mno-speculate-indirect-jumps.
10595 (*call_indirect_aix<mode>_nospec): New define_insn.
10596 (*call_value_indirect_aix<mode>): Disable for
10597 -mno-speculate-indirect-jumps.
10598 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10599 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10600 -mno-speculate-indirect-jumps.
10601 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10603 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10605 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10606 long double type, set the flags for noting the default long double
10607 type, even if we don't pass or return a long double type.
10609 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10612 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10615 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10617 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10618 support for merge[hl].
10619 (fold_mergehl_helper): New helper function.
10620 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10621 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10622 (altivec_vmrglw_direct): Add xxmrglw insn.
10624 2018-01-17 Andrew Waterman <andrew@sifive.com>
10626 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10627 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10629 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10632 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10633 call the lto_location_cache before reading the
10634 DECL_SOURCE_LOCATION of the types.
10636 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10637 Richard Sandiford <richard.sandiford@linaro.org>
10639 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10640 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10641 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10642 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10643 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10645 * config/aarch64/constraints.md (aarch64_movti_operand):
10647 * config/aarch64/predicates.md (Uti): Add new constraint.
10649 2018-01-17 Carl Love <cel@us.ibm.com>
10651 * config/rs6000/vsx.md (define_expand xl_len_r,
10652 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10653 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10655 (define_expand, define_insn): Move the shift left from the
10656 define_insn to the define_expand for lxvl and stxvl instructions.
10657 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10658 and XL_LEN_R definitions to PURE.
10660 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10662 * config/i386/i386.c (indirect_thunk_name): Declare regno
10663 as unsigned int. Compare regno with INVALID_REGNUM.
10664 (output_indirect_thunk): Ditto.
10665 (output_indirect_thunk_function): Ditto.
10666 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10667 in the call to output_indirect_thunk_function.
10669 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10671 PR middle-end/83884
10672 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10673 rather than the size of inner_type to determine the stack slot size
10674 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10676 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10679 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10682 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10684 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10685 endian Linux systems to optionally enable multilibs for selecting
10686 the long double type if the user configured an explicit type.
10687 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10688 have no long double multilibs if not defined.
10689 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10690 warn if the user used -mabi={ieee,ibm}longdouble and we built
10691 multilibs for long double.
10692 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10693 appropriate multilib option.
10694 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10696 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10697 for building long double multilibs.
10698 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10700 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10702 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10705 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10707 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10710 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10713 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10716 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10718 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10719 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10721 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10723 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10724 different rtl trees depending on TARGET_64BIT.
10725 (rs6000_gen_lvx): Likewise.
10727 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10729 * config/visium/visium.md (nop): Tweak comment.
10730 (hazard_nop): Likewise.
10732 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10734 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10735 -mspeculate-indirect-jumps.
10736 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10737 for -mno-speculate-indirect-jumps.
10738 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10739 (*call_value_indirect_elfv2<mode>): Disable for
10740 -mno-speculate-indirect-jumps.
10741 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10742 (indirect_jump): Emit different RTL for
10743 -mno-speculate-indirect-jumps.
10744 (*indirect_jump<mode>): Disable for
10745 -mno-speculate-indirect-jumps.
10746 (*indirect_jump<mode>_nospec): New define_insn.
10747 (tablejump): Emit different RTL for
10748 -mno-speculate-indirect-jumps.
10749 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10750 (tablejumpsi_nospec): New define_expand.
10751 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10752 (tablejumpdi_nospec): New define_expand.
10753 (*tablejump<mode>_internal1): Disable for
10754 -mno-speculate-indirect-jumps.
10755 (*tablejump<mode>_internal1_nospec): New define_insn.
10756 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10759 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10761 * caller-save.c (insert_save): Drop unnecessary parameter. All
10764 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10765 Richard Biener <rguenth@suse.de>
10768 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10769 return early, inline manually is_gimple_sizepos. Make sure if we
10770 call gimplify_expr we don't end up with a gimple constant.
10771 * tree.c (variably_modified_type_p): Don't return true for
10772 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10773 * gimplify.h (is_gimple_sizepos): Remove.
10775 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10777 PR tree-optimization/83857
10778 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10779 vectorizable_live_operation for pure SLP statements.
10780 (vectorizable_live_operation): Handle PHIs.
10782 2018-01-16 Richard Biener <rguenther@suse.de>
10784 PR tree-optimization/83867
10785 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10786 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10788 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10791 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10792 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10793 If off is not INTEGER_CST, issue a may not be aligned warning
10794 rather than isn't aligned. Use isn%'t rather than isn't.
10795 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10797 <case MULT_EXPR>: Improve the case when bottom and one of the
10798 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10799 operand, in that case check if the other operand is multiple of
10800 bottom divided by the INTEGER_CST operand.
10802 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10805 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10806 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10807 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10808 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10809 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10810 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10811 (pa_function_arg_size): New function.
10813 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10815 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10816 in a separate statement.
10818 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10820 PR tree-optimization/83847
10821 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10822 group gathers and scatters.
10824 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10826 PR rtl-optimization/86620
10827 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10829 PR rtl-optimization/83213
10830 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10831 to last if both are JUMP_INSNs.
10833 PR tree-optimization/83843
10834 * gimple-ssa-store-merging.c
10835 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10836 store_immediate_info for bswap/nop orig_stores.
10838 2018-01-15 Andrew Waterman <andrew@sifive.com>
10840 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10842 <UDIV>: Increase cost if !TARGET_DIV.
10844 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10846 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10847 (define_attr "cr_logical_3op"): New.
10848 (cceq_ior_compare): Adjust.
10849 (cceq_ior_compare_complement): Adjust.
10850 (*cceq_rev_compare): Adjust.
10851 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10852 (is_cracked_insn): Adjust.
10853 (insn_must_be_first_in_group): Adjust.
10854 * config/rs6000/40x.md: Adjust.
10855 * config/rs6000/440.md: Adjust.
10856 * config/rs6000/476.md: Adjust.
10857 * config/rs6000/601.md: Adjust.
10858 * config/rs6000/603.md: Adjust.
10859 * config/rs6000/6xx.md: Adjust.
10860 * config/rs6000/7450.md: Adjust.
10861 * config/rs6000/7xx.md: Adjust.
10862 * config/rs6000/8540.md: Adjust.
10863 * config/rs6000/cell.md: Adjust.
10864 * config/rs6000/e300c2c3.md: Adjust.
10865 * config/rs6000/e500mc.md: Adjust.
10866 * config/rs6000/e500mc64.md: Adjust.
10867 * config/rs6000/e5500.md: Adjust.
10868 * config/rs6000/e6500.md: Adjust.
10869 * config/rs6000/mpc.md: Adjust.
10870 * config/rs6000/power4.md: Adjust.
10871 * config/rs6000/power5.md: Adjust.
10872 * config/rs6000/power6.md: Adjust.
10873 * config/rs6000/power7.md: Adjust.
10874 * config/rs6000/power8.md: Adjust.
10875 * config/rs6000/power9.md: Adjust.
10876 * config/rs6000/rs64.md: Adjust.
10877 * config/rs6000/titan.md: Adjust.
10879 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10881 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10882 ix86_indirect_branch_register logic.
10884 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10886 * config/i386/constraints.md (Bs): Update
10887 ix86_indirect_branch_register check. Don't check
10888 ix86_indirect_branch_register with GOT_memory_operand.
10890 * config/i386/predicates.md (GOT_memory_operand): Don't check
10891 ix86_indirect_branch_register here.
10892 (GOT32_symbol_operand): Likewise.
10894 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10896 * config/i386/predicates.md (constant_call_address_operand):
10897 Rewrite ix86_indirect_branch_register logic.
10898 (sibcall_insn_operand): Likewise.
10900 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10902 * config/i386/constraints.md (Bs): Replace
10903 ix86_indirect_branch_thunk_register with
10904 ix86_indirect_branch_register.
10906 * config/i386/i386.md (indirect_jump): Likewise.
10907 (tablejump): Likewise.
10908 (*sibcall_memory): Likewise.
10909 (*sibcall_value_memory): Likewise.
10910 Peepholes of indirect call and jump via memory: Likewise.
10911 * config/i386/i386.opt: Likewise.
10912 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10913 (GOT_memory_operand): Likewise.
10914 (call_insn_operand): Likewise.
10915 (sibcall_insn_operand): Likewise.
10916 (GOT32_symbol_operand): Likewise.
10918 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10920 PR middle-end/83837
10921 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10922 type rather than type addr's type points to.
10923 (expand_omp_atomic_mutex): Likewise.
10924 (expand_omp_atomic): Likewise.
10926 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10929 * config/i386/i386.c (output_indirect_thunk_function): Use
10930 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10931 for __x86_return_thunk.
10933 2018-01-15 Richard Biener <rguenther@suse.de>
10935 PR middle-end/83850
10936 * expmed.c (extract_bit_field_1): Fix typo.
10938 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10941 * config/arm/iterators.md (VF): New mode iterator.
10942 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10943 Remove integer-related logic from pattern.
10944 (neon_vabd<mode>_3): Likewise.
10946 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10948 PR middle-end/82694
10949 * common.opt (fstrict-overflow): No longer an alias.
10950 (fwrapv-pointer): New option.
10951 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10952 also for pointer types based on flag_wrapv_pointer.
10953 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10954 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10955 opts->x_flag_wrapv got set.
10956 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10957 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10958 POINTER_TYPE_OVERFLOW_UNDEFINED.
10959 * match.pd: Likewise in address comparison pattern.
10960 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10962 2018-01-15 Richard Biener <rguenther@suse.de>
10965 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10966 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10967 Reset type names to their identifier if their TYPE_DECL doesn't
10968 have linkage (and thus is used for ODR and devirt).
10969 (save_debug_info_for_decl): Remove.
10970 (save_debug_info_for_type): Likewise.
10971 (add_tree_to_fld_list): Adjust.
10972 * tree-pretty-print.c (dump_generic_node): Make dumping of
10973 type names more robust.
10975 2018-01-15 Richard Biener <rguenther@suse.de>
10977 * BASE-VER: Bump to 8.0.1.
10979 2018-01-14 Martin Sebor <msebor@redhat.com>
10982 * builtins.c (check_access): Avoid warning when the no-warning bit
10985 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10987 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10988 * ira-color (allocno_hard_regs_compare): Likewise.
10990 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10993 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10994 Use .pushsection/.popsection.
10996 2018-01-14 Martin Sebor <msebor@redhat.com>
10999 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
11001 2018-01-14 Jakub Jelinek <jakub@redhat.com>
11003 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
11004 entry from extra_headers.
11005 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
11006 extra_headers, make the list bitwise identical to the i?86-*-* one.
11008 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11010 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
11011 -mcmodel=large with -mindirect-branch=thunk,
11012 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
11013 -mfunction-return=thunk-extern.
11014 * doc/invoke.texi: Document -mcmodel=large is incompatible with
11015 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
11016 -mfunction-return=thunk and -mfunction-return=thunk-extern.
11018 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11020 * config/i386/i386.c (print_reg): Print the name of the full
11021 integer register without '%'.
11022 (ix86_print_operand): Handle 'V'.
11023 * doc/extend.texi: Document 'V' modifier.
11025 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11027 * config/i386/constraints.md (Bs): Disallow memory operand for
11028 -mindirect-branch-register.
11030 * config/i386/predicates.md (indirect_branch_operand): Likewise.
11031 (GOT_memory_operand): Likewise.
11032 (call_insn_operand): Likewise.
11033 (sibcall_insn_operand): Likewise.
11034 (GOT32_symbol_operand): Likewise.
11035 * config/i386/i386.md (indirect_jump): Call convert_memory_address
11036 for -mindirect-branch-register.
11037 (tablejump): Likewise.
11038 (*sibcall_memory): Likewise.
11039 (*sibcall_value_memory): Likewise.
11040 Disallow peepholes of indirect call and jump via memory for
11041 -mindirect-branch-register.
11042 (*call_pop): Replace m with Bw.
11043 (*call_value_pop): Likewise.
11044 (*sibcall_pop_memory): Replace m with Bs.
11045 * config/i386/i386.opt (mindirect-branch-register): New option.
11046 * doc/invoke.texi: Document -mindirect-branch-register option.
11048 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11050 * config/i386/i386-protos.h (ix86_output_function_return): New.
11051 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11052 set function_return_type.
11053 (indirect_thunk_name): Add ret_p to indicate thunk for function
11055 (output_indirect_thunk_function): Pass false to
11056 indirect_thunk_name.
11057 (ix86_output_indirect_branch_via_reg): Likewise.
11058 (ix86_output_indirect_branch_via_push): Likewise.
11059 (output_indirect_thunk_function): Create alias for function
11060 return thunk if regno < 0.
11061 (ix86_output_function_return): New function.
11062 (ix86_handle_fndecl_attribute): Handle function_return.
11063 (ix86_attribute_table): Add function_return.
11064 * config/i386/i386.h (machine_function): Add
11065 function_return_type.
11066 * config/i386/i386.md (simple_return_internal): Use
11067 ix86_output_function_return.
11068 (simple_return_internal_long): Likewise.
11069 * config/i386/i386.opt (mfunction-return=): New option.
11070 (indirect_branch): Mention -mfunction-return=.
11071 * doc/extend.texi: Document function_return function attribute.
11072 * doc/invoke.texi: Document -mfunction-return= option.
11074 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11076 * config/i386/i386-opts.h (indirect_branch): New.
11077 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11078 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11079 with local indirect jump when converting indirect call and jump.
11080 (ix86_set_indirect_branch_type): New.
11081 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11082 (indirectlabelno): New.
11083 (indirect_thunk_needed): Likewise.
11084 (indirect_thunk_bnd_needed): Likewise.
11085 (indirect_thunks_used): Likewise.
11086 (indirect_thunks_bnd_used): Likewise.
11087 (INDIRECT_LABEL): Likewise.
11088 (indirect_thunk_name): Likewise.
11089 (output_indirect_thunk): Likewise.
11090 (output_indirect_thunk_function): Likewise.
11091 (ix86_output_indirect_branch_via_reg): Likewise.
11092 (ix86_output_indirect_branch_via_push): Likewise.
11093 (ix86_output_indirect_branch): Likewise.
11094 (ix86_output_indirect_jmp): Likewise.
11095 (ix86_code_end): Call output_indirect_thunk_function if needed.
11096 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11098 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11099 (ix86_attribute_table): Add indirect_branch.
11100 * config/i386/i386.h (machine_function): Add indirect_branch_type
11101 and has_local_indirect_jump.
11102 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11104 (tablejump): Likewise.
11105 (*indirect_jump): Use ix86_output_indirect_jmp.
11106 (*tablejump_1): Likewise.
11107 (simple_return_indirect_internal): Likewise.
11108 * config/i386/i386.opt (mindirect-branch=): New option.
11109 (indirect_branch): New.
11112 (thunk-inline): Likewise.
11113 (thunk-extern): Likewise.
11114 * doc/extend.texi: Document indirect_branch function attribute.
11115 * doc/invoke.texi: Document -mindirect-branch= option.
11117 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11120 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11122 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11124 * ipa-inline.c (want_inline_small_function_p): Return false if
11125 inlining has already failed with CIF_FINAL_ERROR.
11126 (update_caller_keys): Call want_inline_small_function_p before
11128 (update_callee_keys): Likewise.
11130 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11132 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11134 (rs6000_quadword_masked_address_p): Likewise.
11135 (quad_aligned_load_p): Likewise.
11136 (quad_aligned_store_p): Likewise.
11137 (const_load_sequence_p): Add comment to describe the outer-most loop.
11138 (mimic_memory_attributes_and_flags): New function.
11139 (rs6000_gen_stvx): Likewise.
11140 (replace_swapped_aligned_store): Likewise.
11141 (rs6000_gen_lvx): Likewise.
11142 (replace_swapped_aligned_load): Likewise.
11143 (replace_swapped_load_constant): Capitalize argument name in
11144 comment describing this function.
11145 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11146 and stores that access quad-word aligned addresses and replace
11147 with stvx or lvx instructions when appropriate.
11148 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11149 New function prototype.
11150 (rs6000_quadword_masked_address_p): Likewise.
11151 (rs6000_gen_lvx): Likewise.
11152 (rs6000_gen_stvx): Likewise.
11153 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11154 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11155 when memory address is aligned.
11156 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11157 this split to select lvx instruction when memory address is aligned.
11158 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11159 instruction when memory address is aligned.
11160 (*vsx_le_perm_load_v16qi): Likewise.
11161 (four unnamed splitters): Modify to select the stvx instruction
11162 when memory is aligned.
11164 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11166 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11167 which appears in the queue multiple times.
11169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11170 Alan Hayward <alan.hayward@arm.com>
11171 David Sherwood <david.sherwood@arm.com>
11173 * tree-vectorizer.h (vec_lower_bound): New structure.
11174 (_loop_vec_info): Add check_nonzero and lower_bounds.
11175 (LOOP_VINFO_CHECK_NONZERO): New macro.
11176 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11177 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11178 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11179 fields. Make seg_len the distance travelled, not including the
11181 (dr_direction_indicator): Declare.
11182 (dr_zero_step_indicator): Likewise.
11183 (dr_known_forward_stride_p): Likewise.
11184 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11186 (runtime_alias_check_p): Allow runtime alias checks with
11188 (operator ==): Compare access_size and align.
11189 (prune_runtime_alias_test_list): Rework for new distinction between
11190 the access_size and seg_len.
11191 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11193 (get_segment_min_max): New function.
11194 (create_intersect_range_checks): Use it.
11195 (dr_step_indicator): New function.
11196 (dr_direction_indicator): Likewise.
11197 (dr_zero_step_indicator): Likewise.
11198 (dr_known_forward_stride_p): Likewise.
11199 * tree-loop-distribution.c (data_ref_segment_size): Return
11200 DR_STEP * (niters - 1).
11201 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11203 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11204 (vect_preserves_scalar_order_p): New function, split out from...
11205 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11206 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11207 (vect_vfa_access_size): New function.
11208 (vect_vfa_align): Likewise.
11209 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11210 (dump_lower_bound): New function.
11211 (vect_check_lower_bound): Likewise.
11212 (vect_small_gap_p): Likewise.
11213 (vectorizable_with_step_bound_p): Likewise.
11214 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11215 depencies if the vectorization factor is 1. Convert the checks
11216 for nonzero steps into checks on the bounds of DR_STEP. Try using
11217 a bunds check for variable steps if the minimum required step is
11218 relatively small. Update calls to the dr_with_seg_len
11219 constructor and to vect_compile_time_alias.
11220 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11222 (vect_loop_versioning): Call it.
11223 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11225 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11228 Alan Hayward <alan.hayward@arm.com>
11229 David Sherwood <david.sherwood@arm.com>
11231 * doc/sourcebuild.texi (vect_scatter_store): Document.
11232 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11234 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11236 * genopinit.c (main): Add supports_vec_scatter_store and
11237 supports_vec_scatter_store_cached to target_optabs.
11238 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11239 IFN_MASK_SCATTER_STORE.
11240 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11242 * internal-fn.h (internal_store_fn_p): Declare.
11243 (internal_fn_stored_value_index): Likewise.
11244 * internal-fn.c (scatter_store_direct): New macro.
11245 (expand_scatter_store_optab_fn): New function.
11246 (direct_scatter_store_optab_supported_p): New macro.
11247 (internal_store_fn_p): New function.
11248 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11249 IFN_MASK_SCATTER_STORE.
11250 (internal_fn_mask_index): Likewise.
11251 (internal_fn_stored_value_index): New function.
11252 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11253 for scatter stores.
11254 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11255 * optabs-query.c (supports_vec_scatter_store_p): New function.
11256 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11257 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11258 true for scatter stores.
11259 (vect_gather_scatter_fn_p): Handle scatter stores too.
11260 (vect_check_gather_scatter): Consider using scatter stores if
11261 supports_vec_scatter_store_p.
11262 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11263 scatter stores too.
11264 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11265 internal_fn_stored_value_index.
11266 (check_load_store_masking): Handle scatter stores too.
11267 (vect_get_store_rhs): Make public.
11268 (vectorizable_call): Use internal_store_fn_p.
11269 (vectorizable_store): Handle scatter store internal functions.
11270 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11271 when deciding whether the end of the group has been reached.
11272 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11273 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11274 (mask_scatter_store<mode>): New insns.
11276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11277 Alan Hayward <alan.hayward@arm.com>
11278 David Sherwood <david.sherwood@arm.com>
11280 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11281 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11282 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11284 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11285 Use vect_truncate_gather_scatter_offset if we can't treat the
11286 operation as a normal gather load or scatter store.
11287 (get_group_load_store_type): Take the gather_scatter_info
11288 as argument. Try using a gather load or scatter store for
11289 single-element groups.
11290 (get_load_store_type): Update calls to get_group_load_store_type
11291 and vect_use_strided_gather_scatters_p.
11293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11294 Alan Hayward <alan.hayward@arm.com>
11295 David Sherwood <david.sherwood@arm.com>
11297 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11298 optional tree argument.
11299 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11301 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11302 but continue to use the current value as a fallback.
11303 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11304 to compare the updates.
11305 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11306 (get_load_store_type): Use it when handling a strided access.
11307 (vect_get_strided_load_store_ops): New function.
11308 (vect_get_data_ptr_increment): Likewise.
11309 (vectorizable_load): Handle strided gather loads. Always pass
11310 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11313 Alan Hayward <alan.hayward@arm.com>
11314 David Sherwood <david.sherwood@arm.com>
11316 * doc/md.texi (gather_load@var{m}): Document.
11317 (mask_gather_load@var{m}): Likewise.
11318 * genopinit.c (main): Add supports_vec_gather_load and
11319 supports_vec_gather_load_cached to target_optabs.
11320 * optabs-tree.c (init_tree_optimization_optabs): Use
11321 ggc_cleared_alloc to allocate target_optabs.
11322 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11323 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11325 * internal-fn.h (internal_load_fn_p): Declare.
11326 (internal_gather_scatter_fn_p): Likewise.
11327 (internal_fn_mask_index): Likewise.
11328 (internal_gather_scatter_fn_supported_p): Likewise.
11329 * internal-fn.c (gather_load_direct): New macro.
11330 (expand_gather_load_optab_fn): New function.
11331 (direct_gather_load_optab_supported_p): New macro.
11332 (direct_internal_fn_optab): New function.
11333 (internal_load_fn_p): Likewise.
11334 (internal_gather_scatter_fn_p): Likewise.
11335 (internal_fn_mask_index): Likewise.
11336 (internal_gather_scatter_fn_supported_p): Likewise.
11337 * optabs-query.c (supports_at_least_one_mode_p): New function.
11338 (supports_vec_gather_load_p): Likewise.
11339 * optabs-query.h (supports_vec_gather_load_p): Declare.
11340 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11341 and memory_type field.
11342 (NUM_PATTERNS): Bump to 15.
11343 * tree-vect-data-refs.c: Include internal-fn.h.
11344 (vect_gather_scatter_fn_p): New function.
11345 (vect_describe_gather_scatter_call): Likewise.
11346 (vect_check_gather_scatter): Try using internal functions for
11347 gather loads. Recognize existing calls to a gather load function.
11348 (vect_analyze_data_refs): Consider using gather loads if
11349 supports_vec_gather_load_p.
11350 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11351 (vect_get_gather_scatter_offset_type): Likewise.
11352 (vect_convert_mask_for_vectype): Likewise.
11353 (vect_add_conversion_to_patterm): Likewise.
11354 (vect_try_gather_scatter_pattern): Likewise.
11355 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11356 (vect_vect_recog_func_ptrs): Add it.
11357 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11358 internal_fn_mask_index and internal_gather_scatter_fn_p.
11359 (check_load_store_masking): Take the gather_scatter_info as an
11360 argument and handle gather loads.
11361 (vect_get_gather_scatter_ops): New function.
11362 (vectorizable_call): Check internal_load_fn_p.
11363 (vectorizable_load): Likewise. Handle gather load internal
11365 (vectorizable_store): Update call to check_load_store_masking.
11366 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11367 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11368 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11369 (aarch64_gather_scale_operand_d): New predicates.
11370 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11371 (mask_gather_load<mode>): New insns.
11373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11374 Alan Hayward <alan.hayward@arm.com>
11375 David Sherwood <david.sherwood@arm.com>
11377 * optabs.def (fold_left_plus_optab): New optab.
11378 * doc/md.texi (fold_left_plus_@var{m}): Document.
11379 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11380 * internal-fn.c (fold_left_direct): Define.
11381 (expand_fold_left_optab_fn): Likewise.
11382 (direct_fold_left_optab_supported_p): Likewise.
11383 * fold-const-call.c (fold_const_fold_left): New function.
11384 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11385 * tree-parloops.c (valid_reduction_p): New function.
11386 (gather_scalar_reductions): Use it.
11387 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11388 (vect_finish_replace_stmt): Declare.
11389 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11390 (needs_fold_left_reduction_p): New function, split out from...
11391 (vect_is_simple_reduction): ...here. Accept reductions that
11392 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11393 (vect_force_simple_reduction): Also store the reduction type in
11394 the assignment's STMT_VINFO_REDUC_TYPE.
11395 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11396 (merge_with_identity): New function.
11397 (vect_expand_fold_left): Likewise.
11398 (vectorize_fold_left_reduction): Likewise.
11399 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11400 scalar phi in place for it. Check for target support and reject
11401 cases that would reassociate the operation. Defer the transform
11402 phase to vectorize_fold_left_reduction.
11403 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11404 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11405 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11409 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11410 call to ifc_temp_var.
11412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11413 Alan Hayward <alan.hayward@arm.com>
11414 David Sherwood <david.sherwood@arm.com>
11416 * target.def (legitimize_address_displacement): Take the original
11417 offset as a poly_int.
11418 * targhooks.h (default_legitimize_address_displacement): Update
11420 * targhooks.c (default_legitimize_address_displacement): Likewise.
11421 * doc/tm.texi: Regenerate.
11422 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11423 as an argument, moving assert of ad->disp == ad->disp_term to...
11424 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11425 Try calling targetm.legitimize_address_displacement before expanding
11426 the address rather than afterwards, and adjust for the new interface.
11427 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11428 Match the new hook interface. Handle SVE addresses.
11429 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11430 new hook interface.
11432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11434 * Makefile.in (OBJS): Add early-remat.o.
11435 * target.def (select_early_remat_modes): New hook.
11436 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11437 * doc/tm.texi: Regenerate.
11438 * targhooks.h (default_select_early_remat_modes): Declare.
11439 * targhooks.c (default_select_early_remat_modes): New function.
11440 * timevar.def (TV_EARLY_REMAT): New timevar.
11441 * passes.def (pass_early_remat): New pass.
11442 * tree-pass.h (make_pass_early_remat): Declare.
11443 * early-remat.c: New file.
11444 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11446 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11449 Alan Hayward <alan.hayward@arm.com>
11450 David Sherwood <david.sherwood@arm.com>
11452 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11453 vfm1 with a bound_epilog parameter.
11454 (vect_do_peeling): Update calls accordingly, and move the prologue
11455 call earlier in the function. Treat the base bound_epilog as 0 for
11456 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11457 this base when peeling for gaps.
11458 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11459 with fully-masked loops.
11460 (vect_estimate_min_profitable_iters): Handle the single peeled
11461 iteration in that case.
11463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11464 Alan Hayward <alan.hayward@arm.com>
11465 David Sherwood <david.sherwood@arm.com>
11467 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11468 single-element interleaving even if the size is not a power of 2.
11469 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11470 accesses for single-element interleaving if the group size is
11473 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11474 Alan Hayward <alan.hayward@arm.com>
11475 David Sherwood <david.sherwood@arm.com>
11477 * doc/md.texi (fold_extract_last_@var{m}): Document.
11478 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11479 * optabs.def (fold_extract_last_optab): New optab.
11480 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11481 * internal-fn.c (fold_extract_direct): New macro.
11482 (expand_fold_extract_optab_fn): Likewise.
11483 (direct_fold_extract_optab_supported_p): Likewise.
11484 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11485 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11486 EXTRACT_LAST_REDUCTION.
11487 (get_initial_def_for_reduction): Do not create an initial vector
11488 for EXTRACT_LAST_REDUCTION reductions.
11489 (vectorizable_reduction): Leave the scalar phi in place for
11490 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11491 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11492 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11493 transform phase to vectorizable_condition.
11494 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11496 (vect_finish_stmt_generation): ...here.
11497 (vect_finish_replace_stmt): New function.
11498 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11499 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11501 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11504 Alan Hayward <alan.hayward@arm.com>
11505 David Sherwood <david.sherwood@arm.com>
11507 * doc/md.texi (extract_last_@var{m}): Document.
11508 * optabs.def (extract_last_optab): New optab.
11509 * internal-fn.def (EXTRACT_LAST): New internal function.
11510 * internal-fn.c (cond_unary_direct): New macro.
11511 (expand_cond_unary_optab_fn): Likewise.
11512 (direct_cond_unary_optab_supported_p): Likewise.
11513 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11514 loops using EXTRACT_LAST.
11515 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11516 (extract_last_<mode>): ...this optab.
11517 (vec_extract<mode><Vel>): Update accordingly.
11519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11520 Alan Hayward <alan.hayward@arm.com>
11521 David Sherwood <david.sherwood@arm.com>
11523 * target.def (empty_mask_is_expensive): New hook.
11524 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11525 * doc/tm.texi: Regenerate.
11526 * targhooks.h (default_empty_mask_is_expensive): Declare.
11527 * targhooks.c (default_empty_mask_is_expensive): New function.
11528 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11529 if the target says that empty masks are expensive.
11530 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11532 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11535 Alan Hayward <alan.hayward@arm.com>
11536 David Sherwood <david.sherwood@arm.com>
11538 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11539 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11540 (vect_use_loop_mask_for_alignment_p): New function.
11541 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11542 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11543 niters_skip argument. Make sure that the first niters_skip elements
11544 of the first iteration are inactive.
11545 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11546 Update call to vect_set_loop_masks_directly.
11547 (get_misalign_in_elems): New function, split out from...
11548 (vect_gen_prolog_loop_niters): ...here.
11549 (vect_update_init_of_dr): Take a code argument that specifies whether
11550 the adjustment should be added or subtracted.
11551 (vect_update_init_of_drs): Likewise.
11552 (vect_prepare_for_masked_peels): New function.
11553 (vect_do_peeling): Skip prologue peeling if we're using a mask
11554 instead. Update call to vect_update_inits_of_drs.
11555 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11557 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11558 alignment. Do not include the number of peeled iterations in
11559 the minimum threshold in that case.
11560 (vectorizable_induction): Adjust the start value down by
11561 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11562 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11563 Take the number of skipped iterations into account when calculating
11565 * tree-vect-stmts.c (vect_gen_while_not): New function.
11567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11568 Alan Hayward <alan.hayward@arm.com>
11569 David Sherwood <david.sherwood@arm.com>
11571 * doc/sourcebuild.texi (vect_fully_masked): Document.
11572 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11573 default value to 0.
11574 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11576 (vect_analyze_loop_2): ...here. Don't check the vectorization
11577 factor against the number of loop iterations if the loop is
11580 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11581 Alan Hayward <alan.hayward@arm.com>
11582 David Sherwood <david.sherwood@arm.com>
11584 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11585 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11586 (dump_groups): Update accordingly.
11587 (iv_use::mem_type): New member variable.
11588 (address_p): New function.
11589 (record_use): Add a mem_type argument and initialize the new
11591 (record_group_use): Add a mem_type argument. Use address_p.
11592 Remove obsolete null checks of base_object. Update call to record_use.
11593 (find_interesting_uses_op): Update call to record_group_use.
11594 (find_interesting_uses_cond): Likewise.
11595 (find_interesting_uses_address): Likewise.
11596 (get_mem_type_for_internal_fn): New function.
11597 (find_address_like_use): Likewise.
11598 (find_interesting_uses_stmt): Try find_address_like_use before
11599 calling find_interesting_uses_op.
11600 (addr_offset_valid_p): Use the iv mem_type field as the type
11601 of the addressed memory.
11602 (add_autoinc_candidates): Likewise.
11603 (get_address_cost): Likewise.
11604 (split_small_address_groups_p): Use address_p.
11605 (split_address_groups): Likewise.
11606 (add_iv_candidate_for_use): Likewise.
11607 (autoinc_possible_for_pair): Likewise.
11608 (rewrite_groups): Likewise.
11609 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11610 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11611 (get_alias_ptr_type_for_ptr_address): New function.
11612 (rewrite_use_address): Rewrite address uses in calls that were
11613 identified by find_address_like_use.
11615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11616 Alan Hayward <alan.hayward@arm.com>
11617 David Sherwood <david.sherwood@arm.com>
11619 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11621 * gimple-expr.h (is_gimple_addressable: Likewise.
11622 * gimple-expr.c (is_gimple_address): Likewise.
11623 * internal-fn.c (expand_call_mem_ref): New function.
11624 (expand_mask_load_optab_fn): Use it.
11625 (expand_mask_store_optab_fn): Likewise.
11627 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11628 Alan Hayward <alan.hayward@arm.com>
11629 David Sherwood <david.sherwood@arm.com>
11631 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11632 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11633 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11634 (cond_umax@var{mode}): Document.
11635 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11636 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11637 (cond_umin_optab, cond_umax_optab): New optabs.
11638 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11639 (COND_IOR, COND_XOR): New internal functions.
11640 * internal-fn.h (get_conditional_internal_fn): Declare.
11641 * internal-fn.c (cond_binary_direct): New macro.
11642 (expand_cond_binary_optab_fn): Likewise.
11643 (direct_cond_binary_optab_supported_p): Likewise.
11644 (get_conditional_internal_fn): New function.
11645 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11646 Cope with reduction statements that are vectorized as calls rather
11648 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11649 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11650 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11651 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11652 (UNSPEC_COND_EOR): New unspecs.
11653 (optab): Add mappings for them.
11654 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11655 (sve_int_op, sve_fp_op): New int attributes.
11657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11658 Alan Hayward <alan.hayward@arm.com>
11659 David Sherwood <david.sherwood@arm.com>
11661 * optabs.def (while_ult_optab): New optab.
11662 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11663 * internal-fn.def (WHILE_ULT): New internal function.
11664 * internal-fn.h (direct_internal_fn_supported_p): New override
11665 that takes two types as argument.
11666 * internal-fn.c (while_direct): New macro.
11667 (expand_while_optab_fn): New function.
11668 (convert_optab_supported_p): Likewise.
11669 (direct_while_optab_supported_p): New macro.
11670 * wide-int.h (wi::udiv_ceil): New function.
11671 * tree-vectorizer.h (rgroup_masks): New structure.
11672 (vec_loop_masks): New typedef.
11673 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11674 and fully_masked_p.
11675 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11676 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11677 (vect_max_vf): New function.
11678 (slpeel_make_loop_iterate_ntimes): Delete.
11679 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11680 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11681 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11682 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11683 internal-fn.h, stor-layout.h and optabs-query.h.
11684 (vect_set_loop_mask): New function.
11685 (add_preheader_seq): Likewise.
11686 (add_header_seq): Likewise.
11687 (interleave_supported_p): Likewise.
11688 (vect_maybe_permute_loop_masks): Likewise.
11689 (vect_set_loop_masks_directly): Likewise.
11690 (vect_set_loop_condition_masked): Likewise.
11691 (vect_set_loop_condition_unmasked): New function, split out from
11692 slpeel_make_loop_iterate_ntimes.
11693 (slpeel_make_loop_iterate_ntimes): Rename to..
11694 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11695 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11696 (vect_do_peeling): Update call accordingly.
11697 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11699 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11700 mask_compare_type, can_fully_mask_p and fully_masked_p.
11701 (release_vec_loop_masks): New function.
11702 (_loop_vec_info): Use it to free the loop masks.
11703 (can_produce_all_loop_masks_p): New function.
11704 (vect_get_max_nscalars_per_iter): Likewise.
11705 (vect_verify_full_masking): Likewise.
11706 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11707 retries, and free the mask rgroups before retrying. Check loop-wide
11708 reasons for disallowing fully-masked loops. Make the final decision
11709 about whether use a fully-masked loop or not.
11710 (vect_estimate_min_profitable_iters): Do not assume that peeling
11711 for the number of iterations will be needed for fully-masked loops.
11712 (vectorizable_reduction): Disable fully-masked loops.
11713 (vectorizable_live_operation): Likewise.
11714 (vect_halve_mask_nunits): New function.
11715 (vect_double_mask_nunits): Likewise.
11716 (vect_record_loop_mask): Likewise.
11717 (vect_get_loop_mask): Likewise.
11718 (vect_transform_loop): Handle the case in which the final loop
11719 iteration might handle a partial vector. Call vect_set_loop_condition
11720 instead of slpeel_make_loop_iterate_ntimes.
11721 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11722 (check_load_store_masking): New function.
11723 (prepare_load_store_mask): Likewise.
11724 (vectorizable_store): Handle fully-masked loops.
11725 (vectorizable_load): Likewise.
11726 (supportable_widening_operation): Use vect_halve_mask_nunits for
11728 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11729 (vect_gen_while): New function.
11730 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11731 (aarch64_uqdec<mode>): New insn.
11733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11734 Alan Hayward <alan.hayward@arm.com>
11735 David Sherwood <david.sherwood@arm.com>
11737 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11738 (reduc_xor_scal_optab): New optabs.
11739 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11740 (reduc_xor_scal_@var{m}): Document.
11741 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11742 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11743 internal functions.
11744 * fold-const-call.c (fold_const_call): Handle them.
11745 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11746 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11747 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11748 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11749 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11750 (UNSPEC_XORV): New unspecs.
11751 (optab): Add entries for them.
11752 (BITWISEV): New int iterator.
11753 (bit_reduc_op): New int attributes.
11755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11756 Alan Hayward <alan.hayward@arm.com>
11757 David Sherwood <david.sherwood@arm.com>
11759 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11760 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11761 * optabs.def (vec_shl_insert_optab): New optab.
11762 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11763 (duplicate_and_interleave): Likewise.
11764 * tree-vect-loop.c: Include internal-fn.h.
11765 (neutral_op_for_slp_reduction): New function, split out from
11766 get_initial_defs_for_reduction.
11767 (get_initial_def_for_reduction): Handle option 2 for variable-length
11768 vectors by loading the neutral value into a vector and then shifting
11769 the initial value into element 0.
11770 (get_initial_defs_for_reduction): Replace the code argument with
11771 the neutral value calculated by neutral_op_for_slp_reduction.
11772 Use gimple_build_vector for constant-length vectors.
11773 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11774 but the first group_size elements have a neutral value.
11775 Use duplicate_and_interleave otherwise.
11776 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11777 Update call to get_initial_defs_for_reduction. Handle SLP
11778 reductions for variable-length vectors by creating one vector
11779 result for each scalar result, with the elements associated
11780 with other scalar results stubbed out with the neutral value.
11781 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11782 Require IFN_VEC_SHL_INSERT for double reductions on
11783 variable-length vectors, or SLP reductions that have
11784 a neutral value. Require can_duplicate_and_interleave_p
11785 support for variable-length unchained SLP reductions if there
11786 is no neutral value, such as for MIN/MAX reductions. Also require
11787 the number of vector elements to be a multiple of the number of
11788 SLP statements when doing variable-length unchained SLP reductions.
11789 Update call to vect_create_epilog_for_reduction.
11790 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11791 and remove initial values.
11792 (duplicate_and_interleave): Make public.
11793 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11794 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11797 Alan Hayward <alan.hayward@arm.com>
11798 David Sherwood <david.sherwood@arm.com>
11800 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11801 (can_duplicate_and_interleave_p): New function.
11802 (vect_get_and_check_slp_defs): Take the vector of statements
11803 rather than just the current one. Remove excess parentheses.
11804 Restriction rejectinon of vect_constant_def and vect_external_def
11805 for variable-length vectors to boolean types, or types for which
11806 can_duplicate_and_interleave_p is false.
11807 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11808 (duplicate_and_interleave): New function.
11809 (vect_get_constant_vectors): Use gimple_build_vector for
11810 constant-length vectors and suitable variable-length constant
11811 vectors. Use duplicate_and_interleave for other variable-length
11812 vectors. Don't defer the update when inserting new statements.
11814 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11815 Alan Hayward <alan.hayward@arm.com>
11816 David Sherwood <david.sherwood@arm.com>
11818 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11819 min_profitable_iters doesn't go negative.
11821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11822 Alan Hayward <alan.hayward@arm.com>
11823 David Sherwood <david.sherwood@arm.com>
11825 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11826 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11827 * optabs.def (vec_mask_load_lanes_optab): New optab.
11828 (vec_mask_store_lanes_optab): Likewise.
11829 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11830 (MASK_STORE_LANES): Likewise.
11831 * internal-fn.c (mask_load_lanes_direct): New macro.
11832 (mask_store_lanes_direct): Likewise.
11833 (expand_mask_load_optab_fn): Handle masked operations.
11834 (expand_mask_load_lanes_optab_fn): New macro.
11835 (expand_mask_store_optab_fn): Handle masked operations.
11836 (expand_mask_store_lanes_optab_fn): New macro.
11837 (direct_mask_load_lanes_optab_supported_p): Likewise.
11838 (direct_mask_store_lanes_optab_supported_p): Likewise.
11839 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11841 (vect_load_lanes_supported): Likewise.
11842 * tree-vect-data-refs.c (strip_conversion): New function.
11843 (can_group_stmts_p): Likewise.
11844 (vect_analyze_data_ref_accesses): Use it instead of checking
11845 for a pair of assignments.
11846 (vect_store_lanes_supported): Take a masked_p parameter.
11847 (vect_load_lanes_supported): Likewise.
11848 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11849 vect_store_lanes_supported and vect_load_lanes_supported.
11850 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11851 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11852 parameter. Don't allow gaps for masked accesses.
11853 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11854 and vect_load_lanes_supported.
11855 (get_load_store_type): Take a masked_p parameter and update
11856 call to get_group_load_store_type.
11857 (vectorizable_store): Update call to get_load_store_type.
11858 Handle IFN_MASK_STORE_LANES.
11859 (vectorizable_load): Update call to get_load_store_type.
11860 Handle IFN_MASK_LOAD_LANES.
11862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11863 Alan Hayward <alan.hayward@arm.com>
11864 David Sherwood <david.sherwood@arm.com>
11866 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11868 * config/aarch64/aarch64-protos.h
11869 (aarch64_sve_struct_memory_operand_p): Declare.
11870 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11871 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11872 (VPRED, vpred): Handle SVE structure modes.
11873 * config/aarch64/constraints.md (Utx): New constraint.
11874 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11875 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11876 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11877 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11878 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11879 structure modes. Split into pieces after RA.
11880 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11881 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11883 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11884 SVE structure modes.
11885 (aarch64_classify_address): Likewise.
11886 (sizetochar): Move earlier in file.
11887 (aarch64_print_operand): Handle SVE register lists.
11888 (aarch64_array_mode): New function.
11889 (aarch64_sve_struct_memory_operand_p): Likewise.
11890 (TARGET_ARRAY_MODE): Redefine.
11892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11893 Alan Hayward <alan.hayward@arm.com>
11894 David Sherwood <david.sherwood@arm.com>
11896 * target.def (array_mode): New target hook.
11897 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11898 * doc/tm.texi: Regenerate.
11899 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11900 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11901 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11902 targetm.array_mode.
11903 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11907 Alan Hayward <alan.hayward@arm.com>
11908 David Sherwood <david.sherwood@arm.com>
11910 * fold-const.c (fold_binary_loc): Check the argument types
11911 rather than the result type when testing for a vector operation.
11913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11915 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11916 * doc/tm.texi: Regenerate.
11918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11919 Alan Hayward <alan.hayward@arm.com>
11920 David Sherwood <david.sherwood@arm.com>
11922 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11923 (sve): Document new AArch64 extension.
11924 * doc/md.texi (w): Extend the description of the AArch64
11925 constraint to include SVE vectors.
11926 (Upl, Upa): Document new AArch64 predicate constraints.
11927 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11929 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11930 (msve-vector-bits=): New option.
11931 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11932 SVE when these are disabled.
11933 (sve): New extension.
11934 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11935 modes. Adjust their number of units based on aarch64_sve_vg.
11936 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11937 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11938 aarch64_addr_query_type.
11939 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11940 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11941 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11942 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11943 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11944 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11945 (aarch64_simd_imm_zero_p): Delete.
11946 (aarch64_check_zero_based_sve_index_immediate): Declare.
11947 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11948 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11949 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11950 (aarch64_sve_float_mul_immediate_p): Likewise.
11951 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11952 rather than an rtx.
11953 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11954 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11955 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11956 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11957 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11958 (aarch64_regmode_natural_size): Likewise.
11959 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11960 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11962 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11963 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11964 for VG and the SVE predicate registers.
11965 (V_ALIASES): Add a "z"-prefixed alias.
11966 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11967 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11968 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11969 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11970 (REG_CLASS_NAMES): Add entries for them.
11971 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11972 and the predicate registers.
11973 (aarch64_sve_vg): Declare.
11974 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11975 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11976 (REGMODE_NATURAL_SIZE): Define.
11977 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11979 * config/aarch64/aarch64.c: Include cfgrtl.h.
11980 (simd_immediate_info): Add a constructor for series vectors,
11981 and an associated step field.
11982 (aarch64_sve_vg): New variable.
11983 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11984 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11985 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11986 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11987 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11988 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11989 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11990 (aarch64_get_mask_mode): New functions.
11991 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11992 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11993 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11994 predicate modes and predicate registers. Explicitly restrict
11995 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11996 to store a vector mode if it is recognized by
11997 aarch64_classify_vector_mode.
11998 (aarch64_regmode_natural_size): New function.
11999 (aarch64_hard_regno_caller_save_mode): Return the original mode
12001 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
12002 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
12003 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
12004 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
12006 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
12007 does not overlap dest if the function is frame-related. Handle
12009 (aarch64_split_add_offset): New function.
12010 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
12011 them aarch64_add_offset.
12012 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
12013 and update call to aarch64_sub_sp.
12014 (aarch64_add_cfa_expression): New function.
12015 (aarch64_expand_prologue): Pass extra temporary registers to the
12016 functions above. Handle the case in which we need to emit new
12017 DW_CFA_expressions for registers that were originally saved
12018 relative to the stack pointer, but now have to be expressed
12019 relative to the frame pointer.
12020 (aarch64_output_mi_thunk): Pass extra temporary registers to the
12022 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
12023 IP0 and IP1 values for SVE frames.
12024 (aarch64_expand_vec_series): New function.
12025 (aarch64_expand_sve_widened_duplicate): Likewise.
12026 (aarch64_expand_sve_const_vector): Likewise.
12027 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
12028 Handle SVE constants. Use emit_move_insn to move a force_const_mem
12029 into the register, rather than emitting a SET directly.
12030 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
12031 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
12032 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
12033 (offset_9bit_signed_scaled_p): New functions.
12034 (aarch64_replicate_bitmask_imm): New function.
12035 (aarch64_bitmask_imm): Use it.
12036 (aarch64_cannot_force_const_mem): Reject expressions involving
12037 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12038 (aarch64_classify_index): Handle SVE indices, by requiring
12039 a plain register index with a scale that matches the element size.
12040 (aarch64_classify_address): Handle SVE addresses. Assert that
12041 the mode of the address is VOIDmode or an integer mode.
12042 Update call to aarch64_classify_symbol.
12043 (aarch64_classify_symbolic_expression): Update call to
12044 aarch64_classify_symbol.
12045 (aarch64_const_vec_all_in_range_p): New function.
12046 (aarch64_print_vector_float_operand): Likewise.
12047 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12048 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12049 and the FP immediates 1.0 and 0.5.
12050 (aarch64_print_address_internal): Handle SVE addresses.
12051 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12052 (aarch64_regno_regclass): Handle predicate registers.
12053 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12055 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12056 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12057 (aarch64_convert_sve_vector_bits): New function.
12058 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12059 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12060 rather than an rtx.
12061 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12062 Handle SVE vector and predicate modes. Accept VL-based constants
12063 that need only one temporary register, and VL offsets that require
12064 no temporary registers.
12065 (aarch64_conditional_register_usage): Mark the predicate registers
12066 as fixed if SVE isn't available.
12067 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12068 Return true for SVE vector and predicate modes.
12069 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12070 rather than an unsigned int. Handle SVE modes.
12071 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12073 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12075 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12076 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12077 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12078 (aarch64_sve_float_mul_immediate_p): New functions.
12079 (aarch64_sve_valid_immediate): New function.
12080 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12081 Explicitly reject structure modes. Check for INDEX constants.
12082 Handle PTRUE and PFALSE constants.
12083 (aarch64_check_zero_based_sve_index_immediate): New function.
12084 (aarch64_simd_imm_zero_p): Delete.
12085 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12086 vector modes. Accept constants in the range of CNT[BHWD].
12087 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12088 ask for an Advanced SIMD mode.
12089 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12090 (aarch64_simd_vector_alignment): Handle SVE predicates.
12091 (aarch64_vectorize_preferred_vector_alignment): New function.
12092 (aarch64_simd_vector_alignment_reachable): Use it instead of
12094 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12095 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12097 (MAX_VECT_LEN): Delete.
12098 (expand_vec_perm_d): Add a vec_flags field.
12099 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12100 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12101 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12103 (aarch64_evpc_rev): Rename to...
12104 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12105 (aarch64_evpc_rev_global): New function.
12106 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12107 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12109 (aarch64_evpc_sve_tbl): New function.
12110 (aarch64_expand_vec_perm_const_1): Update after rename of
12111 aarch64_evpc_rev. Handle SVE permutes too, trying
12112 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12113 than aarch64_evpc_tbl.
12114 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12115 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12116 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12117 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12118 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12119 (aarch64_expand_sve_vcond): New functions.
12120 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12121 of aarch64_vector_mode_p.
12122 (aarch64_dwarf_poly_indeterminate_value): New function.
12123 (aarch64_compute_pressure_classes): Likewise.
12124 (aarch64_can_change_mode_class): Likewise.
12125 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12126 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12127 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12128 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12129 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12130 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12131 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12132 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12134 (Dn, Dl, Dr): Accept const as well as const_vector.
12135 (Dz): Likewise. Compare against CONST0_RTX.
12136 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12137 of "vector" where appropriate.
12138 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12139 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12140 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12141 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12142 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12143 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12144 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12145 (v_int_equiv): Extend to SVE modes.
12146 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12148 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12149 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12150 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12151 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12152 (SVE_COND_FP_CMP): New int iterators.
12153 (perm_hilo): Handle the new unpack unspecs.
12154 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12156 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12157 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12158 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12159 (aarch64_equality_operator, aarch64_constant_vector_operand)
12160 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12161 (aarch64_sve_nonimmediate_operand): Likewise.
12162 (aarch64_sve_general_operand): Likewise.
12163 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12164 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12165 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12166 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12167 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12168 (aarch64_sve_float_arith_immediate): Likewise.
12169 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12170 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12171 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12172 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12173 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12174 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12175 (aarch64_sve_float_arith_operand): Likewise.
12176 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12177 (aarch64_sve_float_mul_operand): Likewise.
12178 (aarch64_sve_vec_perm_operand): Likewise.
12179 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12180 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12181 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12182 as well as const_vector.
12183 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12184 in file. Use CONST0_RTX and CONSTM1_RTX.
12185 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12186 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12187 Use aarch64_simd_imm_zero.
12188 * config/aarch64/aarch64-sve.md: New file.
12189 * config/aarch64/aarch64.md: Include it.
12190 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12191 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12192 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12193 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12194 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12195 (sve): New attribute.
12196 (enabled): Disable instructions with the sve attribute unless
12198 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12199 aarch64_expand_mov_immediate.
12200 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12201 CNT[BHSD] immediates.
12202 (movti): Split CONST_POLY_INT moves into two halves.
12203 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12204 Split additions that need a temporary here if the destination
12205 is the stack pointer.
12206 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12207 (*add<mode>3_poly_1): New instruction.
12208 (set_clobber_cc): New expander.
12210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12212 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12213 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12214 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12215 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12216 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12217 Change innermode from fixed_mode_size to machine_mode.
12218 (simplify_subreg): Update call accordingly. Handle a constant-sized
12219 subreg of a variable-length CONST_VECTOR.
12221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12222 Alan Hayward <alan.hayward@arm.com>
12223 David Sherwood <david.sherwood@arm.com>
12225 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12226 (add_offset_to_base): New function, split out from...
12227 (create_mem_ref): ...here. When handling a scale other than 1,
12228 check first whether the address is valid without the offset.
12229 Add it into the base if so, leaving the index and scale as-is.
12231 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12234 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12235 fold_for_warn before checking if arg2 is INTEGER_CST.
12237 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12239 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12240 (store_multiple_operation): Delete.
12241 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12242 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12243 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12244 guarded by TARGET_STRING.
12245 (rs6000_output_load_multiple): Delete.
12246 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12247 OPTION_MASK_STRING / TARGET_STRING handling.
12248 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12249 (const rs6000_opt_masks) <"string">: Change mask to 0.
12250 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12251 (MASK_STRING): Delete.
12252 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12254 (load_multiple): Delete.
12261 (store_multiple): Delete.
12268 (movmemsi_8reg): Delete.
12269 (corresponding unnamed define_insn): Delete.
12270 (movmemsi_6reg): Delete.
12271 (corresponding unnamed define_insn): Delete.
12272 (movmemsi_4reg): Delete.
12273 (corresponding unnamed define_insn): Delete.
12274 (movmemsi_2reg): Delete.
12275 (corresponding unnamed define_insn): Delete.
12276 (movmemsi_1reg): Delete.
12277 (corresponding unnamed define_insn): Delete.
12278 * config/rs6000/rs6000.opt (mno-string): New.
12279 (mstring): Replace by deprecation warning stub.
12280 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12282 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12284 * regrename.c (regrename_do_replace): If replacing the same
12285 reg multiple times, try to reuse last created gen_raw_REG.
12288 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12289 main to workaround a bug in GDB.
12291 2018-01-12 Tom de Vries <tom@codesourcery.com>
12294 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12296 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12298 PR rtl-optimization/80481
12299 * ira-color.c (get_cap_member): New function.
12300 (allocnos_conflict_by_live_ranges_p): Use it.
12301 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12302 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12304 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12307 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12308 (*saddl_se_1): Ditto.
12309 (*ssubsi_1): Ditto.
12310 (*ssubl_se_1): Ditto.
12312 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12314 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12315 rather than wi::to_widest for DR_INITs.
12316 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12317 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12318 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12320 (vect_analyze_group_access_1): Note that here.
12322 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12324 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12325 polynomial type sizes.
12327 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12329 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12330 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12331 (gimple_add_tmp_var): Likewise.
12333 2018-01-12 Martin Liska <mliska@suse.cz>
12335 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12336 (gimple_alloc_sizes): Likewise.
12337 (dump_gimple_statistics): Use PRIu64 in printf format.
12338 * gimple.h: Change uint64_t to int.
12340 2018-01-12 Martin Liska <mliska@suse.cz>
12342 * tree-core.h: Use uint64_t instead of int.
12343 * tree.c (tree_node_counts): Likewise.
12344 (tree_node_sizes): Likewise.
12345 (dump_tree_statistics): Use PRIu64 in printf format.
12347 2018-01-12 Martin Liska <mliska@suse.cz>
12349 * Makefile.in: As qsort_chk is implemented in vec.c, add
12350 vec.o to linkage of gencfn-macros.
12351 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12352 passing the info to record_node_allocation_statistics.
12353 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12355 * ggc-common.c (struct ggc_usage): Add operator== and use
12356 it in operator< and compare function.
12357 * mem-stats.h (struct mem_usage): Likewise.
12358 * vec.c (struct vec_usage): Remove operator< and compare
12359 function. Can be simply inherited.
12361 2018-01-12 Martin Jambor <mjambor@suse.cz>
12364 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12365 * tree-ssa-math-opts.c: Include domwalk.h.
12366 (convert_mult_to_fma_1): New function.
12367 (fma_transformation_info): New type.
12368 (fma_deferring_state): Likewise.
12369 (cancel_fma_deferring): New function.
12370 (result_of_phi): Likewise.
12371 (last_fma_candidate_feeds_initial_phi): Likewise.
12372 (convert_mult_to_fma): Added deferring logic, split actual
12373 transformation to convert_mult_to_fma_1.
12374 (math_opts_dom_walker): New type.
12375 (math_opts_dom_walker::after_dom_children): New method, body moved
12376 here from pass_optimize_widening_mul::execute, added deferring logic
12378 (pass_optimize_widening_mul::execute): Moved most of code to
12379 math_opts_dom_walker::after_dom_children.
12380 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12381 * config/i386/i386.c (ix86_option_override_internal): Added
12382 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12384 2018-01-12 Richard Biener <rguenther@suse.de>
12387 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12388 inline instance vars.
12390 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12393 * config/rx/rx.c (rx_is_restricted_memory_address):
12394 Handle SUBREG case.
12396 2018-01-12 Richard Biener <rguenther@suse.de>
12398 PR tree-optimization/80846
12399 * target.def (split_reduction): New target hook.
12400 * targhooks.c (default_split_reduction): New function.
12401 * targhooks.h (default_split_reduction): Declare.
12402 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12403 target requests first reduce vectors by combining low and high
12405 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12406 (get_vectype_for_scalar_type_and_size): Export.
12407 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12408 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12409 * doc/tm.texi: Regenerate.
12410 * config/i386/i386.c (ix86_split_reduction): Implement
12411 TARGET_VECTORIZE_SPLIT_REDUCTION.
12413 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12416 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12417 in PIC mode except for TARGET_VXWORKS_RTP.
12418 * config/sparc/sparc.c: Include cfgrtl.h.
12419 (TARGET_INIT_PIC_REG): Define.
12420 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12421 (sparc_pic_register_p): New predicate.
12422 (sparc_legitimate_address_p): Use it.
12423 (sparc_legitimize_pic_address): Likewise.
12424 (sparc_delegitimize_address): Likewise.
12425 (sparc_mode_dependent_address_p): Likewise.
12426 (gen_load_pcrel_sym): Remove 4th parameter.
12427 (load_got_register): Adjust call to above. Remove obsolete stuff.
12428 (sparc_expand_prologue): Do not call load_got_register here.
12429 (sparc_flat_expand_prologue): Likewise.
12430 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12431 (sparc_use_pseudo_pic_reg): New function.
12432 (sparc_init_pic_reg): Likewise.
12433 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12434 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12436 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12438 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12439 Add item for branch_cost.
12441 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12443 PR rtl-optimization/83565
12444 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12445 not extend the result to a larger mode for rotate operations.
12446 (num_sign_bit_copies1): Likewise.
12448 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12451 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12453 Use values-Xc.o for -pedantic.
12454 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12456 2018-01-12 Martin Liska <mliska@suse.cz>
12459 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12461 (possible_polymorphic_call_targets): Use it.
12462 (ipa_devirt): Likewise.
12464 2018-01-12 Martin Liska <mliska@suse.cz>
12466 * profile-count.h (enum profile_quality): Use 0 as invalid
12467 enum value of profile_quality.
12469 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12471 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12472 -mext-string options.
12474 2018-01-12 Richard Biener <rguenther@suse.de>
12476 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12477 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12478 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12480 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12482 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12484 * configure.ac (--with-long-double-format): Add support for the
12485 configuration option to change the default long double format on
12487 * config.gcc (powerpc*-linux*-*): Likewise.
12488 * configure: Regenerate.
12489 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12490 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12491 used without modification.
12493 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12495 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12496 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12497 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12498 MISC_BUILTIN_SPEC_BARRIER.
12499 (rs6000_init_builtins): Likewise.
12500 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12502 (speculation_barrier): New define_insn.
12503 * doc/extend.texi: Document __builtin_speculation_barrier.
12505 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12508 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12509 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12510 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12512 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12513 integral modes instead of "ss" and "sd".
12514 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12515 vectors with 32-bit and 64-bit elements.
12516 (vecdupssescalarmodesuffix): New mode attribute.
12517 (vec_dup<mode>): Use it.
12519 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12522 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12523 frame if argument is passed on stack.
12525 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12528 * ree.c (combine_reaching_defs): Optimize also
12529 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12530 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12532 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12534 PR middle-end/83189
12535 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12537 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12539 PR middle-end/83718
12540 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12541 after they are computed.
12543 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12545 PR tree-optimization/83695
12546 * gimple-loop-linterchange.cc
12547 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12548 reset cached scev information after interchange.
12549 (pass_linterchange::execute): Remove call to scev_reset_htab.
12551 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12553 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12554 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12555 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12556 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12557 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12558 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12559 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12560 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12561 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12562 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12563 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12564 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12565 (V_lane_reg): Likewise.
12566 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12568 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12569 (vfmal_lane_low<mode>_intrinsic,
12570 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12571 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12572 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12573 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12574 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12575 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12577 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12579 * config/arm/arm-cpus.in (fp16fml): New feature.
12580 (ALL_SIMD): Add fp16fml.
12581 (armv8.2-a): Add fp16fml as an option.
12582 (armv8.3-a): Likewise.
12583 (armv8.4-a): Add fp16fml as part of fp16.
12584 * config/arm/arm.h (TARGET_FP16FML): Define.
12585 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12587 * config/arm/arm-modes.def (V2HF): Define.
12588 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12589 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12590 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12591 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12592 vfmsl_low, vfmsl_high): New set of builtins.
12593 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12594 (vfml_op): New code attribute.
12595 (VFMLHALVES): New int iterator.
12596 (VFML, VFMLSEL): New mode attributes.
12597 (V_reg): Define mapping for V2HF.
12598 (V_hi, V_lo): New mode attributes.
12599 (VF_constraint): Likewise.
12600 (vfml_half, vfml_half_selector): New int attributes.
12601 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12603 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12604 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12606 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12607 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12608 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12609 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12611 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12612 Document new effective target and option set.
12614 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12616 * config/arm/arm-cpus.in (armv8_4): New feature.
12617 (ARMv8_4a): New fgroup.
12618 (armv8.4-a): New arch.
12619 * config/arm/arm-tables.opt: Regenerate.
12620 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12621 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12622 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12623 Add matching rules for -march=armv8.4-a and extensions.
12624 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12626 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12629 * config/rx/rx.md (BW): New mode attribute.
12630 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12632 2018-01-11 Richard Biener <rguenther@suse.de>
12634 PR tree-optimization/83435
12635 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12636 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12637 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12639 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12640 Alan Hayward <alan.hayward@arm.com>
12641 David Sherwood <david.sherwood@arm.com>
12643 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12645 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12646 (aarch64_print_address_internal): Use it to check for a zero offset.
12648 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12649 Alan Hayward <alan.hayward@arm.com>
12650 David Sherwood <david.sherwood@arm.com>
12652 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12653 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12654 Return a poly_int64 rather than a HOST_WIDE_INT.
12655 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12656 rather than a HOST_WIDE_INT.
12657 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12658 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12659 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12660 final_offset from HOST_WIDE_INT to poly_int64.
12661 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12662 to_constant when getting the number of units in an Advanced SIMD
12664 (aarch64_builtin_vectorized_function): Check for a constant number
12666 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12668 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12669 attribute instead of GET_MODE_NUNITS.
12670 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12671 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12672 GET_MODE_SIZE for fixed-size registers.
12673 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12674 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12675 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12676 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12677 (aarch64_print_operand, aarch64_print_address_internal)
12678 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12679 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12680 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12681 Handle polynomial GET_MODE_SIZE.
12682 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12683 wider than SImode without modification.
12684 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12685 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12686 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12687 passing and returning SVE modes.
12688 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12689 rather than GEN_INT.
12690 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12691 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12692 (aarch64_allocate_and_probe_stack_space): Likewise.
12693 (aarch64_layout_frame): Cope with polynomial offsets.
12694 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12695 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12696 polynomial offsets.
12697 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12698 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12699 poly_int64 rather than a HOST_WIDE_INT.
12700 (aarch64_get_separate_components, aarch64_process_components)
12701 (aarch64_expand_prologue, aarch64_expand_epilogue)
12702 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12703 (aarch64_anchor_offset): New function, split out from...
12704 (aarch64_legitimize_address): ...here.
12705 (aarch64_builtin_vectorization_cost): Handle polynomial
12706 TYPE_VECTOR_SUBPARTS.
12707 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12709 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12710 number of elements from the PARALLEL rather than the mode.
12711 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12712 rather than GET_MODE_BITSIZE.
12713 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12714 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12715 (aarch64_expand_vec_perm_const_1): Handle polynomial
12716 d->perm.length () and d->perm elements.
12717 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12718 Apply to_constant to d->perm elements.
12719 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12720 polynomial CONST_VECTOR_NUNITS.
12721 (aarch64_move_pointer): Take amount as a poly_int64 rather
12723 (aarch64_progress_pointer): Avoid temporary variable.
12724 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12725 the mode attribute instead of GET_MODE.
12727 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12728 Alan Hayward <alan.hayward@arm.com>
12729 David Sherwood <david.sherwood@arm.com>
12731 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12732 x exists before using it.
12733 (aarch64_add_constant_internal): Rename to...
12734 (aarch64_add_offset_1): ...this. Replace regnum with separate
12735 src and dest rtxes. Handle the case in which they're different,
12736 including when the offset is zero. Replace scratchreg with an rtx.
12737 Use 2 additions if there is no spare register into which we can
12738 move a 16-bit constant.
12739 (aarch64_add_constant): Delete.
12740 (aarch64_add_offset): Replace reg with separate src and dest
12741 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12742 Use aarch64_add_offset_1.
12743 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12744 an rtx rather than an int. Take the delta as a poly_int64
12745 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12746 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12747 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12748 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12749 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12750 and aarch64_add_sp.
12751 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12752 aarch64_add_constant.
12754 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12756 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12757 Use scalar_float_mode.
12759 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12761 * config/aarch64/aarch64-simd.md
12762 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12763 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12764 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12765 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12766 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12767 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12768 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12769 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12770 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12771 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12773 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12776 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12777 targ_options->x_arm_arch_string is non NULL.
12779 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12781 * config/aarch64/aarch64.h
12782 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12784 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12787 * expmed.c (emit_store_flag_force): Swap if const op0
12788 and change VOIDmode to mode of op0.
12790 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12792 PR rtl-optimization/83761
12793 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12794 than bytes to mode_for_size.
12796 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12798 PR middle-end/83189
12799 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12800 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12803 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12805 PR middle-end/83575
12806 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12807 when in layout mode.
12808 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12809 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12812 2018-01-10 Michael Collison <michael.collison@arm.com>
12814 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12815 * config/aarch64/aarch64-option-extension.def: Add
12816 AARCH64_OPT_EXTENSION of 'fp16fml'.
12817 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12818 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12819 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12820 * config/aarch64/constraints.md (Ui7): New constraint.
12821 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12822 (VFMLA_SEL_W): Ditto.
12825 (VFMLA16_LOW): New int iterator.
12826 (VFMLA16_HIGH): Ditto.
12827 (UNSPEC_FMLAL): New unspec.
12828 (UNSPEC_FMLSL): Ditto.
12829 (UNSPEC_FMLAL2): Ditto.
12830 (UNSPEC_FMLSL2): Ditto.
12831 (f16mac): New code attribute.
12832 * config/aarch64/aarch64-simd-builtins.def
12833 (aarch64_fmlal_lowv2sf): Ditto.
12834 (aarch64_fmlsl_lowv2sf): Ditto.
12835 (aarch64_fmlalq_lowv4sf): Ditto.
12836 (aarch64_fmlslq_lowv4sf): Ditto.
12837 (aarch64_fmlal_highv2sf): Ditto.
12838 (aarch64_fmlsl_highv2sf): Ditto.
12839 (aarch64_fmlalq_highv4sf): Ditto.
12840 (aarch64_fmlslq_highv4sf): Ditto.
12841 (aarch64_fmlal_lane_lowv2sf): Ditto.
12842 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12843 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12844 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12845 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12846 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12847 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12848 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12849 (aarch64_fmlal_lane_highv2sf): Ditto.
12850 (aarch64_fmlsl_lane_highv2sf): Ditto.
12851 (aarch64_fmlal_laneq_highv2sf): Ditto.
12852 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12853 (aarch64_fmlalq_lane_highv4sf): Ditto.
12854 (aarch64_fmlsl_lane_highv4sf): Ditto.
12855 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12856 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12857 * config/aarch64/aarch64-simd.md:
12858 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12859 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12860 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12861 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12862 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12863 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12864 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12865 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12866 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12867 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12868 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12869 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12870 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12871 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12872 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12873 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12874 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12875 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12876 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12877 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12878 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12879 (vfmlsl_low_u32): Ditto.
12880 (vfmlalq_low_u32): Ditto.
12881 (vfmlslq_low_u32): Ditto.
12882 (vfmlal_high_u32): Ditto.
12883 (vfmlsl_high_u32): Ditto.
12884 (vfmlalq_high_u32): Ditto.
12885 (vfmlslq_high_u32): Ditto.
12886 (vfmlal_lane_low_u32): Ditto.
12887 (vfmlsl_lane_low_u32): Ditto.
12888 (vfmlal_laneq_low_u32): Ditto.
12889 (vfmlsl_laneq_low_u32): Ditto.
12890 (vfmlalq_lane_low_u32): Ditto.
12891 (vfmlslq_lane_low_u32): Ditto.
12892 (vfmlalq_laneq_low_u32): Ditto.
12893 (vfmlslq_laneq_low_u32): Ditto.
12894 (vfmlal_lane_high_u32): Ditto.
12895 (vfmlsl_lane_high_u32): Ditto.
12896 (vfmlal_laneq_high_u32): Ditto.
12897 (vfmlsl_laneq_high_u32): Ditto.
12898 (vfmlalq_lane_high_u32): Ditto.
12899 (vfmlslq_lane_high_u32): Ditto.
12900 (vfmlalq_laneq_high_u32): Ditto.
12901 (vfmlslq_laneq_high_u32): Ditto.
12902 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12903 (AARCH64_FL_FOR_ARCH8_4): New.
12904 (AARCH64_ISA_F16FML): New ISA flag.
12905 (TARGET_F16FML): New feature flag for fp16fml.
12906 (doc/invoke.texi): Document new fp16fml option.
12908 2018-01-10 Michael Collison <michael.collison@arm.com>
12910 * config/aarch64/aarch64-builtins.c:
12911 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12912 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12913 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12914 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12915 (AARCH64_ISA_SHA3): New ISA flag.
12916 (TARGET_SHA3): New feature flag for sha3.
12917 * config/aarch64/iterators.md (sha512_op): New int attribute.
12918 (CRYPTO_SHA512): New int iterator.
12919 (UNSPEC_SHA512H): New unspec.
12920 (UNSPEC_SHA512H2): Ditto.
12921 (UNSPEC_SHA512SU0): Ditto.
12922 (UNSPEC_SHA512SU1): Ditto.
12923 * config/aarch64/aarch64-simd-builtins.def
12924 (aarch64_crypto_sha512hqv2di): New builtin.
12925 (aarch64_crypto_sha512h2qv2di): Ditto.
12926 (aarch64_crypto_sha512su0qv2di): Ditto.
12927 (aarch64_crypto_sha512su1qv2di): Ditto.
12928 (aarch64_eor3qv8hi): Ditto.
12929 (aarch64_rax1qv2di): Ditto.
12930 (aarch64_xarqv2di): Ditto.
12931 (aarch64_bcaxqv8hi): Ditto.
12932 * config/aarch64/aarch64-simd.md:
12933 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12934 (aarch64_crypto_sha512su0qv2di): Ditto.
12935 (aarch64_crypto_sha512su1qv2di): Ditto.
12936 (aarch64_eor3qv8hi): Ditto.
12937 (aarch64_rax1qv2di): Ditto.
12938 (aarch64_xarqv2di): Ditto.
12939 (aarch64_bcaxqv8hi): Ditto.
12940 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12941 (vsha512h2q_u64): Ditto.
12942 (vsha512su0q_u64): Ditto.
12943 (vsha512su1q_u64): Ditto.
12944 (veor3q_u16): Ditto.
12945 (vrax1q_u64): Ditto.
12946 (vxarq_u64): Ditto.
12947 (vbcaxq_u16): Ditto.
12948 * config/arm/types.md (crypto_sha512): New type attribute.
12949 (crypto_sha3): Ditto.
12950 (doc/invoke.texi): Document new sha3 option.
12952 2018-01-10 Michael Collison <michael.collison@arm.com>
12954 * config/aarch64/aarch64-builtins.c:
12955 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12956 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12957 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12958 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12959 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12960 (AARCH64_ISA_SM4): New ISA flag.
12961 (TARGET_SM4): New feature flag for sm4.
12962 * config/aarch64/aarch64-simd-builtins.def
12963 (aarch64_sm3ss1qv4si): Ditto.
12964 (aarch64_sm3tt1aq4si): Ditto.
12965 (aarch64_sm3tt1bq4si): Ditto.
12966 (aarch64_sm3tt2aq4si): Ditto.
12967 (aarch64_sm3tt2bq4si): Ditto.
12968 (aarch64_sm3partw1qv4si): Ditto.
12969 (aarch64_sm3partw2qv4si): Ditto.
12970 (aarch64_sm4eqv4si): Ditto.
12971 (aarch64_sm4ekeyqv4si): Ditto.
12972 * config/aarch64/aarch64-simd.md:
12973 (aarch64_sm3ss1qv4si): Ditto.
12974 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12975 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12976 (aarch64_sm4eqv4si): Ditto.
12977 (aarch64_sm4ekeyqv4si): Ditto.
12978 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12979 (sm3part_op): Ditto.
12980 (CRYPTO_SM3TT): Ditto.
12981 (CRYPTO_SM3PART): Ditto.
12982 (UNSPEC_SM3SS1): New unspec.
12983 (UNSPEC_SM3TT1A): Ditto.
12984 (UNSPEC_SM3TT1B): Ditto.
12985 (UNSPEC_SM3TT2A): Ditto.
12986 (UNSPEC_SM3TT2B): Ditto.
12987 (UNSPEC_SM3PARTW1): Ditto.
12988 (UNSPEC_SM3PARTW2): Ditto.
12989 (UNSPEC_SM4E): Ditto.
12990 (UNSPEC_SM4EKEY): Ditto.
12991 * config/aarch64/constraints.md (Ui2): New constraint.
12992 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12993 * config/arm/types.md (crypto_sm3): New type attribute.
12994 (crypto_sm4): Ditto.
12995 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12996 (vsm3tt1aq_u32): Ditto.
12997 (vsm3tt1bq_u32): Ditto.
12998 (vsm3tt2aq_u32): Ditto.
12999 (vsm3tt2bq_u32): Ditto.
13000 (vsm3partw1q_u32): Ditto.
13001 (vsm3partw2q_u32): Ditto.
13002 (vsm4eq_u32): Ditto.
13003 (vsm4ekeyq_u32): Ditto.
13004 (doc/invoke.texi): Document new sm4 option.
13006 2018-01-10 Michael Collison <michael.collison@arm.com>
13008 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
13009 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
13010 (AARCH64_FL_FOR_ARCH8_4): New.
13011 (AARCH64_FL_V8_4): New flag.
13012 (doc/invoke.texi): Document new armv8.4-a option.
13014 2018-01-10 Michael Collison <michael.collison@arm.com>
13016 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13017 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
13018 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
13019 * config/aarch64/aarch64-option-extension.def: Add
13020 AARCH64_OPT_EXTENSION of 'sha2'.
13021 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
13022 (crypto): Disable sha2 and aes if crypto disabled.
13023 (crypto): Enable aes and sha2 if enabled.
13024 (simd): Disable sha2 and aes if simd disabled.
13025 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
13027 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
13028 (TARGET_SHA2): New feature flag for sha2.
13029 (TARGET_AES): New feature flag for aes.
13030 * config/aarch64/aarch64-simd.md:
13031 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
13032 conditional on TARGET_AES.
13033 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
13034 (aarch64_crypto_sha1hsi): Make pattern conditional
13036 (aarch64_crypto_sha1hv4si): Ditto.
13037 (aarch64_be_crypto_sha1hv4si): Ditto.
13038 (aarch64_crypto_sha1su1v4si): Ditto.
13039 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13040 (aarch64_crypto_sha1su0v4si): Ditto.
13041 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13042 (aarch64_crypto_sha256su0v4si): Ditto.
13043 (aarch64_crypto_sha256su1v4si): Ditto.
13044 (doc/invoke.texi): Document new aes and sha2 options.
13046 2018-01-10 Martin Sebor <msebor@redhat.com>
13048 PR tree-optimization/83781
13049 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13052 2018-01-11 Martin Sebor <msebor@gmail.com>
13053 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13055 PR tree-optimization/83501
13056 PR tree-optimization/81703
13058 * tree-ssa-strlen.c (get_string_cst): Rename...
13059 (get_string_len): ...to this. Handle global constants.
13060 (handle_char_store): Adjust.
13062 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13063 Jim Wilson <jimw@sifive.com>
13065 * config/riscv/riscv-protos.h (riscv_output_return): New.
13066 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13067 (riscv_attribute_table, riscv_output_return),
13068 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13069 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13070 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13072 (riscv_expand_prologue): Add early return for naked function.
13073 (riscv_expand_epilogue): Likewise.
13074 (riscv_function_ok_for_sibcall): Return false for naked function.
13075 (riscv_set_current_function): New.
13076 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13077 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13078 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13079 * doc/extend.texi (RISC-V Function Attributes): New.
13081 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13083 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13084 check for 128-bit long double before checking TCmode.
13085 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13086 128-bit long doubles before checking TFmode or TCmode.
13087 (FLOAT128_IBM_P): Likewise.
13089 2018-01-10 Martin Sebor <msebor@redhat.com>
13091 PR tree-optimization/83671
13092 * builtins.c (c_strlen): Unconditionally return zero for the empty
13094 Use -Warray-bounds for warnings.
13095 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13096 for non-constant array indices with COMPONENT_REF, arrays of
13097 arrays, and pointers to arrays.
13098 (gimple_fold_builtin_strlen): Determine and set length range for
13099 non-constant character arrays.
13101 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13103 PR middle-end/81897
13104 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13107 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13109 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13111 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13114 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13115 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13116 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13117 indexed_or_indirect_operand predicate.
13118 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13119 (*vsx_le_perm_load_v8hi): Likewise.
13120 (*vsx_le_perm_load_v16qi): Likewise.
13121 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13122 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13123 (*vsx_le_perm_store_v8hi): Likewise.
13124 (*vsx_le_perm_store_v16qi): Likewise.
13125 (eight unnamed splitters): Likewise.
13127 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13129 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13130 * config/rs6000/emmintrin.h: Likewise.
13131 * config/rs6000/mmintrin.h: Likewise.
13132 * config/rs6000/xmmintrin.h: Likewise.
13134 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13137 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13139 * tree.c (tree_nop_conversion): Return true for location wrapper
13141 (maybe_wrap_with_location): New function.
13142 (selftest::check_strip_nops): New function.
13143 (selftest::test_location_wrappers): New function.
13144 (selftest::tree_c_tests): Call it.
13145 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13146 (maybe_wrap_with_location): New decl.
13147 (EXPR_LOCATION_WRAPPER_P): New macro.
13148 (location_wrapper_p): New inline function.
13149 (tree_strip_any_location_wrapper): New inline function.
13151 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13154 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13155 stack_realign_offset for the largest alignment of stack slot
13157 (ix86_find_max_used_stack_alignment): New function.
13158 (ix86_finalize_stack_frame_flags): Use it. Set
13159 max_used_stack_alignment if we don't realign stack.
13160 * config/i386/i386.h (machine_function): Add
13161 max_used_stack_alignment.
13163 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13165 * config/arm/arm.opt (-mbranch-cost): New option.
13166 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13169 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13172 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13173 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13175 2018-01-10 Richard Biener <rguenther@suse.de>
13178 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13179 early out so it also covers the case where we have a non-NULL
13182 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13184 PR tree-optimization/83753
13185 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13186 for non-strided grouped accesses if the number of elements is 1.
13188 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13191 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13192 * i386.h (TARGET_USE_GATHER): Define.
13193 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13195 2018-01-10 Martin Liska <mliska@suse.cz>
13198 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13199 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13201 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13202 CLEANUP_NO_PARTITIONING is not set.
13204 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13206 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13207 for vectors, as a partial revert of r254296.
13208 * rtl.h (const_vec_p): Delete.
13209 (const_vec_duplicate_p): Don't test for vector CONSTs.
13210 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13211 * expmed.c (make_tree): Likewise.
13214 * common.md (E, F): Use CONSTANT_P instead of checking for
13216 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13217 checking for CONST_VECTOR.
13219 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13221 PR middle-end/83575
13222 * predict.c (force_edge_cold): Handle in more sane way edges
13223 with no prediction.
13225 2018-01-09 Carl Love <cel@us.ibm.com>
13227 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13229 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13230 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13231 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13232 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13233 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13234 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13235 * config/rs6000/rs6000-protos.h: Add extern defition for
13236 rs6000_generate_float2_double_code.
13237 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13239 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13240 (float2_v2df): Add define_expand.
13242 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13245 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13246 op_mode in the force_to_mode call.
13248 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13250 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13251 instead of checking each element individually.
13252 (aarch64_evpc_uzp): Likewise.
13253 (aarch64_evpc_zip): Likewise.
13254 (aarch64_evpc_ext): Likewise.
13255 (aarch64_evpc_rev): Likewise.
13256 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13257 instead of checking each element individually. Return true without
13259 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13260 whether all selected elements come from the same input, instead of
13261 checking each element individually. Remove calls to gen_rtx_REG,
13262 start_sequence and end_sequence and instead assert that no rtl is
13265 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13267 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13268 order of HIGH and CONST checks.
13270 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13272 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13273 if the destination isn't an SSA_NAME.
13275 2018-01-09 Richard Biener <rguenther@suse.de>
13277 PR tree-optimization/83668
13278 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13280 (canonicalize_loop_form): ... here, renamed from ...
13281 (canonicalize_loop_closed_ssa_form): ... this and amended to
13282 swap successor edges for loop exit blocks to make us use
13283 the RPO order we need for initial schedule generation.
13285 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13287 PR tree-optimization/64811
13288 * match.pd: When optimizing comparisons with Inf, avoid
13289 introducing or losing exceptions from comparisons with NaN.
13291 2018-01-09 Martin Liska <mliska@suse.cz>
13294 * asan.c (shadow_mem_size): Add gcc_assert.
13296 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13298 Don't save registers in main().
13301 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13302 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13303 * config/avr/avr.c (avr_set_current_function): Don't error if
13304 naked, OS_task or OS_main are specified at the same time.
13305 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13307 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13309 * common/config/avr/avr-common.c (avr_option_optimization_table):
13310 Switch on -mmain-is-OS_task for optimizing compilations.
13312 2018-01-09 Richard Biener <rguenther@suse.de>
13314 PR tree-optimization/83572
13315 * graphite.c: Include cfganal.h.
13316 (graphite_transform_loops): Connect infinite loops to exit
13317 and remove fake edges at the end.
13319 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13321 * ipa-inline.c (edge_badness): Revert accidental checkin.
13323 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13326 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13327 symbols; not inline clones.
13329 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13332 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13333 hard registers. Formatting fixes.
13335 PR preprocessor/83722
13336 * gcc.c (try_generate_repro): Pass
13337 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13338 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13341 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13342 Kito Cheng <kito.cheng@gmail.com>
13344 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13345 (riscv_leaf_function_p): Delete.
13346 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13348 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13350 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13352 (do_ifelse): New function.
13353 (do_isel): New function.
13354 (do_sub3): New function.
13355 (do_add3): New function.
13356 (do_load_mask_compare): New function.
13357 (do_overlap_load_compare): New function.
13358 (expand_compare_loop): New function.
13359 (expand_block_compare): Call expand_compare_loop() when appropriate.
13360 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13361 option description.
13362 (-mblock-compare-inline-loop-limit): New option.
13364 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13367 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13368 Reverse order of second and third operands in first alternative.
13369 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13370 of first and second elements in UNSPEC_VPERMR vector.
13371 (altivec_expand_vec_perm_le): Likewise.
13373 2018-01-08 Jeff Law <law@redhat.com>
13375 PR rtl-optimizatin/81308
13376 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13377 (process_switch): If group_case_labels makes a change, then set
13379 (pass_convert_switch::execute): If a switch is converted, then
13380 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13382 PR rtl-optimization/81308
13383 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13386 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13388 PR target/83663 - Revert r255946
13389 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13390 generation for cases where splatting a value is not useful.
13391 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13392 across a vec_duplicate and a paradoxical subreg forming a vector
13393 mode to a vec_concat.
13395 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13397 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13398 -march=armv8.3-a variants.
13399 * config/arm/t-multilib: Likewise.
13400 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13402 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13404 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13406 (cceq_ior_compare_complement): Give it a name so I can use it, and
13407 change boolean_or_operator predicate to boolean_operator so it can
13408 be used to generate a crand.
13409 (eqne): New code iterator.
13410 (bd/bd_neg): New code_attrs.
13411 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13412 a single define_insn.
13413 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13414 decrement (bdnzt/bdnzf/bdzt/bdzf).
13415 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13416 with the new names of the branch decrement patterns, and added the
13417 names of the branch decrement conditional patterns.
13419 2018-01-08 Richard Biener <rguenther@suse.de>
13421 PR tree-optimization/83563
13422 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13425 2018-01-08 Richard Biener <rguenther@suse.de>
13427 PR middle-end/83713
13428 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13430 2018-01-08 Richard Biener <rguenther@suse.de>
13432 PR tree-optimization/83685
13433 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13434 references to abnormals.
13436 2018-01-08 Richard Biener <rguenther@suse.de>
13439 * dwarf2out.c (output_indirect_strings): Handle empty
13440 skeleton_debug_str_hash.
13441 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13443 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13445 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13446 (emit_store_direct): Likewise.
13447 (arc_trampoline_adjust_address): Likewise.
13448 (arc_asm_trampoline_template): New function.
13449 (arc_initialize_trampoline): Use asm_trampoline_template.
13450 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13451 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13452 * config/arc/arc.md (flush_icache): Delete pattern.
13454 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13456 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13457 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13460 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13463 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13464 by not USED_FOR_TARGET.
13465 (make_pass_resolve_sw_modes): Likewise.
13467 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13469 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13472 2018-01-08 Richard Biener <rguenther@suse.de>
13474 PR middle-end/83580
13475 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13477 2018-01-08 Richard Biener <rguenther@suse.de>
13479 PR middle-end/83517
13480 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13482 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13484 PR middle-end/81897
13485 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13486 basic blocks with a small number of successors.
13487 (convert_control_dep_chain_into_preds): Improve handling of
13489 (dump_predicates): Split apart into...
13490 (dump_pred_chain): ...here...
13491 (dump_pred_info): ...and here.
13492 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13493 (can_chain_union_be_invalidated_p): Improve check for invalidation
13495 (uninit_uses_cannot_happen): Avoid unnecessary if
13496 convert_control_dep_chain_into_preds yielded nothing.
13498 2018-01-06 Martin Sebor <msebor@redhat.com>
13500 PR tree-optimization/83640
13501 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13502 subtracting negative offset from size.
13503 (builtin_access::overlap): Adjust offset bounds of the access to fall
13504 within the size of the object if possible.
13506 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13508 PR rtl-optimization/83699
13509 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13510 extract_bit_field_as_subreg to cases in which the extracted
13511 value is also a vector.
13513 * lra-constraints.c (process_alt_operands): Test for the equivalence
13514 substitutions when detecting a possible reload cycle.
13516 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13519 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13520 by default if flag_selective_schedling{,2}. Formatting fixes.
13522 PR rtl-optimization/83682
13523 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13524 if it has non-VECTOR_MODE element mode.
13525 (vec_duplicate_p): Likewise.
13527 PR middle-end/83694
13528 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13529 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13531 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13534 * config/i386/i386-builtin.def
13535 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13536 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13537 Require also OPTION_MASK_ISA_AVX512F in addition to
13538 OPTION_MASK_ISA_GFNI.
13539 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13540 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13541 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13542 to OPTION_MASK_ISA_GFNI.
13543 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13544 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13545 OPTION_MASK_ISA_AVX512BW.
13546 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13547 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13548 addition to OPTION_MASK_ISA_GFNI.
13549 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13550 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13551 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13552 to OPTION_MASK_ISA_GFNI.
13553 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13554 a requirement for all ISAs rather than any of them with a few
13556 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13558 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13559 bitmasks to be enabled with 3 exceptions, instead of requiring any
13560 enabled ISA with lots of exceptions.
13561 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13562 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13563 Change avx512bw in isa attribute to avx512f.
13564 * config/i386/sgxintrin.h: Add license boilerplate.
13565 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13566 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13567 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13568 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13570 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13571 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13572 temporarily sse2 rather than sse if not enabled already.
13575 * config/i386/sse.md (VI248_VLBW): Rename to ...
13576 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13577 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13578 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13579 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13580 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13581 mode iterator instead of VI248_VLBW.
13583 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13585 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13586 (record_modified): Skip clobbers; add debug output.
13587 (param_change_prob): Use sreal frequencies.
13589 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13591 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13592 punt for user-aligned variables.
13594 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13596 * tree-chrec.c (chrec_contains_symbols): Return true for
13599 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13602 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13603 of (x|y) == x for BICS pattern.
13605 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13607 PR tree-optimization/83605
13608 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13609 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13612 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13614 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13615 * config/epiphany/rtems.h: New file.
13617 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13618 Uros Bizjak <ubizjak@gmail.com>
13621 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13622 QIreg_operand instead of register_operand predicate.
13623 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13624 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13625 comments instead of -fmitigate[-_]rop.
13627 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13630 * cgraphunit.c (symbol_table::compile): Switch to text_section
13631 before calling assembly_start debug hook.
13632 * run-rtl-passes.c (run_rtl_passes): Likewise.
13635 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13637 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13638 range_int_cst_p rather than !symbolic_range_p before calling
13639 extract_range_from_multiplicative_op_1.
13641 2018-01-04 Jeff Law <law@redhat.com>
13643 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13644 redundant test in assertion.
13646 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13648 * doc/rtl.texi: Document machine_mode wrapper classes.
13650 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13652 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13653 using tree_to_uhwi.
13655 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13657 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13658 the VEC_PERM_EXPR fold to fail.
13660 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13663 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13664 to switched_sections.
13666 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13669 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13670 test for d.testing.
13672 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13675 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13676 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13678 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13681 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13682 is BLKmode and bitpos not zero or mode change is needed.
13684 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13687 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13690 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13693 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13694 instead of MULT rtx. Update all corresponding splitters.
13695 (*saddl_se): Ditto.
13696 (*ssub<modesuffix>): Ditto.
13697 (*ssubl_se): Ditto.
13698 (*cmp_sadd_di): Update split patterns.
13699 (*cmp_sadd_si): Ditto.
13700 (*cmp_sadd_sidi): Ditto.
13701 (*cmp_ssub_di): Ditto.
13702 (*cmp_ssub_si): Ditto.
13703 (*cmp_ssub_sidi): Ditto.
13704 * config/alpha/predicates.md (const23_operand): New predicate.
13705 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13706 Look for ASHIFT, not MULT inner operand.
13707 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13709 2018-01-04 Martin Liska <mliska@suse.cz>
13711 PR gcov-profile/83669
13712 * gcov.c (output_intermediate_file): Add version to intermediate
13714 * doc/gcov.texi: Document new field 'version' in intermediate
13715 file format. Fix location of '-k' option of gcov command.
13717 2018-01-04 Martin Liska <mliska@suse.cz>
13720 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13722 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13724 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13726 2018-01-03 Martin Sebor <msebor@redhat.com>
13728 PR tree-optimization/83655
13729 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13730 checking calls with invalid arguments.
13732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13734 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13735 (vectorizable_mask_load_store): Delete.
13736 (vectorizable_call): Return false for masked loads and stores.
13737 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13738 instead of gimple_assign_rhs1.
13739 (vectorizable_load): Handle IFN_MASK_LOAD.
13740 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13744 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13746 (vectorizable_mask_load_store): ...here.
13747 (vectorizable_load): ...and here.
13749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13751 * tree-vect-stmts.c (vect_build_all_ones_mask)
13752 (vect_build_zero_merge_argument): New functions, split out from...
13753 (vectorizable_load): ...here.
13755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13757 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13759 (vectorizable_mask_load_store): ...here.
13760 (vectorizable_store): ...and here.
13762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13764 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13766 (vectorizable_mask_load_store): ...here.
13768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13770 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13771 (vect_model_store_cost): Take a vec_load_store_type instead of a
13773 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13774 (vect_model_store_cost): Take a vec_load_store_type instead of a
13776 (vectorizable_mask_load_store): Update accordingly.
13777 (vectorizable_store): Likewise.
13778 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13782 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13783 IFN_MASK_LOAD calls here rather than...
13784 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13787 Alan Hayward <alan.hayward@arm.com>
13788 David Sherwood <david.sherwood@arm.com>
13790 * expmed.c (extract_bit_field_1): For vector extracts,
13791 fall back to extract_bit_field_as_subreg if vec_extract
13794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13795 Alan Hayward <alan.hayward@arm.com>
13796 David Sherwood <david.sherwood@arm.com>
13798 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13799 they are variable or constant sized.
13800 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13801 slots for constant-sized data.
13803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13804 Alan Hayward <alan.hayward@arm.com>
13805 David Sherwood <david.sherwood@arm.com>
13807 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13808 handling COND_EXPRs with boolean comparisons, try to find a better
13809 basis for the mask type than the boolean itself.
13811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13813 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13814 is calculated and how it can be overridden.
13815 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13816 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13818 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13822 Alan Hayward <alan.hayward@arm.com>
13823 David Sherwood <david.sherwood@arm.com>
13825 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13826 Remove the mode argument.
13827 (aarch64_simd_valid_immediate): Remove the mode and inverse
13829 * config/aarch64/iterators.md (bitsize): New iterator.
13830 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13831 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13832 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13833 aarch64_simd_valid_immediate.
13834 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13835 (aarch64_reg_or_bic_imm): Likewise.
13836 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13837 with an insn_type enum and msl with a modifier_type enum.
13838 Replace element_width with a scalar_mode. Change the shift
13839 to unsigned int. Add constructors for scalar_float_mode and
13840 scalar_int_mode elements.
13841 (aarch64_vect_float_const_representable_p): Delete.
13842 (aarch64_can_const_movi_rtx_p)
13843 (aarch64_simd_scalar_immediate_valid_for_move)
13844 (aarch64_simd_make_constant): Update call to
13845 aarch64_simd_valid_immediate.
13846 (aarch64_advsimd_valid_immediate_hs): New function.
13847 (aarch64_advsimd_valid_immediate): Likewise.
13848 (aarch64_simd_valid_immediate): Remove mode and inverse
13849 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13850 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13851 and aarch64_float_const_representable_p on the result.
13852 (aarch64_output_simd_mov_immediate): Remove mode argument.
13853 Update call to aarch64_simd_valid_immediate and use of
13854 simd_immediate_info.
13855 (aarch64_output_scalar_simd_mov_immediate): Update call
13858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13859 Alan Hayward <alan.hayward@arm.com>
13860 David Sherwood <david.sherwood@arm.com>
13862 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13863 (mode_nunits): Likewise CONST_MODE_NUNITS.
13864 * machmode.def (ADJUST_NUNITS): Document.
13865 * genmodes.c (mode_data::need_nunits_adj): New field.
13866 (blank_mode): Update accordingly.
13867 (adj_nunits): New variable.
13868 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13870 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13871 listed in adj_nunits.
13872 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13873 listed in adj_nunits. Don't emit case statements for such modes.
13874 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13875 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13876 nothing if adj_nunits is nonnull.
13877 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13878 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13879 (emit_mode_fbit): Update use of print_maybe_const_decl.
13880 (emit_move_size): Likewise. Treat the array as non-const
13882 (emit_mode_adjustments): Handle adj_nunits.
13884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13886 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13887 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13888 (VECTOR_MODES): Use it.
13889 (make_vector_modes): Take the prefix as an argument.
13891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13892 Alan Hayward <alan.hayward@arm.com>
13893 David Sherwood <david.sherwood@arm.com>
13895 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13896 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13897 for MODE_VECTOR_BOOL.
13898 * machmode.def (VECTOR_BOOL_MODE): Document.
13899 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13900 (make_vector_bool_mode): New function.
13901 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13903 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13904 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13906 * stor-layout.c (int_mode_for_mode): Likewise.
13907 * tree.c (build_vector_type_for_mode): Likewise.
13908 * varasm.c (output_constant_pool_2): Likewise.
13909 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13910 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13911 for MODE_VECTOR_BOOL.
13912 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13913 of mode class checks.
13914 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13915 instead of a list of mode class checks.
13916 (expand_vector_scalar_condition): Likewise.
13917 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13920 Alan Hayward <alan.hayward@arm.com>
13921 David Sherwood <david.sherwood@arm.com>
13923 * machmode.h (mode_size): Change from unsigned short to
13925 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13926 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13927 or if measurement_type is not polynomial.
13928 (fixed_size_mode::includes_p): Check for constant-sized modes.
13929 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13930 return a poly_uint16 rather than an unsigned short.
13931 (emit_mode_size): Change the type of mode_size from unsigned short
13932 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13933 (emit_mode_adjustments): Cope with polynomial vector sizes.
13934 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13936 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13938 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13939 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13940 * caller-save.c (setup_save_areas): Likewise.
13941 (replace_reg_with_saved_mem): Likewise.
13942 * calls.c (emit_library_call_value_1): Likewise.
13943 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13944 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13945 (gen_lowpart_for_combine): Likewise.
13946 * convert.c (convert_to_integer_1): Likewise.
13947 * cse.c (equiv_constant, cse_insn): Likewise.
13948 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13949 (cselib_subst_to_values): Likewise.
13950 * dce.c (word_dce_process_block): Likewise.
13951 * df-problems.c (df_word_lr_mark_ref): Likewise.
13952 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13953 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13954 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13955 (rtl_for_decl_location): Likewise.
13956 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13957 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13958 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13959 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13960 (expand_expr_real_1): Likewise.
13961 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13962 (pad_below): Likewise.
13963 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13964 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13965 * ira.c (get_subreg_tracking_sizes): Likewise.
13966 * ira-build.c (ira_create_allocno_objects): Likewise.
13967 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13968 (ira_sort_regnos_for_alter_reg): Likewise.
13969 * ira-costs.c (record_operand_costs): Likewise.
13970 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13971 (resolve_simple_move): Likewise.
13972 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13973 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13974 (lra_constraints): Likewise.
13975 (CONST_POOL_OK_P): Reject variable-sized modes.
13976 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13977 (add_pseudo_to_slot, lra_spill): Likewise.
13978 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13979 * optabs-query.c (get_best_extraction_insn): Likewise.
13980 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13981 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13982 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13983 * recog.c (offsettable_address_addr_space_p): Likewise.
13984 * regcprop.c (maybe_mode_change): Likewise.
13985 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13986 * regrename.c (build_def_use): Likewise.
13987 * regstat.c (dump_reg_info): Likewise.
13988 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13989 (find_reloads, find_reloads_subreg_address): Likewise.
13990 * reload1.c (eliminate_regs_1): Likewise.
13991 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13992 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13993 (simplify_binary_operation_1, simplify_subreg): Likewise.
13994 * targhooks.c (default_function_arg_padding): Likewise.
13995 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13996 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13997 (verify_gimple_assign_ternary): Likewise.
13998 * tree-inline.c (estimate_move_cost): Likewise.
13999 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14000 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
14001 (get_address_cost_ainc): Likewise.
14002 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
14003 (vect_supportable_dr_alignment): Likewise.
14004 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14005 (vectorizable_reduction): Likewise.
14006 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
14007 (vectorizable_operation, vectorizable_load): Likewise.
14008 * tree.c (build_same_sized_truth_vector_type): Likewise.
14009 * valtrack.c (cleanup_auto_inc_dec): Likewise.
14010 * var-tracking.c (emit_note_insn_var_location): Likewise.
14011 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
14012 (ADDR_VEC_ALIGN): Likewise.
14014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14015 Alan Hayward <alan.hayward@arm.com>
14016 David Sherwood <david.sherwood@arm.com>
14018 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
14020 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
14021 or if measurement_type is polynomial.
14022 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
14023 * combine.c (make_extraction): Likewise.
14024 * dse.c (find_shift_sequence): Likewise.
14025 * dwarf2out.c (mem_loc_descriptor): Likewise.
14026 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
14027 (extract_bit_field, extract_low_bits): Likewise.
14028 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
14029 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
14030 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
14031 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
14032 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
14033 * reload.c (find_reloads): Likewise.
14034 * reload1.c (alter_reg): Likewise.
14035 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
14036 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
14037 * tree-if-conv.c (predicate_mem_writes): Likewise.
14038 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14039 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14040 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14041 * valtrack.c (dead_debug_insert_temp): Likewise.
14042 * varasm.c (mergeable_constant_section): Likewise.
14043 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14046 Alan Hayward <alan.hayward@arm.com>
14047 David Sherwood <david.sherwood@arm.com>
14049 * expr.c (expand_assignment): Cope with polynomial mode sizes
14050 when assigning to a CONCAT.
14052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14053 Alan Hayward <alan.hayward@arm.com>
14054 David Sherwood <david.sherwood@arm.com>
14056 * machmode.h (mode_precision): Change from unsigned short to
14058 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14060 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14061 or if measurement_type is not polynomial.
14062 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14063 in which the mode is already known to be a scalar_int_mode.
14064 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14065 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14067 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14068 for GET_MODE_PRECISION.
14069 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14070 for GET_MODE_PRECISION.
14071 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14073 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14074 (expand_field_assignment, make_extraction): Likewise.
14075 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14076 (get_last_value): Likewise.
14077 * convert.c (convert_to_integer_1): Likewise.
14078 * cse.c (cse_insn): Likewise.
14079 * expr.c (expand_expr_real_1): Likewise.
14080 * lra-constraints.c (simplify_operand_subreg): Likewise.
14081 * optabs-query.c (can_atomic_load_p): Likewise.
14082 * optabs.c (expand_atomic_load): Likewise.
14083 (expand_atomic_store): Likewise.
14084 * ree.c (combine_reaching_defs): Likewise.
14085 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14086 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14087 * tree.h (type_has_mode_precision_p): Likewise.
14088 * ubsan.c (instrument_si_overflow): Likewise.
14090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14091 Alan Hayward <alan.hayward@arm.com>
14092 David Sherwood <david.sherwood@arm.com>
14094 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14095 polynomial numbers of units.
14096 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14097 (valid_vector_subparts_p): New function.
14098 (build_vector_type): Remove temporary shim and take the number
14099 of units as a poly_uint64 rather than an int.
14100 (build_opaque_vector_type): Take the number of units as a
14101 poly_uint64 rather than an int.
14102 * tree.c (build_vector_from_ctor): Handle polynomial
14103 TYPE_VECTOR_SUBPARTS.
14104 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14105 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14106 (build_vector_from_val): If the number of units is variable,
14107 use build_vec_duplicate_cst for constant operands and
14108 VEC_DUPLICATE_EXPR otherwise.
14109 (make_vector_type): Remove temporary is_constant ().
14110 (build_vector_type, build_opaque_vector_type): Take the number of
14111 units as a poly_uint64 rather than an int.
14112 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14114 * cfgexpand.c (expand_debug_expr): Likewise.
14115 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14116 (store_constructor, expand_expr_real_1): Likewise.
14117 (const_scalar_mask_from_tree): Likewise.
14118 * fold-const-call.c (fold_const_reduction): Likewise.
14119 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14120 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14121 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14122 (fold_relational_const): Likewise.
14123 (native_interpret_vector): Likewise. Change the size from an
14124 int to an unsigned int.
14125 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14126 TYPE_VECTOR_SUBPARTS.
14127 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14128 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14129 duplicating a non-constant operand into a variable-length vector.
14130 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14131 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14132 * ipa-icf.c (sem_variable::equals): Likewise.
14133 * match.pd: Likewise.
14134 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14135 * print-tree.c (print_node): Likewise.
14136 * stor-layout.c (layout_type): Likewise.
14137 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14138 * tree-cfg.c (verify_gimple_comparison): Likewise.
14139 (verify_gimple_assign_binary): Likewise.
14140 (verify_gimple_assign_ternary): Likewise.
14141 (verify_gimple_assign_single): Likewise.
14142 * tree-pretty-print.c (dump_generic_node): Likewise.
14143 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14144 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14145 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14146 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14147 (vect_shift_permute_load_chain): Likewise.
14148 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14149 (expand_vector_condition, optimize_vector_constructor): Likewise.
14150 (lower_vec_perm, get_compute_type): Likewise.
14151 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14152 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14153 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14154 (vect_recog_mask_conversion_pattern): Likewise.
14155 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14156 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14157 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14158 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14159 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14160 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14161 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14162 (supportable_widening_operation): Likewise.
14163 (supportable_narrowing_operation): Likewise.
14164 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14166 * varasm.c (output_constant): Likewise.
14168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14169 Alan Hayward <alan.hayward@arm.com>
14170 David Sherwood <david.sherwood@arm.com>
14172 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14173 so that both the length == 3 and length != 3 cases set up their
14174 own permute vectors. Add comments explaining why we know the
14175 number of elements is constant.
14176 (vect_permute_load_chain): Likewise.
14178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14179 Alan Hayward <alan.hayward@arm.com>
14180 David Sherwood <david.sherwood@arm.com>
14182 * machmode.h (mode_nunits): Change from unsigned char to
14184 (ONLY_FIXED_SIZE_MODES): New macro.
14185 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14186 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14187 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14189 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14190 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14191 or if measurement_type is not polynomial.
14192 * genmodes.c (ZERO_COEFFS): New macro.
14193 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14195 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14196 Use ZERO_COEFFS when emitting initializers.
14197 * data-streamer.h (bp_pack_poly_value): New function.
14198 (bp_unpack_poly_value): Likewise.
14199 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14200 for GET_MODE_NUNITS.
14201 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14202 for GET_MODE_NUNITS.
14203 * tree.c (make_vector_type): Remove temporary shim and make
14204 the real function take the number of units as a poly_uint64
14205 rather than an int.
14206 (build_vector_type_for_mode): Handle polynomial nunits.
14207 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14208 * emit-rtl.c (const_vec_series_p_1): Likewise.
14209 (gen_rtx_CONST_VECTOR): Likewise.
14210 * fold-const.c (test_vec_duplicate_folding): Likewise.
14211 * genrecog.c (validate_pattern): Likewise.
14212 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14213 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14214 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14215 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14216 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14217 * rtlanal.c (subreg_get_info): Likewise.
14218 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14219 (vect_grouped_load_supported): Likewise.
14220 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14221 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14222 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14223 (simplify_const_unary_operation, simplify_binary_operation_1)
14224 (simplify_const_binary_operation, simplify_ternary_operation)
14225 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14226 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14227 instead of CONST_VECTOR_NUNITS.
14228 * varasm.c (output_constant_pool_2): Likewise.
14229 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14230 explicit-encoded elements in the XVEC for variable-length vectors.
14232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14234 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14237 Alan Hayward <alan.hayward@arm.com>
14238 David Sherwood <david.sherwood@arm.com>
14240 * coretypes.h (fixed_size_mode): Declare.
14241 (fixed_size_mode_pod): New typedef.
14242 * builtins.h (target_builtins::x_apply_args_mode)
14243 (target_builtins::x_apply_result_mode): Change type to
14244 fixed_size_mode_pod.
14245 * builtins.c (apply_args_size, apply_result_size, result_vector)
14246 (expand_builtin_apply_args_1, expand_builtin_apply)
14247 (expand_builtin_return): Update accordingly.
14249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14251 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14252 * cselib.c (cselib_hash_rtx): Likewise.
14253 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14254 CONST_VECTOR encoding.
14256 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14257 Jeff Law <law@redhat.com>
14260 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14261 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14262 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14263 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14266 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14267 explicitly probe *sp in a noreturn function if there were any callee
14268 register saves or frame pointer is needed.
14270 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14273 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14274 BLKmode for ternary, binary or unary expressions.
14277 * var-tracking.c (delete_vta_debug_insn): New inline function.
14278 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14279 insns from get_insns () to NULL instead of each bb separately.
14280 Use delete_vta_debug_insn. No longer static.
14281 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14282 delete_vta_debug_insns callers.
14283 * rtl.h (delete_vta_debug_insns): Declare.
14284 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14285 instead of variable_tracking_main.
14287 2018-01-03 Martin Sebor <msebor@redhat.com>
14289 PR tree-optimization/83603
14290 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14291 arguments past the endof the argument list in functions declared
14292 without a prototype.
14293 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14294 Avoid checking when arguments are null.
14296 2018-01-03 Martin Sebor <msebor@redhat.com>
14299 * doc/extend.texi (attribute const): Fix a typo.
14300 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14301 issuing -Wsuggest-attribute for void functions.
14303 2018-01-03 Martin Sebor <msebor@redhat.com>
14305 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14306 offset_int::from instead of wide_int::to_shwi.
14307 (maybe_diag_overlap): Remove assertion.
14308 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14309 * gimple-ssa-sprintf.c (format_directive): Same.
14310 (parse_directive): Same.
14311 (sprintf_dom_walker::compute_format_length): Same.
14312 (try_substitute_return_value): Same.
14314 2018-01-03 Jeff Law <law@redhat.com>
14316 PR middle-end/83654
14317 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14318 non-constant residual for zero at runtime and avoid probing in
14319 that case. Reorganize code for trailing problem to mirror handling
14322 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14324 PR tree-optimization/83501
14325 * tree-ssa-strlen.c (get_string_cst): New.
14326 (handle_char_store): Call get_string_cst.
14328 2018-01-03 Martin Liska <mliska@suse.cz>
14330 PR tree-optimization/83593
14331 * tree-ssa-strlen.c: Include tree-cfg.h.
14332 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14333 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14334 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14336 (strlen_dom_walker::before_dom_children): Call
14337 gimple_purge_dead_eh_edges. Dump tranformation with details
14339 (strlen_dom_walker::before_dom_children): Update call by adding
14340 new argument cleanup_eh.
14341 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14343 2018-01-03 Martin Liska <mliska@suse.cz>
14346 * cif-code.def (VARIADIC_THUNK): New enum value.
14347 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14350 2018-01-03 Jan Beulich <jbeulich@suse.com>
14352 * sse.md (mov<mode>_internal): Tighten condition for when to use
14353 vmovdqu<ssescalarsize> for TI and OI modes.
14355 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14357 Update copyright years.
14359 2018-01-03 Martin Liska <mliska@suse.cz>
14362 * ipa-visibility.c (function_and_variable_visibility): Skip
14363 functions with noipa attribure.
14365 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14367 * gcc.c (process_command): Update copyright notice dates.
14368 * gcov-dump.c (print_version): Ditto.
14369 * gcov.c (print_version): Ditto.
14370 * gcov-tool.c (print_version): Ditto.
14371 * gengtype.c (create_file): Ditto.
14372 * doc/cpp.texi: Bump @copying's copyright year.
14373 * doc/cppinternals.texi: Ditto.
14374 * doc/gcc.texi: Ditto.
14375 * doc/gccint.texi: Ditto.
14376 * doc/gcov.texi: Ditto.
14377 * doc/install.texi: Ditto.
14378 * doc/invoke.texi: Ditto.
14380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14382 * vector-builder.h (vector_builder::m_full_nelts): Change from
14383 unsigned int to poly_uint64.
14384 (vector_builder::full_nelts): Update prototype accordingly.
14385 (vector_builder::new_vector): Likewise.
14386 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14387 (vector_builder::operator ==): Likewise.
14388 (vector_builder::finalize): Likewise.
14389 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14390 Take the number of elements as a poly_uint64 rather than an
14392 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14393 from unsigned int to poly_uint64.
14394 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14395 (vec_perm_indices::new_vector): Likewise.
14396 (vec_perm_indices::length): Likewise.
14397 (vec_perm_indices::nelts_per_input): Likewise.
14398 (vec_perm_indices::input_nelts): Likewise.
14399 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14400 number of elements per input as a poly_uint64 rather than an
14401 unsigned int. Use the original encoding for variable-length
14402 vectors, rather than clamping each individual element.
14403 For the second and subsequent elements in each pattern,
14404 clamp the step and base before clamping their sum.
14405 (vec_perm_indices::series_p): Handle polynomial element counts.
14406 (vec_perm_indices::all_in_range_p): Likewise.
14407 (vec_perm_indices_to_tree): Likewise.
14408 (vec_perm_indices_to_rtx): Likewise.
14409 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14410 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14411 (tree_vector_builder::new_binary_operation): Handle polynomial
14412 element counts. Return false if we need to know the number
14413 of elements at compile time.
14414 * fold-const.c (fold_vec_perm): Punt if the number of elements
14415 isn't known at compile time.
14417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14419 * vec-perm-indices.h (vec_perm_builder): Change element type
14420 from HOST_WIDE_INT to poly_int64.
14421 (vec_perm_indices::element_type): Update accordingly.
14422 (vec_perm_indices::clamp): Handle polynomial element_types.
14423 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14424 (vec_perm_indices::all_in_range_p): Likewise.
14425 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14427 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14428 polynomial vec_perm_indices element types.
14429 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14430 * fold-const.c (fold_vec_perm): Likewise.
14431 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14432 * tree-vect-generic.c (lower_vec_perm): Likewise.
14433 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14434 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14435 element type to HOST_WIDE_INT.
14437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14438 Alan Hayward <alan.hayward@arm.com>
14439 David Sherwood <david.sherwood@arm.com>
14441 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14442 rather than an int. Use plus_constant.
14443 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14444 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14447 Alan Hayward <alan.hayward@arm.com>
14448 David Sherwood <david.sherwood@arm.com>
14450 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14451 a HOST_WIDE_INT to a poly_int64.
14453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14454 Alan Hayward <alan.hayward@arm.com>
14455 David Sherwood <david.sherwood@arm.com>
14457 * calls.c (load_register_parameters): Cope with polynomial
14458 mode sizes. Require a constant size for BLKmode parameters
14459 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14460 forces a parameter to be padded at the lsb end in order to
14461 fill a complete number of words, require the parameter size
14462 to be ordered wrt UNITS_PER_WORD.
14464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14465 Alan Hayward <alan.hayward@arm.com>
14466 David Sherwood <david.sherwood@arm.com>
14468 * reload1.c (spill_stack_slot_width): Change element type
14469 from unsigned int to poly_uint64_pod.
14470 (alter_reg): Treat mode sizes as polynomial.
14472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14473 Alan Hayward <alan.hayward@arm.com>
14474 David Sherwood <david.sherwood@arm.com>
14476 * reload.c (complex_word_subreg_p): New function.
14477 (reload_inner_reg_of_subreg, push_reload): Use it.
14479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14480 Alan Hayward <alan.hayward@arm.com>
14481 David Sherwood <david.sherwood@arm.com>
14483 * lra-constraints.c (process_alt_operands): Reject matched
14484 operands whose sizes aren't ordered.
14485 (match_reload): Refer to this check here.
14487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14488 Alan Hayward <alan.hayward@arm.com>
14489 David Sherwood <david.sherwood@arm.com>
14491 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14492 that the mode size is in the set {1, 2, 4, 8, 16}.
14494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14495 Alan Hayward <alan.hayward@arm.com>
14496 David Sherwood <david.sherwood@arm.com>
14498 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14499 Use plus_constant instead of gen_rtx_PLUS.
14501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14502 Alan Hayward <alan.hayward@arm.com>
14503 David Sherwood <david.sherwood@arm.com>
14505 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14506 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14507 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14508 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14509 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14510 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14511 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14512 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14513 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14514 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14516 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14517 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14518 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14519 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14520 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14521 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14522 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14523 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14524 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14525 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14527 * expr.c (emit_move_resolve_push): Treat the input and result
14528 of PUSH_ROUNDING as a poly_int64.
14529 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14530 (emit_push_insn): Likewise.
14531 * lra-eliminations.c (mark_not_eliminable): Likewise.
14532 * recog.c (push_operand): Likewise.
14533 * reload1.c (elimination_effects): Likewise.
14534 * rtlanal.c (nonzero_bits1): Likewise.
14535 * calls.c (store_one_arg): Likewise. Require the padding to be
14536 known at compile time.
14538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14539 Alan Hayward <alan.hayward@arm.com>
14540 David Sherwood <david.sherwood@arm.com>
14542 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14543 Use plus_constant instead of gen_rtx_PLUS.
14545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14546 Alan Hayward <alan.hayward@arm.com>
14547 David Sherwood <david.sherwood@arm.com>
14549 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14550 rather than an int.
14552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14553 Alan Hayward <alan.hayward@arm.com>
14554 David Sherwood <david.sherwood@arm.com>
14556 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14557 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14558 via stack temporaries. Treat the mode size as polynomial too.
14560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * expr.c (expand_expr_real_2): When handling conversions involving
14565 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14566 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14567 as a poly_uint64 too.
14569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14570 Alan Hayward <alan.hayward@arm.com>
14571 David Sherwood <david.sherwood@arm.com>
14573 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14576 Alan Hayward <alan.hayward@arm.com>
14577 David Sherwood <david.sherwood@arm.com>
14579 * combine.c (can_change_dest_mode): Handle polynomial
14580 REGMODE_NATURAL_SIZE.
14581 * expmed.c (store_bit_field_1): Likewise.
14582 * expr.c (store_constructor): Likewise.
14583 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14584 and polynomial REGMODE_NATURAL_SIZE.
14585 (gen_lowpart_common): Likewise.
14586 * reginfo.c (record_subregs_of_mode): Likewise.
14587 * rtlanal.c (read_modify_subreg_p): Likewise.
14589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14590 Alan Hayward <alan.hayward@arm.com>
14591 David Sherwood <david.sherwood@arm.com>
14593 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14594 numbers of elements.
14596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14597 Alan Hayward <alan.hayward@arm.com>
14598 David Sherwood <david.sherwood@arm.com>
14600 * match.pd: Cope with polynomial numbers of vector elements.
14602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14603 Alan Hayward <alan.hayward@arm.com>
14604 David Sherwood <david.sherwood@arm.com>
14606 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14607 in a POINTER_PLUS_EXPR.
14609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14610 Alan Hayward <alan.hayward@arm.com>
14611 David Sherwood <david.sherwood@arm.com>
14613 * omp-simd-clone.c (simd_clone_subparts): New function.
14614 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14615 (ipa_simd_modify_function_body): Likewise.
14617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14618 Alan Hayward <alan.hayward@arm.com>
14619 David Sherwood <david.sherwood@arm.com>
14621 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14622 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14623 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14624 (expand_vector_condition, vector_element): Likewise.
14625 (subparts_gt): New function.
14626 (get_compute_type): Use subparts_gt.
14627 (count_type_subparts): Delete.
14628 (expand_vector_operations_1): Use subparts_gt instead of
14629 count_type_subparts.
14631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14632 Alan Hayward <alan.hayward@arm.com>
14633 David Sherwood <david.sherwood@arm.com>
14635 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14636 (vect_compile_time_alias): ...this new function. Do the calculation
14637 on poly_ints rather than trees.
14638 (vect_prune_runtime_alias_test_list): Update call accordingly.
14640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14641 Alan Hayward <alan.hayward@arm.com>
14642 David Sherwood <david.sherwood@arm.com>
14644 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14646 (vect_schedule_slp_instance): Likewise.
14648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14649 Alan Hayward <alan.hayward@arm.com>
14650 David Sherwood <david.sherwood@arm.com>
14652 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14653 constant and extern definitions for variable-length vectors.
14654 (vect_get_constant_vectors): Note that the number of units
14655 is known to be constant.
14657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14658 Alan Hayward <alan.hayward@arm.com>
14659 David Sherwood <david.sherwood@arm.com>
14661 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14662 of units as polynomial. Choose between WIDE and NARROW based
14665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14666 Alan Hayward <alan.hayward@arm.com>
14667 David Sherwood <david.sherwood@arm.com>
14669 * tree-vect-stmts.c (simd_clone_subparts): New function.
14670 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14673 Alan Hayward <alan.hayward@arm.com>
14674 David Sherwood <david.sherwood@arm.com>
14676 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14677 vectors as polynomial. Use build_index_vector for
14678 IFN_GOMP_SIMD_LANE.
14680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14681 Alan Hayward <alan.hayward@arm.com>
14682 David Sherwood <david.sherwood@arm.com>
14684 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14685 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14686 for variable-length vectors.
14687 (vectorizable_mask_load_store): Treat the number of units as
14688 polynomial, asserting that it is constant if the condition has
14689 already been enforced.
14690 (vectorizable_store, vectorizable_load): Likewise.
14692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14693 Alan Hayward <alan.hayward@arm.com>
14694 David Sherwood <david.sherwood@arm.com>
14696 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14697 of units as polynomial. Punt if we can't tell at compile time
14698 which vector contains the final result.
14700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14701 Alan Hayward <alan.hayward@arm.com>
14702 David Sherwood <david.sherwood@arm.com>
14704 * tree-vect-loop.c (vectorizable_induction): Treat the number
14705 of units as polynomial. Punt on SLP inductions. Use an integer
14706 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14707 cast of such a series for variable-length floating-point
14710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14711 Alan Hayward <alan.hayward@arm.com>
14712 David Sherwood <david.sherwood@arm.com>
14714 * tree.h (build_index_vector): Declare.
14715 * tree.c (build_index_vector): New function.
14716 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14717 of units as polynomial, forcibly converting it to a constant if
14718 vectorizable_reduction has already enforced the condition.
14719 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14720 to create a {1,2,3,...} vector.
14721 (vectorizable_reduction): Treat the number of units as polynomial.
14722 Choose vectype_in based on the largest scalar element size rather
14723 than the smallest number of units. Enforce the restrictions
14726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14727 Alan Hayward <alan.hayward@arm.com>
14728 David Sherwood <david.sherwood@arm.com>
14730 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14731 number of units as polynomial.
14733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14734 Alan Hayward <alan.hayward@arm.com>
14735 David Sherwood <david.sherwood@arm.com>
14737 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14738 * target.def (autovectorize_vector_sizes): Return the vector sizes
14739 by pointer, using vector_sizes rather than a bitmask.
14740 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14741 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14742 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14744 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14745 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14746 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14747 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14748 * omp-general.c (omp_max_vf): Likewise.
14749 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14750 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14751 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14752 * tree-vect-slp.c (vect_slp_bb): Likewise.
14753 * doc/tm.texi: Regenerate.
14754 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14756 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14757 the vector size as a poly_uint64 rather than an unsigned int.
14758 (current_vector_size): Change from an unsigned int to a poly_uint64.
14759 (get_vectype_for_scalar_type): Update accordingly.
14760 * tree.h (build_truth_vector_type): Take the size and number of
14761 units as a poly_uint64 rather than an unsigned int.
14762 (build_vector_type): Add a temporary overload that takes
14763 the number of units as a poly_uint64 rather than an unsigned int.
14764 * tree.c (make_vector_type): Likewise.
14765 (build_truth_vector_type): Take the number of units as a poly_uint64
14766 rather than an unsigned int.
14768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14769 Alan Hayward <alan.hayward@arm.com>
14770 David Sherwood <david.sherwood@arm.com>
14772 * target.def (get_mask_mode): Take the number of units and length
14773 as poly_uint64s rather than unsigned ints.
14774 * targhooks.h (default_get_mask_mode): Update accordingly.
14775 * targhooks.c (default_get_mask_mode): Likewise.
14776 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14777 * doc/tm.texi: Regenerate.
14779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14780 Alan Hayward <alan.hayward@arm.com>
14781 David Sherwood <david.sherwood@arm.com>
14783 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14784 * omp-general.c (omp_max_vf): Likewise.
14785 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14786 (expand_omp_simd): Handle polynomial safelen.
14787 * omp-low.c (omplow_simd_context): Add a default constructor.
14788 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14789 (lower_rec_simd_input_clauses): Update accordingly.
14790 (lower_rec_input_clauses): Likewise.
14792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14793 Alan Hayward <alan.hayward@arm.com>
14794 David Sherwood <david.sherwood@arm.com>
14796 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14797 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14798 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14799 (vect_analyze_slp_cost): Likewise.
14800 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14801 (vect_model_load_cost): Likewise.
14803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14804 Alan Hayward <alan.hayward@arm.com>
14805 David Sherwood <david.sherwood@arm.com>
14807 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14808 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14809 from an unsigned int * to a poly_uint64_pod *.
14810 (calculate_unrolling_factor): New function.
14811 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14814 Alan Hayward <alan.hayward@arm.com>
14815 David Sherwood <david.sherwood@arm.com>
14817 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14818 from an unsigned int to a poly_uint64.
14819 (_loop_vec_info::slp_unrolling_factor): Likewise.
14820 (_loop_vec_info::vectorization_factor): Change from an int
14822 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14823 (vect_get_num_vectors): New function.
14824 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14825 (vect_get_num_copies): Use vect_get_num_vectors.
14826 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14827 to an unsigned int *.
14828 (vect_analyze_data_refs): Change min_vf from an int * to a
14830 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14831 than an unsigned HOST_WIDE_INT.
14832 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14833 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14834 to an unsigned int *.
14835 (vect_analyze_data_ref_dependences): Likewise.
14836 (vect_compute_data_ref_alignment): Handle polynomial vf.
14837 (vect_enhance_data_refs_alignment): Likewise.
14838 (vect_prune_runtime_alias_test_list): Likewise.
14839 (vect_shift_permute_load_chain): Likewise.
14840 (vect_supportable_dr_alignment): Likewise.
14841 (dependence_distance_ge_vf): Take the vectorization factor as a
14842 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14843 (vect_analyze_data_refs): Change min_vf from an int * to a
14845 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14846 vfm1 as a poly_uint64 rather than an int. Make the same change
14847 for the returned bound_scalar.
14848 (vect_gen_vector_loop_niters): Handle polynomial vf.
14849 (vect_do_peeling): Likewise. Update call to
14850 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14851 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14853 * tree-vect-loop.c (vect_determine_vectorization_factor)
14854 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14855 (vect_get_known_peeling_cost): Likewise.
14856 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14857 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14858 (vect_transform_loop): Likewise. Use the lowest possible VF when
14859 updating the upper bounds of the loop.
14860 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14861 rather than an int.
14862 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14863 polynomial unroll factors.
14864 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14865 (vect_make_slp_decision): Likewise.
14866 (vect_supported_load_permutation_p): Likewise, and polynomial
14868 (vect_analyze_slp_cost): Handle polynomial vf.
14869 (vect_slp_analyze_node_operations): Likewise.
14870 (vect_slp_analyze_bb_1): Likewise.
14871 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14872 than an unsigned HOST_WIDE_INT.
14873 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14874 (vectorizable_load): Handle polynomial vf.
14875 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14877 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14880 Alan Hayward <alan.hayward@arm.com>
14881 David Sherwood <david.sherwood@arm.com>
14883 * match.pd: Handle bit operations involving three constants
14884 and try to fold one pair.
14886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14888 * tree-vect-loop-manip.c: Include gimple-fold.h.
14889 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14890 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14891 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14892 Add a path that uses a step of VF instead of 1, but disable it
14894 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14895 and niters_no_overflow parameters. Update calls to
14896 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14897 Create a new SSA name if the latter choses to use a ste other
14898 than zero, and return it via niters_vector_mult_vf_var.
14899 * tree-vect-loop.c (vect_transform_loop): Update calls to
14900 vect_do_peeling, vect_gen_vector_loop_niters and
14901 slpeel_make_loop_iterate_ntimes.
14902 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14903 (vect_gen_vector_loop_niters): Update declarations after above changes.
14905 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14907 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14908 128-bit round to integer instructions.
14909 (ceil<mode>2): Likewise.
14910 (btrunc<mode>2): Likewise.
14911 (round<mode>2): Likewise.
14913 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14915 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14916 unaligned VSX load/store on P8/P9.
14917 (expand_block_clear): Allow the use of unaligned VSX
14918 load/store on P8/P9.
14920 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14922 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14924 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14925 swap associated with both a load and a store.
14927 2018-01-02 Andrew Waterman <andrew@sifive.com>
14929 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14930 * config/riscv/riscv.md (clear_cache): Use it.
14932 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14934 * web.c: Remove out-of-date comment.
14936 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14938 * expr.c (fixup_args_size_notes): Check that any existing
14939 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14940 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14941 (emit_single_push_insn): ...here.
14943 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14945 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14946 (const_vector_encoded_nelts): New function.
14947 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14948 (const_vector_int_elt, const_vector_elt): Declare.
14949 * emit-rtl.c (const_vector_int_elt_1): New function.
14950 (const_vector_elt): Likewise.
14951 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14952 of CONST_VECTOR_ELT.
14954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14956 * expr.c: Include rtx-vector-builder.h.
14957 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14958 directly on the tree encoding.
14959 (const_vector_from_tree): Likewise.
14960 * optabs.c: Include rtx-vector-builder.h.
14961 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14962 sequence of "u" values.
14963 * vec-perm-indices.c: Include rtx-vector-builder.h.
14964 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14965 directly on the vec_perm_indices encoding.
14967 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14969 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14970 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14971 * rtx-vector-builder.h: New file.
14972 * rtx-vector-builder.c: Likewise.
14973 * rtl.h (rtx_def::u2): Add a const_vector field.
14974 (CONST_VECTOR_NPATTERNS): New macro.
14975 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14976 (CONST_VECTOR_DUPLICATE_P): Likewise.
14977 (CONST_VECTOR_STEPPED_P): Likewise.
14978 (CONST_VECTOR_ENCODED_ELT): Likewise.
14979 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14980 (unwrap_const_vec_duplicate): Likewise.
14981 (const_vec_series_p): Check for a non-duplicated vector encoding.
14982 Say that the function only returns true for integer vectors.
14983 * emit-rtl.c: Include rtx-vector-builder.h.
14984 (gen_const_vec_duplicate_1): Delete.
14985 (gen_const_vector): Call gen_const_vec_duplicate instead of
14986 gen_const_vec_duplicate_1.
14987 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14988 (gen_const_vec_duplicate): Use rtx_vector_builder.
14989 (gen_const_vec_series): Likewise.
14990 (gen_rtx_CONST_VECTOR): Likewise.
14991 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14992 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14993 Build a new vector rather than modifying a CONST_VECTOR in-place.
14994 (handle_special_swappables): Update call accordingly.
14995 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14996 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14997 Build a new vector rather than modifying a CONST_VECTOR in-place.
14998 (handle_special_swappables): Update call accordingly.
15000 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15002 * simplify-rtx.c (simplify_const_binary_operation): Use
15003 CONST_VECTOR_ELT instead of XVECEXP.
15005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15007 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
15008 the selector elements to be different from the data elements
15009 if the selector is a VECTOR_CST.
15010 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
15011 ssizetype for the selector.
15013 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15015 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
15016 before testing each element individually.
15017 * tree-vect-generic.c (lower_vec_perm): Likewise.
15019 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15021 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
15022 * selftest-run-tests.c (selftest::run_tests): Call it.
15023 * vector-builder.h (vector_builder::operator ==): New function.
15024 (vector_builder::operator !=): Likewise.
15025 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
15026 (vec_perm_indices::all_from_input_p): New function.
15027 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15028 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
15029 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
15030 instead of reading the VECTOR_CST directly. Detect whether both
15031 vector inputs are the same before constructing the vec_perm_indices,
15032 and update the number of inputs argument accordingly. Use the
15033 utility functions added above. Only construct sel2 if we need to.
15035 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15037 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15038 the broadcast of the low byte.
15039 (expand_mult_highpart): Use an explicit encoding for the permutes.
15040 * optabs-query.c (can_mult_highpart_p): Likewise.
15041 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15042 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15043 (vectorizable_bswap): Likewise.
15044 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15045 explicit encoding for the power-of-2 permutes.
15046 (vect_permute_store_chain): Likewise.
15047 (vect_grouped_load_supported): Likewise.
15048 (vect_permute_load_chain): Likewise.
15050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15052 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15053 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15054 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15055 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15056 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15057 (vect_gen_perm_mask_any): Likewise.
15059 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15061 * int-vector-builder.h: New file.
15062 * vec-perm-indices.h: Include int-vector-builder.h.
15063 (vec_perm_indices): Redefine as an int_vector_builder.
15064 (auto_vec_perm_indices): Delete.
15065 (vec_perm_builder): Redefine as a stand-alone class.
15066 (vec_perm_indices::vec_perm_indices): New function.
15067 (vec_perm_indices::clamp): Likewise.
15068 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15069 (vec_perm_indices::new_vector): New function.
15070 (vec_perm_indices::new_expanded_vector): Update for new
15071 vec_perm_indices class.
15072 (vec_perm_indices::rotate_inputs): New function.
15073 (vec_perm_indices::all_in_range_p): Operate directly on the
15074 encoded form, without computing elided elements.
15075 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15076 encoding. Update for new vec_perm_indices class.
15077 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15078 the given vec_perm_builder.
15079 (expand_vec_perm_var): Update vec_perm_builder constructor.
15080 (expand_mult_highpart): Use vec_perm_builder instead of
15081 auto_vec_perm_indices.
15082 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15083 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15084 or double series encoding as appropriate.
15085 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15086 vec_perm_indices instead of auto_vec_perm_indices.
15087 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15088 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15089 (vect_permute_store_chain): Likewise.
15090 (vect_grouped_load_supported): Likewise.
15091 (vect_permute_load_chain): Likewise.
15092 (vect_shift_permute_load_chain): Likewise.
15093 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15094 (vect_transform_slp_perm_load): Likewise.
15095 (vect_schedule_slp_instance): Likewise.
15096 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15097 (vectorizable_mask_load_store): Likewise.
15098 (vectorizable_bswap): Likewise.
15099 (vectorizable_store): Likewise.
15100 (vectorizable_load): Likewise.
15101 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15102 vec_perm_indices instead of auto_vec_perm_indices. Use
15103 tree_to_vec_perm_builder to read the vector from a tree.
15104 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15105 vec_perm_builder instead of a vec_perm_indices.
15106 (have_whole_vector_shift): Use vec_perm_builder and
15107 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15108 truncation to calc_vec_perm_mask_for_shift.
15109 (vect_create_epilog_for_reduction): Likewise.
15110 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15111 from auto_vec_perm_indices to vec_perm_indices.
15112 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15113 instead of changing individual elements.
15114 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15115 the vector in d.perm.
15116 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15117 from auto_vec_perm_indices to vec_perm_indices.
15118 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15119 instead of changing individual elements.
15120 (arm_vectorize_vec_perm_const): Use new_vector to install
15121 the vector in d.perm.
15122 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15123 Update vec_perm_builder constructor.
15124 (rs6000_expand_interleave): Likewise.
15125 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15126 (rs6000_expand_interleave): Likewise.
15128 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15130 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15131 to qimode could truncate the indices.
15132 * optabs.c (expand_vec_perm_var): Likewise.
15134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15136 * Makefile.in (OBJS): Add vec-perm-indices.o.
15137 * vec-perm-indices.h: New file.
15138 * vec-perm-indices.c: Likewise.
15139 * target.h (vec_perm_indices): Replace with a forward class
15141 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15142 * optabs.h: Include vec-perm-indices.h.
15143 (expand_vec_perm): Delete.
15144 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15145 (expand_vec_perm_const): Declare.
15146 * target.def (vec_perm_const_ok): Replace with...
15147 (vec_perm_const): ...this new hook.
15148 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15149 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15150 * doc/tm.texi: Regenerate.
15151 * optabs.def (vec_perm_const): Delete.
15152 * doc/md.texi (vec_perm_const): Likewise.
15153 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15154 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15155 expand_vec_perm for constant permutation vectors. Assert that
15156 the mode of variable permutation vectors is the integer equivalent
15157 of the mode that is being permuted.
15158 * optabs-query.h (selector_fits_mode_p): Declare.
15159 * optabs-query.c: Include vec-perm-indices.h.
15160 (selector_fits_mode_p): New function.
15161 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15162 is defined, instead of checking whether the vec_perm_const_optab
15163 exists. Use targetm.vectorize.vec_perm_const instead of
15164 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15165 fit in the vector mode before using a variable permute.
15166 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15167 vec_perm_indices instead of an rtx.
15168 (expand_vec_perm): Replace with...
15169 (expand_vec_perm_const): ...this new function. Take the selector
15170 as a vec_perm_indices rather than an rtx. Also take the mode of
15171 the selector. Update call to shift_amt_for_vec_perm_mask.
15172 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15173 Use vec_perm_indices::new_expanded_vector to expand the original
15174 selector into bytes. Check whether the indices fit in the vector
15175 mode before using a variable permute.
15176 (expand_vec_perm_var): Make global.
15177 (expand_mult_highpart): Use expand_vec_perm_const.
15178 * fold-const.c: Includes vec-perm-indices.h.
15179 * tree-ssa-forwprop.c: Likewise.
15180 * tree-vect-data-refs.c: Likewise.
15181 * tree-vect-generic.c: Likewise.
15182 * tree-vect-loop.c: Likewise.
15183 * tree-vect-slp.c: Likewise.
15184 * tree-vect-stmts.c: Likewise.
15185 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15187 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15188 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15189 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15190 (aarch64_vectorize_vec_perm_const): ...this new function.
15191 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15192 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15193 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15194 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15195 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15196 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15197 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15199 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15200 check for NEON modes.
15201 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15202 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15203 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15204 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15206 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15207 the old VEC_PERM_CONST conditions.
15208 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15209 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15210 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15211 (ia64_vectorize_vec_perm_const_ok): Merge into...
15212 (ia64_vectorize_vec_perm_const): ...this new function.
15213 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15214 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15215 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15216 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15217 * config/mips/mips.c (mips_expand_vec_perm_const)
15218 (mips_vectorize_vec_perm_const_ok): Merge into...
15219 (mips_vectorize_vec_perm_const): ...this new function.
15220 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15221 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15222 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15223 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15224 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15225 (rs6000_expand_vec_perm_const): Delete.
15226 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15228 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15229 (altivec_expand_vec_perm_const_le): Take each operand individually.
15230 Operate on constant selectors rather than rtxes.
15231 (altivec_expand_vec_perm_const): Likewise. Update call to
15232 altivec_expand_vec_perm_const_le.
15233 (rs6000_expand_vec_perm_const): Delete.
15234 (rs6000_vectorize_vec_perm_const_ok): Delete.
15235 (rs6000_vectorize_vec_perm_const): New function.
15236 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15237 an element count and rtx array.
15238 (rs6000_expand_extract_even): Update call accordingly.
15239 (rs6000_expand_interleave): Likewise.
15240 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15241 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15242 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15243 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15244 (rs6000_expand_vec_perm_const): Delete.
15245 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15246 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15247 (altivec_expand_vec_perm_const_le): Take each operand individually.
15248 Operate on constant selectors rather than rtxes.
15249 (altivec_expand_vec_perm_const): Likewise. Update call to
15250 altivec_expand_vec_perm_const_le.
15251 (rs6000_expand_vec_perm_const): Delete.
15252 (rs6000_vectorize_vec_perm_const_ok): Delete.
15253 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15254 reference to the SPE evmerge intructions.
15255 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15256 an element count and rtx array.
15257 (rs6000_expand_extract_even): Update call accordingly.
15258 (rs6000_expand_interleave): Likewise.
15259 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15260 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15262 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15264 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15266 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15267 vector mode and that that mode matches the mode of the data
15269 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15270 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15271 directly using expand_vec_perm_1 when forcing selectors into
15273 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15277 * optabs-query.h (can_vec_perm_p): Delete.
15278 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15279 * optabs-query.c (can_vec_perm_p): Split into...
15280 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15281 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15282 particular selector is valid.
15283 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15284 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15285 (vect_grouped_load_supported): Likewise.
15286 (vect_shift_permute_load_chain): Likewise.
15287 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15288 (vect_transform_slp_perm_load): Likewise.
15289 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15290 (vectorizable_bswap): Likewise.
15291 (vect_gen_perm_mask_checked): Likewise.
15292 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15293 implementations of variable permutation vectors into account
15294 when deciding which selector to use.
15295 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15296 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15297 with a false third argument.
15298 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15299 to test whether the constant selector is valid and can_vec_perm_var_p
15300 to test whether a variable selector is valid.
15302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15304 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15305 * optabs-query.c (can_vec_perm_p): Likewise.
15306 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15307 instead of vec_perm_indices.
15308 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15309 (vect_gen_perm_mask_checked): Likewise,
15310 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15311 (vect_gen_perm_mask_checked): Likewise,
15313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15315 * optabs-query.h (qimode_for_vec_perm): Declare.
15316 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15317 (qimode_for_vec_perm): ...this new function.
15318 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15320 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15322 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15323 does not have a conditional at the top.
15325 2018-01-02 Richard Biener <rguenther@suse.de>
15327 * ipa-inline.c (big_speedup_p): Fix expression.
15329 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15332 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15335 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15339 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15340 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15341 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15342 cond_taken_branch_cost 3->4.
15344 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15346 PR tree-optimization/83581
15347 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15348 TODO_cleanup_cfg if any changes have been made.
15350 PR middle-end/83608
15351 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15352 convert_modes if target mode has the right side, but different mode
15355 PR middle-end/83609
15356 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15357 last argument when extracting from CONCAT. If either from_real or
15358 from_imag is NULL, use expansion through memory. If result is not
15359 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15360 the parts directly to inner mode, if even that fails, use expansion
15363 PR middle-end/83623
15364 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15365 check for bswap in mode rather than HImode and use that in expand_unop
15368 Copyright (C) 2018 Free Software Foundation, Inc.
15370 Copying and distribution of this file, with or without modification,
15371 are permitted in any medium without royalty provided the copyright
15372 notice and this notice are preserved.